Let us sing a suicide song to Intel's largest revenue p

>>59227118
>not even 60 fps

File: AMD-Ryzen-7-1700X-NAMD-Benchmark.jpg (71KB, 855x566px) Image search: [Google]

71KB, 855x566px

>>59227118
Some more production benches from the OP article.

>>59227132
>source: my jew overlords

File: AMD-Ryzen-7-1700X-7zip-Compression-Benchmark.jpg (90KB, 858x767px) Image search: [Google]

90KB, 858x767px

>>59227139

>>59227139
Dual Naples Ryzen will obliterate the Xeons.

File: AMD-Ryzen-7-1700X-OpenSSL-Sign-Benchmark.jpg (77KB, 858x586px) Image search: [Google]

77KB, 858x586px

Insane.

>>59227118
LOL, AMD just launched their naples for mainstream instead of summit ridge? WTF are those numbers!

I'm glad people are starting to see Ryzen actual potential and how it isn't and won't be games for the next few years.
It's a great CPU it just got marketed wrong for some odd reason

File: AMD-Ryzen-7-1700X-OpenSSL-Verify-Benchmark-800x551.jpg (69KB, 800x551px) Image search: [Google]

69KB, 800x551px

>>59227163
Verify

File: AMD-Ryzen-7-1700X-Sysbench-Single-Threaded.jpg (73KB, 858x489px) Image search: [Google]

73KB, 858x489px

In ST the Xeons literally have no chance even with Turbo Boost 3.0 pimping one core to its max.

>>59227174
>My early thoughts were that getting on and off chip with Ryzen was exhibiting some sort of intrinsic round-trip latency or bug, since when tasked with pure compute workloads Ryzen is so strong versus significantly more expensive Intel chips. Others thought Ryzen was having issues with its power states, dropping into a lower power mode inadvertently under this test condition, indicative perhaps of a BIOS firmware issue. Regardless, the more we kicked this around in the community, along with some of wide swings in performance seen between motherboard vendors and version, the more this began to feel, in my opinion, like typical early silicon architecture growing pains.

I think he have a point. It was the same with Nehalem and Sandy Bridge

https://www.forbes.com/sites/davealtavilla/2017/03/03/misinformed-rush-to-judgement-casts-amd-ryzen-in-overly-critical-light/#5e8abcc913e2

Basically AMD destroyed X99. Who would use it now that they can have a whole system by the price of just the 6950X.

File: Screenshot from 04-03-2017 03:36:48.png (75KB, 751x485px) Image search: [Google]

75KB, 751x485px

>>59227118
The final words say it all, a failed 'gamer' chip will demolish Intel's bottom line.

>>59227174

I had stock and managed to sell it before launch for a decent profit. There hasn't been a launch for a tech product that hasn't had hiccups. Do people forget x99 and the first i series?
The rubes sold stock and it dropped like a bit, I'm going to try and maybe buy back in next week if it drops some more.

Zen right now is a server/HPC arch first and foremost, being sold as mainstream. I've seen a ton of HPC/datacenter guys lose their shit over Ryzen. Have you seen the fucking power draw numbers? How it scales with lower frequency? Straight up performance? PRICE? This motherfucking chip is a mobile and HPC MONSTER. Naples is going to fucking Intel sideways.

If you think the kinds of datacenters that use that kind of Intel hardware are just going to dump it all and refit with AMD hardware top to bottom you really don't have a clue about how business computing infrastructure works.

But that's not really surprising, considering you're posting on 4chan, right?

Right?

A lot of these benefit from AMD's fantastic SMT implementation, once you scale 32 cores to this it will be not pretty for Intel.

>>59227289
TCO is everything, I hope you're aware of that.

This might be the biggest marketing fuck up by AMD ever. Targeting the R7s towards the gamer demographic only caused people to focus in even more on gaming performance, even though they were never going to beat Intel on IPC.

But they have on the market 8c/16t processors that are less than half the price and offer comparable performance to Intel's 8c/16t processors. Intel got a reprieve with Ryzen's shaky debut today, but the fuckers are almost certainly shitting their pants.

The future is bright.

>>59227289
I think you also do not.

File: Screenshot from 04-03-2017 03:47:54.png (54KB, 1489x889px) Image search: [Google]

54KB, 1489x889px

>>59227312
Nah, it's not a fuck up, this will get them to fix any teething BIOS/firmware and software scheduler issues before they release it to enterprise customers, who already had samples in testing for months.

If these upcoming BIOS issues fail to address the performance disparity in gaming, these CPUs won't be all that good a buy for gaming, but still fantastic productivity chips, and considering the revenue in enterprise AMD might as well consider the gamer market chump change, they'll design a high clocking arch in the future, but for now Zen doesn't scale all that well over 3.8 and needs massive voltages to do it, but scales down fantastic and that benefits everything but the gamer market.

>>59227139
That Xeon-D 1587 is a 16c/32t part.
Ryzen 1700X is right on its ass.
Goddamn.

>>59227382
>ryzen does more with less
I hope that this means cheap V3 Xeons on eBay.
Oh, and of course, the demise of jewtel.

>>59227118
DUAL
FUCKING BROADWELL 10 CORE XEONS

QUAD FUCKING XEONS

AMD WHAT HAVE YOU DONE
Also who are these guys that they have literally $100,000 (without mobo/ram cost) in chips lying around?

>>59227118
Hoooooooooooooooooooooooooooly shit.

Intel on suicide watch, guaranteed the CEO doesn't make it out of Q3.

>>59227362
I guess the question then is how much the clock ceilings are a product of Glofo's underlying process, as opposed to the heat generated by 8 full cores.

Though once EUV stops being vaporware, AMD might be able to get a handy boost from Glofo moving to 7nm. And given Intel's struggles with 10nm, the foundries overtaking Intel might be a legit possibility.

AMD is back baby, AMD IS FUCKING BACK.

Even during K7/K8 AMD never had this much of a competitive server core.

lol zen is amazing

File: 14885914166612.jpg (153KB, 885x1132px) Image search: [Google]

153KB, 885x1132px

>>59227118
It gets way better.
It absolutely decimates in performance/watt

>the closest thing to it is a i3 dual core

>>59227276
I managed to sell mine at 14.9. bought it back today at 13.12. It may dip more, but I wanted to get back in for upcoming servers and APUs.

Time to buy amd stock while it drop

>>59227491
That's beautiful.
Should I wait for Naples or just start to build my compile farm with R7 1700Xs?

>>59227418
>STH provides insights into best practices and hardware reviews for SMB servers, web hosting and virtualization labs. It is part of Loyolan Ventures, LLC.

>>59227515

I was gonna buy today before close but I never got around to it, I think it'll go up on Monday once people have had time to digest a little bit. Or not, some of the stock guys can be pretty dumb sometimes.

>>59227491
>6 core APUs with truckload of SATA ports for my pfsense router.

God dammit when.

WHEN I'M LITERALLY DROOLING
HECK I DON'T EVEN NEED THE APU JUST GIVE ME A 25W 8 CORE CPU WITH HIGH SINGLE THREAD TURBO SO I CAN SHOVE IT INTO MY NAS AND PFSENSE BOX AND RUN IT HEADLESS FUCK AMD YOU HAVE ONEJOB

>>59227587
Well, fan stock buyers mess up the price more so than actual traders. It's why stocks of companies like AMD perform kinda weird compared to others.

>>59227558

I would get just get 1700s imo. Looks like they can OC just as high at the Xs and pretty much the same perf and TDP profile.

If you can wait, I'd wait. I mean it's not like you couldn't sell the systems for 80% of the price you payed either, in a couple months down the line.

>>59227587
Efficient market hypothesis is a fucking meme. It'll take a while for the news to percolate down to most investors. In any case, AMD is coming off a ridiculous hype bubble, it'll probably come down somewhat as people who bought in at the single digits start to cash in.

I'll wait till there's either a clear recovery or until it starts getting close to AMD's Q2 earnings.

>>59227609
Wait for Naples.

>>59227609

25W might be too low for the 8 but who knows? Some guys over on Anandtech were doing some low power/frequency tests and Ryzen was BTFO of everything. I think they got it down to 35-45W but they didn't try particularly hard either.

>>59227558
Unfortunately you're not gonna get more than 8 cores in consumer market, nor are you gonna get two sockets.
So if 8 cores with ECC (no registered) support are gonna cut it for performance, the 1700's are the most efficient and offer the best value.

But first wait for the BIOS/fw/microcode updates.
They might even increase multithreaded perf or lower power, at least that's what happened with MSI's newest BIOS

>>59227665

I'm going to have options or hold some shares for Q1. There is no way in hell that balance sheet will not look good for AMD. Q2 is going to be a fucking riot when the APUs and HPC stuff comes out. Vega 2H. Oh fuck.

https://twitter.com/BitsAndChipsEng/status/837429476237651968

IT's over

>>59227445
I think it is the TDP, the 7700K goes 5ghz with 4 cores on 95W, Ryzen 7 goes 4.0 on 95 with double the cores. Maybe if it were 140W it would go higher

>>59227558
Go 1700 non X. energy cost will be lower.

>>59227818
Zen's sweet spot is 3300MHz, it needs insane voltage to pump over 4.0GHz, this is mostly a fab and design decision.

AMD made a chip that can scale voltage really low, an obvious server characteristic.

>>59227873
Sorry if some of us have more than a layman's knowledge of EE, faggot.

>>59227873
You're smelling something else.

>>59227885
So quick to disparage. Team Red+ finally found a talking point, cute but even ShariaBlue does it better.

>>59227917
There has been threads like these way more before, in case you missed the last 24 hours it's mostly just been game shitposting.

>>59227917
Believe it or not there are plenty of people here who didn't drop out of college during their first year of CS. Ryzen's launch was the most anticipated launch is recent history so of course its going to draw a lot of attention.

>>59227917
>a thread not oriented towards consumertards is now sus

Were you even around a few years ago? /g/ still had the occasional RE thread (reverse, PBUH) and other decent conversations every now and then.

Of course, now it's almost all consumerfaggotry and generals.

>>59227917
This isn't really some high level discussion at all, it just brought up memories of a more innocent time on this board where not everyone was underage.

There's some nice talk here too >>59226509
But still this is okay-tier at best, if this was difficult then reading RWT must fry your brain.

>>59227970
The 10+ cellphone threads per day really has shitted up the place.
Low brow shallow discussion of meaningless mobile web browsers is so commonplace that discussion of low cost server/workstation hardware is now out of the ordinary.

Can't wait to replace my sandybridge-e bullshit. Waiting for ecc am4 boards.

>>59227118
Intel literally on suicide watch

This fucking destroys them

>>59228009
It wouldn't be so bad if they could just keep it in their containment generals. But no, fucking everyone has to make a new thread whenever someone needs a new smartphone/tablet/laptop.

/g/ should have been split into consumer/g/ and tech/g/.

>>59228097
/g/ should have enforced self-moderation like old /g/ and /a/ did years ago and this wouldn't have happened, the gamergame nonsense certainly didn't help.

>>59227118
>*1* processor
>beating 2-4 processors
>not even the highest clocked variant
WHAT

>>59228138
Zen cores have a insanely powerful FPU

>>59227174
>>59227276
>>59227207
Doesn't really make sense to go for an 8 core Ryzen for gaming, the R5 series which will have 6 cores and probably higher clock speeds should be for the gaming market.

File: 1444020472351.jpg (29KB, 439x480px) Image search: [Google]

29KB, 439x480px

>>59228138
>Tfw dual 2680 v1s.

>>59227139
Just a reminder that NAMD can scale over 500k cores, yes, you heard me, five hundred thousand cores.
And zen is not far off from a dual xeon broadwell ( $650 each) with a total of 20 cores while still on a buggy BIOS.


If this doesn't tell you something, especially about AMD's SMT efficiency, I don't know what will.

File: CPU-Z.png (28KB, 407x400px) Image search: [Google]

28KB, 407x400px

>>59228221
>TFW I have an E5-2690 v1
It shitstomps the fuck out of it. I want one.

>>59228227
muh gayms

>>59227917
Go back to >>>/v/ adults are talking here.

>>59228227
>five hundred thousand cores.
>check top 500
>computahs with 40k CPUs with 16 cores

What a world we live in, and we're still shitposting about 4 cores.

>>59228285
100,000 socket clusters in the 2020s.
Get ready for it.

>>59228227
This means that the future is bright to AMD.
Could they pull something like Larrabe with Zen cores in the future?
I want to see that.

>>59227362
Ill toss my two cents in here.

I got a quad core when it was retarded, I watched the you bought threads you retard, I watched the single core king dual core is shit...

Having watched them all, we finally have a cpu that is worth a damn that is in the 6+core range, while being relatively cheap to boot. we are going to see the same shit again, it clocks lower, its not the best, the people with it are happy, the people who have 4 cores or only look at steril benchmarks will swear by 4 cores.

my only real consideration is what one should i get, i'm stuck between the 1700 because if I oc it, everything kind of has the same bottleneck area, or get the 1800x and have guaranteed clocks. I'm waiting on a bit more info from people who overclock to make a decision.

probably getting a noctua d15 regardless of cpu.

>>59228304
Man you can build a few houses from all those motherboards, yikes.

>>59228308
An accelerator? Doubt it unless XeonPhi's model gets used more than regular GPUs, which I doubt.

>>59228308
Not with Zen cores, they're too big. But if they refine their Cat cores and lean out the design they could. They could also pull on their GPU experience to help with that.

Don't see why they should though, AMD's GPUs are pretty solid from a compute standpoint as it is.

>>59228247
That thing must have compiled gentoo like lightning when it came out.

>>59228333
Only to troll Intel IRL.

>>59228308
amd already has gpus, larrabe was made because intel wanted to make an x86 gpu, failed to make a gpu compelling, but found a new use for many low power but smarter then gpu cores.

>>59228342
they already did that with with athletes when they advertised how fast intel would need to be to compete in the number.

>>59228323
It really doesn't. Most tasks are either embarrassingly parallel, in which case you can just use a GPU anyways, or they're serial-bound, in which case you just use a CPU.

There's not much that really fits the use case of "I need less cores than a GPU but they're all shitty P5 cores" that isn't filled by either existing CPUs or GPUs.

>tfw I've been pointing to the server market as the real place AMD will make waves
>everyone was still going on about the ever shrinking enthusiast consumer market
>tfw Vega is going to be the same story

How could it be any less obvious what these parts were intended for? Up until recently almost all conferences and release notes have been heavily targeted at enterprise workloads with gaming being more or less supplemental. The money to be made from the .1% of people buying top of the line hardware for gaming is nonexistent compared to businesses who spend millions of dollars annually on getting the highest performance per watt parts available. MCM Naples are going to fuck shit up big time.

>>59227917
Sometimes the people who aren't retarded on /g/ have a thread ;)

>>59227118

>please intel fail despite competently holding the title for over a decade
>thinks Intel gives a fuck about the desktop market and hasn't got new processors lined up to control and dominate the server market

You seem to forget that AMD is the one always playing catch up.

Intel already has all these contracts in place and Ryzen has yet to prove its stability or reliability. People don't switch overnight. They wait a year and then switch if there's a reasonable profit margin.

That's why nearly every office and business computer maintained Windows 98 until recently despite it being worse than XP or 7 by huge margins because it wasn't worth rolling out hundreds of thousands of copies of new software and potentially thousands of new base units for a tiny performance increase.

Servers are exactly the same. The only people watching Ryzen with absolute interest are video editors and idiots who want the next best desktop CPU despite not using even 10% of their processors power 90% of the time and thus an idiot who could have saves £200 by buying an i3 for the same jobs.

So, in resume
>the cubs won
>trump won
>AMD is competitive again, and Intel is afraid
This is the best timeline, lads.

>>59228371
There's probably not much of a case for that I agree, but I'm wondering what kind of edge cases could the corelet's do good in? All I know is that GPU cores are small, fast and dumb as rocks while P5 cores are still CPU cores to a degree, so somewhat smart.

>>59228227
>SMT efficiency
B-But it leads to buggy single thread performance

>>59228336
Possibly. It was worth about $2k when it hit shelves. They can be had for $200 on ebay now. Boards a shitfuck expensive though due to the glut of chips.

>>59228342
But why though? Like what >>59228371 and >>59228343 said, AMD already has silicon that can do the job already. High serial workloads can be dumped on Ryzen, High parallel workloads can be dumped on GCN. AMD's limited R&D can be put to better uses.

>>59227196
Naples will probably be 3.3ghz boost at the highest. But that'd still be really good single core perf. Still fastest on that list.

Did AMD release R7 before naples so gaymurs can help them debug their microcode??!?

>>59228422
In games, not actual CPU single threaded benches.

>>59228436
Yup.

>>59228427
I was daydreaming lads. You're right, it isn't needed.

>>59228436
At this point I'd be surprised if that wasn't the intention behind this rather unusual launch.

>>59228402
Can you explain how Vega is targeting the server market?

I suppose i can edit 4k videos without a problem with this cpu, right?

>>59228436
Reminder that the Fury was nothing more than a test bed for HBM and silicon interposers.

>>59228518
Sorry not server. Actually sorry for how poorly written that post was in general.

Vega will be great for GPGPU applications and CAD workloads. AMD wants to dethrone Nvidia as kings of machine learning.

>>59228524
It will surprise me if they choke on 8k.

>>59228436
Getting the chip into the hands of the faggot gaymers will let AMD identify issues and fix firmware and software before the more important Naples release, AMD doesn't have a million employees like Intel to testrun them and debug in every single application on the planet, they let us do it.

>>59228518
High perf/watt FP16 focused design with truckloads of bandwidth, also architecture slides.

File: POTE laugh.jpg (94KB, 1180x842px) Image search: [Google]

94KB, 1180x842px

>>59228576
>make those who critize you doing your bidding

>>59228576
Grandmaster Lisa su

>>59228594
>>59228414

Why do you cucks always have to bring politics into things. Fuck off.

>>59228630
Everything is political lad. Even technology.

>>59227873
>The whole thread smells funny.
No shit, half of us prolly haven't showered in a week.

>>59228645
Guilty as charged.

File: 1477472646521.jpg (9KB, 247x248px) Image search: [Google]

9KB, 247x248px

>>59227172
>yfw, they got them mixed up

https://www.youtube.com/watch?v=e2vPp0fQUkM

Hour long talk about debugging complex ICs (x86 CPUs) by an engineer at AMD to give you some idea of what it's like.


One figure I liked in the talk was that the CPUs run so fast that 1 second of uptime for a CPU is equivalent to like 9 years of testing in a lab. Random people running these chips are equivalent to an unbelievable amount of testing.

>>59228748
>spending the entire day looking at verilog

Clearly I haven't experienced true hell yet.

>Each zen thread is being registered as an individual core with its own L2 and L3 cache. i.e. totaling 136 MB cache!!. this is using Windows Sysinternals. This explains the SMT troubles in the event that a thread bounced to a HT thinking its the real deal.


Someone dungufd

File: 1480878808551.png (56KB, 605x773px) Image search: [Google]

56KB, 605x773px

Time to install gentoo, kek.

Too bad Ryzen isn't good at games out of the box
So it's DOA just because gaymers are fucking faggot retards

>>59228771

If only.

Staring at Cadence is pretty much the life of every IC designer.

>this is the job that is probably the most lucrative for EEs
>literally using one of the shittiest CAD applications on the market every day, because market penetration

>>59228843
Luckily gamers don't have jobs

so their opinions on zen will have zero effect on enterprise purchases

>>59228844
You'd think the folks who have to think about efficiency all day 24/7 would use something better.

>>59227289
Big companies have multiple generations of machines in rotation at any given time. Doing a huge refit doesn't make any sense, you keep the old ones in service until they aren't worth in in terms of power, maintenance, and storage costs.

Mac Pro refresh with dual Ryzen CPUs confirmed. Will probably be the only way to use dual Ryzen CPUs too. This is why Apple waited.

>>59228884
That'd be kinda neat.

>>59228844
How do you even keep track of multi-billion transistor IC's anyway? That shit should be automated, where does the brain capacity come from?

>>59228863
Their only job is finding issues in Ryzen for free.

>>59228888

Quads confirm. Apple will be all over these for the workstation performance, surprisingly low TDP and low bulk price. No more charging $1,500 for the next step in CPUs like with Intel.

>>59228884
God ears you.
I will save to get a nice Ryzen hackintosh build for my video and audio crap.

>>59228884
[citation needed]

>>59228875

One would think, but:

>ic designers don't make good ui/ux designers
>80% market saturation, maybe more in analog/mixed-signal
>literally the only competition is synopsys
>backwards compatibility with the cell libraries that now-retired IC designer made 20 years ago a hard requirement
>all foundries use cadence shit too

>>59228911
>hears
The fuck is wrong with my english.

>>59228911
If Apple uses AMD CPUs I'd be a win-win
Apple gets there higher profit margins and can make there iMacs thinner
We would be able to build custom AMD Macs

I was actually only considering Ryzen for my media server (plex and more). It was stated they were going to be better just not better than memelake and baby lake. I've been eyeballing used xeons for months. Even if you nab a cheap xeon a mobo for it is usually astronomically high used or new. Old complete workstations or old servers are the best bang for the buck in my experience but you have to go back to sandy or ivy bridge era to get something affordable.

>>59228920

It's an educated guess, moron. Its literally the best option available for the coming Mac Pro. They haven't updated in years for a very good Ryzen. ;^)

Plus they're already using Polaris GPUs on their laptops because of their efficiency. It would be great if Apple went full AMD, really kicking it to Intel.

My life would be complete if AMD opensorez the PSP and started full time coreboot support.

>>59228991
Also if it matters to you, Ryzen has ECC support. No more shelling out to Intel for a decent processor with ECC for NAS, home server, etc

>>59228884

Confirmed where?

>>59229015
>completely free software tinfoil router with a good core arch

I'm literally NSA proof aside from space beams

Ryzen is a monster. I didn't doubt it was going to be impressive but this is something else. The initial gaymen benchmarks weren't very impressive but when you looked at like every other benchmark it was always consistently on the top or very close. Here it's the same pattern the thing is fucking wrecking multiprocessor setups

>>59229035
Read
>>59228993
Its pretty much a guess

>>59229033
Yea that's the other thing too as it supports ecc just fine. I mean hell a brand new cpu and mobo server grade for under a grand if pretty game changing imo. I'm still rocking a couple sandy bridge 2500ks a 4690k and a 4790k. I'm good for gaming.

>>59229033

>Ryzen has ECC support

Yeah, hold onto your butts when Apple does their next event. Watch Intel stock plummet for once instead of AMD's.

>>59229067
I'll add that the new linux kernel adding virtual gpu support is great but I really like my haswell/devil's canyon vt-d for gpu passthrough. Quite nice. I'm not sold on the extra drm in the new ryzen or skylake/kaby cpu's. Negative imo.

File: 1485941479382.jpg (59KB, 476x594px) Image search: [Google]

59KB, 476x594px

>>59228903

>>59229048
>guess

Which is not a confirmation faggot, so stop spreading FUD like an Intelbred.

>>59227132
>Intel optimised games do better
>/v/ faggot thinks his fps matters at the enterprise scale
Intel is getting its shit wrecked.

>>59227139
>>59227150
>>59227163
>>59227180
You just can't beat Ryzen's perf/watt

>>59227196
Zen is beating Intel on sysbench single-thread lol!

>>59229115
And the best part: it's a $400 CPU

I hope that AMD wins lots of cash and becomes filthy rich to invest in R&D and Open Source.
They could even solve the video acceleration mess in Linux only for they jewish desires and still would help the whole world for that.

>>59227312
They're using dumb gamers as lab rats to see what the problems are with Zen. The server CPUs will demolish Intel.

>>59229108

Are you new here? You think when someone says 'confirmed' casually like that they actually mean it? Go back to plebbit, kid.

>>59229067
What really shitty about ECC support on intel is that intel was just petty about it. The memory controller on every intel CPU supports ECC its just intel locked it in microcode to only have it work on server chipsets.

>>59229108
Yes its a guess, speculation really, but based on how efficient Ryzen is how how much less AMD is charging for there processors I't wouldn't be half bad an Idea if Apple went Ryzen

File: 1459114630505.jpg (109KB, 1187x674px) Image search: [Google]

109KB, 1187x674px

>>59229115
Well I think Intel's been at the same arch for so long it's easy for devs to optimize. They're also the most popular brand so it's a no brainer since companies aim to make a profit. You target the biggest base almost always. AMD changed took a huge departure with Ryzen compared to Intel.
>pic related
Since when have you seen a m.2 and usb direct to cpu? Audio/dac? This is some crazy shit right here. We're looking at the future and you can bet Intel is at work doing this with their own asap.

>>59229159
>ultra low latency audio stack

AMD clearly has a market in mind.

>>59229140
>using thousands of consumers as an army of bug testers
This is actually a standard practice for companies now, so its not really surprising if AMD did this intentionally. Vidya games aren't critical workloads for companies so its acceptable. It would only be a problem if they shipped buggy hardware to enterprise clients.

>>59229159
You got the full I/O diagram? Looks really interesting.

>>59229168
It reminds me when they went with on die mem controllers. People were like wait, you can do that? x64 same thing. I can have more than 3-4GB ram?

>>59228905
>No more charging $1,500
Dude. It's Apple the one you're talking about

>>59229168
Even if you don't use the integrated audio the direct CPU path for USB devices should even give USB audio a decent latency

>>59229159
>dac directly to the processor
Hope that Steve Jobs ghost's haunts Cook and Ive in their dreams. They should got AMD.
Coreaudio is a dream, and this will improve that dream ever.

>>59229168
What's the market here honestly? Active/Passive sonars? Space?

>>59229190
He's talking about jewish prices in Intel shit.

>>59229224
Seismology, listening devices, I think

>>59229224
Low latency audio for audio production and stuff, even lower latency than Thunderbolt and PCIe

>>59229201

Eh...

It connects to the processor. Because the southbridge is directly on the CPU.

Zen is a true System on Chip design, all this work started years ago on Llano.

File: 1482361721692.png (106KB, 1362x1060px) Image search: [Google]

106KB, 1362x1060px

>>59229185
No I don't. Maybe the manual. https://www.asus.com/us/Motherboards/ROG-CROSSHAIR-VI-HERO/HelpDesk_Download/

>>59229159
>usb directly on the CPU

Dayum that's some HMD-tier shit, possibly external GPU as well

>>59229243
AM1 had a true SoC design, motherboards where just slots and ports. Not a chipset to be found

>>59229243
I believe Nvidia's Tegra is also their attempt which is nice for tablets. Wish they had a new shield tabby instead of switch but whatever. Wonder how they compare.

>>59229264
X200 motherboards don't have a chipset either IIRC just some traces connected to a dummy IC so stuff doesn't break.

>>59229280
>X300

>>59229264

It has only integrated the North Bridge.

The southbridge wasn't integrated until FM2

>>59229277

But no one cares about tegras.

>>59229159
Its the path to the eventual "Processing cube" that has memory, external interfaces integrated.

>>59229320
I was waiting for a shield tablet refresh w/ new tegra. fml

File: Kabini_die_shot_labled.jpg (62KB, 600x613px) Image search: [Google]

62KB, 600x613px

>>59229305
Kabini on AM1 was SoC.

Carrizo is also SoC, but it isn't on socket FM2, or really FM2+ for that matter.

>>59229323
I think integrating DRAM or whatever it is replacing it is gonna be the biggest latency and overall performance jump since the memory controller was integrated.

>>59229335
>DRAM at L3 latency

I'm getting sweaty just thinking about it.

File: 71-zecZyZOL._SL1050_.jpg (172KB, 1050x1050px) Image search: [Google]

172KB, 1050x1050px

>>59229305
AM1 came after FM2 and AM1 has compleately forgone the chipset on the mobo itself
Pic related

>>59229352
Looks so clean compared to the 440bx days where you had a northbridge and southbridge complete with their own hsf.

>>59227118
lol but everything other than the E5-2630 V4 are old ass $200 chips.

File: 1461148846463.jpg (1MB, 2048x1498px) Image search: [Google]

1MB, 2048x1498px

>>59229352
>>59229365
Slut 1 ;^)

>>59227802

>bobbehhill.png

File: AMD-Zen-16-Core-Naples-CPU-Wccftech-watermark-3-840x560.jpg (59KB, 840x560px) Image search: [Google]

59KB, 840x560px

INTEL IS FUCKED! INTEL IS LITERALLY FUCKED!

Intel, are you ready for Naples?

>>59229460
Quad-socket, 32-core Naples when?

was AMD just pulling Intel by the leash with a red herring this whole time? fucking ingenious gorilla marketing if they were, suddenly they are blindsided where they thought they were safe at the heart of their sales

File: 1458440007273.gif (2MB, 200x180px) Image search: [Google]

2MB, 200x180px

>>59229470

>128 cores
>256 threads
>512GiB or 1TiB RAM
>install gentoo
>compile muh kernels
>compile for programming
>renders
>decode and encode

This is a good thread.

>>59228818
IM SOLD
ORDERING
FUCK

File: 1488259520916.jpg (110KB, 577x459px) Image search: [Google]

110KB, 577x459px

>>59228818

>>59229365
Less stuff to blow up.

>>59229484
Nah, Intel knew, they're switching their release cadence from mobile > desktop > server to server > mobile > desktop

>>59229352
>bios on an 8 pin DIP
who makes this godboard?

>>59229497
Based on the Ryzen compile time from >>59228818, I did a shitty calculation and a Quad-Socket Naples system with 32-core CPUs would compile Gentoo in 7.7 seconds with a 30% overhead

>>59229352
Are these guys that sell these "ite" chips between the CMOS and PCI the richest fuckers ever? I haven't seen a motherboard without them.

>>59229594
Assuming scaling is good after a certain amount of threads.

>>59229588
i didnt even notice that shit that's cool
it's an SNSV board

has anyone tried disabling cores on the R7 ryzens to see if it clocks higher? what about smt on/off? I want to see max frequency with a NH-d15 or equivelant
with
4 core SMT on
4 core SMT off
6 core SMT on
6 core SMT off
8 core SMT off

I manually went through 20 Ryzen launch reviews and not a single one had anything about this in it. Did AMD tell them they couldn't do it?

I'm about to order one just to play with and answer my own questions.

>>59229594
I remember spending an entire day waiting for gentoo to build on my Dell Dimension

SEVEN POINT SEVEN SECONDS
FUCK

>>59227418
>Also who are these guys that they have literally $100,000 (without mobo/ram cost) in chips lying around?
The Phoronix guy isn't too far from STH, and it's just him in his basement
They serve a niche segment that's willing to pay/donate for their insights, since there's literally no one else doing the same thing and the only other way to find out it's to buy the hardware and test it yourself, which works well for big companies but not for small ones
>>59227609
Why would you need a iGP for a router
>>59227558
For a beowulf cluster or hobo server/workstation the 1700 is by far the best value, Naples boards will be quite more expensive
>>59227917
Maybe you should try actually discussing things instead of shitposting 24/7
>>59228158
It's the SMT implementation, apparently it can give up to a 45% increase in performance, so in the best case Zen 8 cores are performing almost like 12 cores
>>59228518
Vega focusing on FP8 and FP16 performance which would curbstomp anything out there for machine learning
Didn't you see the slides about Radeon Instict and how the entire thing was focused at servers?
>>59228413
>You seem to forget that AMD is the one always playing catch up.
Literally false, AMD obliterated Intel in the server market when they launched AMD64, and were incredibly competitive with their MCM Opteron's
>Intel already has all these contracts in place and Ryzen has yet to prove its stability or reliability.
Which is probably why they launched Zen first for HEDT, with enterprise coming two quarters later
Also, none of this contracts stopped the near instantaneous death of Itanium/EPIC, when Intel had secured and eliminated literally all of the competition on the server market by promising impossible things
>That's why nearly every office and business computer maintained Windows 98 until recently
That's due to legacy, and legacy systems run on some s360 or something equally backwards in the server market
Upgrading x86 systems is usually painless

>>59228576
>AMD doesn't have a million employees like Intel to testrun them and debug in every single application on the planet
Intel literally fired their validation teams to """"compete"""" with ARM
That's why there's a zillion bugs on their shit
Reminder: https://danluu.com/cpu-bugs/

Wow, look at that openssl and AES/SHA performance, just get that low power 65W part and you got yourself a amazing cert server.

I wonder how it does in dnssec/dnscryt

>>59229844
That would explain the exploding atoms.

>>59227855
Intel shilling pigeonhole style.

>>59229884
Stop trying to start shit, zeppelin's voltage/clock scaling is at the start of the thread.
What he said is 100% correct.

>>59228312
Get real nobody overclocks. Buy the part you want and fuck that Korean hobby.

>>59229497
I find it impossible for them to not run into serious NUMA issues after two sockets, even the single 32c Naples seems kind of a miracle if their fabric has none of the common issues of multi socket systems

>>59229913
somebody lost the silicon lottery and is mad about it :^)

>>59228413
>Windows 98
>worse than XP
You're what, 16?

>>59229913

Anyone who cares about getting the best value for their money overclocks, kiddo.

>>59227491

What a shitty graph.

can you possibly make it more difficult to read then rape it more with jpg?

>>59229470
Quad socket 4chip MCM would run nearly a kilowatt on the platform alone, not including storage and networking and expansion cards. That's flat out ridiculous

File: IMG_9131.jpg (155KB, 680x742px) Image search: [Google]

155KB, 680x742px

>>59229901
Lol Zen is server only, according to Intel shills. Got it.

>>59229523

Is there actual /g/ discussion going on in here instead of butt hurt thank god

File: AMD-Ryzen-7-1700X-Python-Linux-Kernel-Compile-Benchmarks-800x514.jpg (45KB, 800x514px) Image search: [Google]

45KB, 800x514px

>>59229616
That Xeon Phi is 64 low clocked Atom chips w/ 256 threads, you do the math.

File: IMG_9084.jpg (184KB, 761x948px) Image search: [Google]

184KB, 761x948px

>>59229934
>>59230101
Overclocking is a gay hobby for cuck gaymers, maybe a hundredth of a percent of counter users have ever overclocked.

File: 1478220057977.jpg (469KB, 1440x1920px) Image search: [Google]

469KB, 1440x1920px

>>59228903

>>59230152
>market reeks of entry-level motherboards with one-click overclock
>gay hobby
It's standard procedure, even you can do it.

>>59230183
Overclocking:
>harder on all components
>outs you as homosexual
>only real benefits attainable are edge cases
I just fail to see the point. Heck I underclocked my PowerBook G4 for years for the battery life.

>>59230138
2008 Mac Pro has a 1500w PSU...

>>59229626
More in depth reviews are coming. Anadtech has confirmed they're going to dive in as deep as they can.

The whole thing is a clusterfuck at the moment. Reviewers rushed just to get the bare minimum review out but there's a fuck load of shit to go over. That's not even touching on the apparent massive changes caused by last minute bios updates. I imagine it'll be months before everything settles completely.

>>59230512
>Reviewers rushed just to get the bare minimum review out but there's a fuck load of shit to go over
Can't really blame them desu, mobo vendors couldn't get their shit together, and some reviewers got their kits less than a day before the NDA lifted

>>59230151
This. How the fuck does this affect Naples?

why they so desperate to beat intel on pc gaymer market when they have enterprises and console partnership on their hands?

>>59228884
but will neo-apple drop thunderbolt? I wish they would, but I doubt it.

File: 1457394816759.png (57KB, 928x173px) Image search: [Google]

57KB, 928x173px

>>59230638
They were never trying to beat Intel on muh games, shills are leading you to believe that, Zen was always targeted at the enterprise with HEDT competitiveness being a side product

>>59228536
The fury is still 80-95% as good as the 1070 in benchmarks thanks to the driver updates since release.

lmao.

I kinda want to get one since they're down to $235. But my 7970 is still doing me fine and Vega is supposed to come out within the next 6 months.

>>59228009
I blame /v/ and fermi shitposting for ruining /g/

>>59230638
because they want to sell all the high leak current dies that can't be made into niples mcm's that they are getting from the manufacturing process. consumer [gaymer] market is a perfect outlet for that as no one really cares about extra 10-15W power consumption

File: 5071l2z7le.jpg (119KB, 1200x900px) Image search: [Google]

119KB, 1200x900px

>>59228638

>>59228518
>>59229827
Vega will have very good 4 and 8 bit integer performance as well apparently?

Which is very unusual for a GPU.

It also has features for game streaming to greatly lower latency. Like whatever that failed service that Gamefly I think bought does, but the technology is more there to make it work now.

It has 512TB of virtual address space so it can have people sharing the same GPU. So each Vega server with 8 cards can handle thousands of people at lower load then transfer them over onto less crowded cards when load is high.

There's so many server aimed features in Vega, but many of them also benefit consumers.

>>59230841
((( Ruberg )))

File: 17057.jpg (18KB, 450x337px) Image search: [Google]

18KB, 450x337px

>>59227118
It reminds me of Phenom I, the consumer version had a lot of bugs at launch and was generally considered a flop, but the server version absolutely destroyed anything Intel had to offer.

>>59230245
https://www.youtube.com/watch?v=5DBqNHormTs
i needed to look that up, i guess apple made actually good specced computers back then

>>59231289
The G5's were dual socket power guzzling monsters, Apple had to use leaky water cooler AIO's back then to keep the beasts under control

>>59230887
Are we going back to thin clients and mainframes again?

>>59230532
motherboard vendors were given sub 3 weeks to make shit work with finalized amd chips, this is all on amd for being so secretive.

>>59231452
have been kind of going that way for a while now, it makes it easier in massive companies, and when someone needs a beefier computer then what they have, the power is there to shift resources.

>>59227454
AMD hasn't released a really competitive product in 15 years tbqh

>>59227139
Now this is underrated. 1700X now I guess.

>>59227491
Good heavens. This is the other side of AMD. Fuck gaymen shills.

>>59228169
It does if you do more than just gaming.

R7s are just fine for gaming, but excel at other tasks.

An 1700 with a 1080 is about as close to the best multirole computer you can get

>>59230228
well, harder on all components, but when you buy a system specifically for it, the components can take the load.

The benefits... well lets take the 1700 as a case example here, it operates at 3.0ghz and has a turbo of 3.7 and no telling if it's an all core turbo or not, xfr is an all core turbo, but its not on the 1700, so the only way to get good speeds is to overclock. the average oc is around 4.0ghz so far, so you are looking at a 25% performance boost by overclocking. Because it has weaker single core then intel, this overclock puts it in a higher tier of cpu for effectively nothing lost if you have a good enough motherboard, as its below safe everyday voltage.

Also likely going to have a low clocked underclock variant, which downclocks the cpu and voltage, so it effectively draws no power. I don't like turning my computer off, and my current cpu is a bit of a bitch because processes will kick in in the middle of the night and ping cores to 100% for hours heating my room up. Will be nice to have a tool that I can save profiles to for when I sleep and when I wake up.

But this is my two cents, not every program I use is fantastically multithreaded, and not every task I use is only single threaded, If I only had multithreaded tasks, I would buy an ES intel and roll dice on that as it would be the performance king, but I cant forgo the single threaded, So I'm splitting the baby, and if I can save what is it, 180$ on the cpu, that just means I can get better things in other areas.

>>59231722
Even if you game, there's going to be a Windows patch that significantly improves things.

The benchmarks of 20-30% off are way off. It should rarely be more than 5% behind the 7700k. There's no architectural reason it's doing so bad, it's just Windows isn't scheduling threads properly mostly and that SMT is giving a 8-12% performance hit when it automatically shouldn't be getting used when it would.

The 7700k is still going to win in a lot of games, especially older games, after all the issues are ironed out. But some of these benchmarks now are really skewed.

>>59228428
Considering 3.3 is the sweet spot for 14nm LPP, it's totally okay.

Does AMD plan to release low-power Ryzen CPUs? They would make sick low cost NAS/HTPC/home server builds.

Duck you amd shill! Intel still gets 250FPS vs you're 220FPS in 720p gaming. Hahahahaaa.

So zen had no botnet, intel user cuck themself again

>>59231732
>no telling if it's an all core turbo or not
It's all core
>xfr is an all core turbo, but its not on the 1700
XFR is present on all models but it only gives a 25 pr 50 Mhz boost on non-X SKU's, on X SKU's it gives 100 or 200 Mhz
>>59231896
They do, but not until the server parts drop, neither beore APU's drop in the consumer market

File: 1418486857294.jpg (50KB, 500x535px) Image search: [Google]

50KB, 500x535px

>>59227312
Basically AMD marketed what is essentially a HEDT competitor as a consumer i5/i7 competitor. The fact that they're doing it all on one platform (AM4) probably didn't help either.

The R5 will probably be where Ryzen shines for competing with Intels mainstream desktop stuff like the 7600/7700K.

At least I hope.

>>59232059
There's no architectural differences, so the 1600x should literally be 1800/1700 tier but with two cores disabled and $250.

I don't see it outperforming the 1800x unless it's from extra overclock headroom thanks to fewer cores.

>>59232117
The 1600X will have 33% more unified cache shared per core. That may actually significantly increase single core performance.

Clocks will likely not be any higher.

File: BarraCuda2TB-large.jpg (113KB, 461x750px) Image search: [Google]

113KB, 461x750px

So my old desktop finally died but the HDD is still good, what's a good adapter to use to recover my data?

pic related

For gaymers like for like, in a top tier match up intel still win but for the price you pay to get somewhere statisticaly close to parity it's going to be a real winner for high end gaming on a budget. But that said thats only in the interim. To me the results appear to gpu issues and to see real potential we will need to wait for next gen gpu's.

With that said gaymers are fark all with regards to revenue. The server market, rendering and cad machines however these areas will pale in comparison to the largest market of them all. The business client - aka that dusty pos that you sit in front of on a daily basis. Powerful and cheap - mark my words the intels jewery will be shaking in their mashughanas over losing this market. The business client makes ~75 - 80% of the global desktop market.

>>59232171
https://www.amazon.com/Sabrent-External-Lay-Flat-Docking-EC-DFLT/dp/B00LS5NFQ2/ref=sr_1_4?s=pc&ie=UTF8&qid=1488626589&sr=1-4&keywords=external+hard+drive+case

>>59232309
thank you

>>59232117
Whats going to happen if they let you unlock cores like with the phenoms

>>59232338
They started lasering the cores since bulldozer so no unlocks for you.

File: 1439331999835.png (592KB, 876x882px) Image search: [Google]

592KB, 876x882px

If it doesn't support ECC memory and fails at games, who it the target audience? 6950X owners? Because people that need 6950X for rendering, you know, already have it. And they won't give two shits that is costs less if 6950X is even just slightly better (which it is).

>>59232395
>fails at gaymen
By having 5-7 fps lower than 7700k?

>>59232412
By costing 53% more fore 5-7 lower FPS.

>>59232422
Maybe, just maybe, R5 and R3 exist?

File: 1476243336588.jpg (372KB, 740x933px) Image search: [Google]

372KB, 740x933px

that probably settles it. I'm still running a xeon from 8 years ago. hopefully these zens will do at least a 5 year stretch.

>>59232427
As far as I know, they don't. Nice way of switching goals.

>Ryzen R7 is bad.
>B-b-b-but R5!

looks like my E5-2670 is now obsolete

>>59232449
(you).

>>59232427

I don't think you get the memo.

Whenever ryzen doesn't top a chart it must be compared to to whichever intel chip it doesn't best, regardless of which intel chip it is - price isn't a factor.

If it beats a 6990k but not a 7700k in vidya
>aahahaha can't beat a cheaper chip!
If it beats a 7700k in HEDT focused workloads but loses to a 6900k
>ahahah can't beat Intel!

The mere fact there is best part of 300 burger difference between the two Intel chips is irrelevant, because once you do startl ooking at price across many workloads the 1700 basically destroys every other chip.

bretty good

>>59232478
>I don't think you get the memo.
Neither do you. AMD is competing in 0.5% of the market where price doesn't really make a difference.

>>59231732
All core turbo on 1700 is 3.2GHz.
But if you're buying the 1700 you're probably looking at the fantastic power consumption, so overclocking over 3.4/3.5 dumps that in the river.
However, a 1700 overclocked to 3.9 is insane value and handily beats a stock 6900k and gets real close to a 6950X, which is over FIVE times its price, and that's not even counting the extra platform cost of x99

>>59233016
If it uses around 70W running POVray at 3.2 it can probably use around 50W or lower running it at a more sane 2.8GHz which is still a insane base clock for a 24+ core chip, much less a 32 core one.

>>59233076
Example.

Intel® Xeon® Processor E7-8894 v4
>24 cores
>2.4 base
>165W TDP
>launched a month ago
And Intel wants $8900 for this

>>59227558
No dual socket motherboards yet so just wait for naples imo

>>59233016

The 1700 is semmingly the chip for smart people of the current ryzen releases. It is the same as 8320 vs 8350 - lower clcoks but hits the same wall its more expensive bigger brothers do. So save dosh and crank it up yourself if you want top performance from the chip or keep it at stock and enjoy that delicious low power draw.

Flatout its win/win if you can into computer.

>>59233245
>>59233016
Don't forget that the 1700 comes with a cooler that's worth at least $30.
So it's effectively $300.

And the B350 motherboards are pretty decent boards for only $80-$100 and people seem to be getting 3.9ghz all core overclocks on B350 boards.

>>59233016
The 1700 is really great, the cheapest, can OC as much as the others + the only one that gets a cooler bundled, and you can even do a decent OC on said cooler without burning down your house.

Amazing value.

>>59227118
>Linux
Literally who the fuck cares?

And it's doing all that with lower power consumption.

>>59233278
>being this retarded
wew, stop posting

>>59233286
Don't respond to it.

File: 1488590910330.png (55KB, 1509x905px) Image search: [Google]

55KB, 1509x905px

>>59233273
It may overclock as well as the others, but it uses more power doing so.

The difference in the binning is basically that the frequency max for a given voltage min shifts to the right about 100mhz.

The critical 1 point is at 3.4ghz or higher for the 1800X while for the 1700 it can be below 3.3ghz it is on this graph. And the 3rd critical shifts back to around 4.0-4.1

So yeah. A 1700 will reach 4.1 fine, but it will use more power and put out more heat. You probably want to keep the 1700 at 3.8 or 3.9 all-core at highest.
And if you get an 1800X, honestly just let XFR do its thing. It will bring 2 cores to 4.1 in most cases much more efficiently and without fuss.

>>59233272

The cooler is nice but if you are overclocking its no match for big air or water.

>>59233300

Is that from The Stilt's analysis?

>>59233300
If you're overclocking you already decided to throw perf/watt out of the Window, but for $200 cheaper I see no problem with slightly higher voltage.

Skylake Xeon will pretty much destroy AYYMD HOUSEFIRES Rypoo

>>59233316
Not necessarily when you do it with 1800X. But now you threw out another 50% of the price of 1700/7700k.

>>59233308
Joker was getting 60C temps with the stock cooler at 3.9ghz all core on a 1700.
Maybe he hit the silicon lottery and his is extra efficient. Won't know that for a while.

And yes, it's from Stilt's analysis.

>>59233316
>If you're overclocking you already decided to throw perf/watt out of the Window
Not necessarily...

The 1700 only does all core turbo of 3.3, and single to 3.7. There's no reason to not just stick it at 3.8 or 3.9 that's at the end of its efficient point.
Just because the chips will apparently take over 1.9v without any damage doesn't mean it's worthwhile. You'll need a 300w+ TDP cooler. You have to put so much more power into them for little gains beyond 3.9-4.1 depending on your silicon lottery. But if you live in Antarctica and need a space heater, sure I guess.

>>59233369
Only does 3.3 all core and 2 cores to 3.7 automatically, I should have said.

>>59227118
http://www.phoronix.com/scan.php?page=article&item=amd-ryzen-1700

File: IQVrz.png (84KB, 299x288px) Image search: [Google]

84KB, 299x288px

>>59232395
Ryzen supports ECC retard.

>>59233369

>Joker was getting 60C temps with the stock cooler at 3.9ghz all core on a 1700

Then again, vidya isn't going to slam a cpu like prime95 or IBT is and those are what I use my my cpu overclock testing. I've got a 1700 on order and i'm going to bolt my silver arrow to it - if that can't cool it then it can't be done on air.

Seems along with the CCX and cache issues, SMT issues and BIOS issues, the Balanced power mode in Windows increases clock and voltages switching by a lot.

>>59228518
HBC effectively allows Vega to load massive CAD files into it's memory relatively quickly. The massive boost in mixed precision could well post interesting gains as a GPGPU as well.

>>59229033
I'm actually looking forward to the lower powered R5 6c part for a microserver. Going to get an itx board and a few 2.5" drives to load into it and run a vpn pass through for my network. May as well run a web server too, but damned if I know what I'll put on it.

>>59227276
I dont give two shits about ryzen but am interested in snowy owl and naples.

pretty badass, looking forward to their Opteron lineup. I just wish their Ryzen launch wasn't so buggered and marred by the gaming marketing bullshit and also these apparent bugs.

When are the Opterons due?

1.9 GHz 8/16 opteron at 35w when?

>>59235194
Q2 if not delayed.

>>59235206

Unless you need ECC the 1700 is effectively that chip.

>>59235218
You don't delay enterprise launches, or you'd get some very angry rich people twiddling their thumbs waiting for you to deliver them hardware, and you don't want them angry if you care about your bottom line.

>>59227382
>That Xeon-D 1587 is a 16c/32t part.
You're missing something:
>TDP: 65W
>Ryzen TDP: 95W

You're also assuming it scales out well to 16C which remains to be seen. (I hope it does, but this kind of speculation is what got us into this /g/ meltdown about Ryzen's gaming performance.)

>>59235242
>Unless you need ECC the 1700 is effectively that chip.
lolno

>>59235256

You should look at The Stilt's testing.

>>59235242
Ryzen supports ECC, dude.

And you're off by some 30 watts.

>>59227276
The only problem with Ryzen is that they released their most expensive CPU first. The 7700k only costs around $330. Ryzen has 12% lower performance. Even though it's not bad it doesn't make sense to buy the worse product when both cost the same but on the other hand if they sell the 4 or 6 core Ryzen in the $100 to $200 price range they'll btfo intel and take over the market.

>>59230841
Are you trying to acusse me of being a SJW?

>>59235389
>Bonnie Ruberg
Obviously it's the cries of another diseased transsexual animal, ignore it.

>>59235278
Hello anon-kun! Did you know that reducing the clockspeed and voltage of a CPU reduces the power consumption?

>>59229410
>Onboard Voodoo3 2000

I had no idea someone made something like that. Looks really cool.

http://www.anandtech.com/show/363

>>59235278

Ryzen does but no current motherboard does fool.

>inb4 showing me motherboard specs that have the letters ECC in them

One should note said ECC memory will run in non-ECC mode.

>>59235688
It's up to the motherboard dudes, which have really fucked up this launch with late BIOS updates.

You're gonna pay extra for that ECC checkbox if some release the boards with it, trust me.

>>59227118
BUT CAN IT MINE CRYPTO CURRENCY

AT NEVER BEFORE OVER 88 MILES PER HOUR SPEEDS?

EITHER IT CAN OR AMD SHOULD CAN' IT.

>>59233339
>Just wait.

If there's anything I've learned over the last 2 days on /g/ is that there's a clear distinction on this board between crossboarders and actual tech enthusiasts.

>>59236902
And the former usually win because they're louder.

Gamers are cancer.

/g/ is literally filled with /v/tards

All they can do is shitpost about how shit Ryzen is at gaming. Like AMD care. They know where the bread and butter is.

>>59237066
AMD wouldn't mind if Zen was better at gaming though, that's still revenue.
However compared to the datacenter it's chump change, it's also a very volatile market (dropping shipments by up to 15% per quarter and then bouncing back later), you never know how many units you really need, unlike datacenters where if someone is interested will ring you up, say "I want this and that until this deadline" and you get to work, no bullshit or huge marketing needed, just a decent product.
Beats dealing with deranged

>>59237111
>Beats dealing with deranged children running steam on a PC bought by mom's credit card

Man I stopped gaming properly years ago, there's one decent game every 2 years at best and that lasts me for a few weeks and done.

>>59227312
They didn't market it for high performance computing because they haven't launched their server line.

And these chips are overclocked to hell out of the box. The Samsung 14nm process is aimed at sub-3GHz. That stilts guy on anand already tested it and Ryzen has over 2x perf/W when running <3Ghz compared to stock clocks.

>>59237191
14nm LPP's sweet spot is around 3.3ghz according to stilt.

>>59237223
I thought the optimal spot was 1.9-2GHz on all 8cores consuming literally 30W with SMT enabled.

>>59237128
>there's one decent game every 2 years at best and that lasts me for a few weeks and done
Not to mention that the "decent game" usually doesn't even require powerful pc, so spending shtiload of cash on pc just to play videogames is beyond retarded.

>>59237223
It's linear scaling from 2.2 to 3.3, but still 0.7v at 2.2 will be more efficient than at 3.3, especially considering a lot of Naples centered workloads scale nicely with cores and near full utilization of all threads is simply more efficient than running fewer cores at higher clocks.

File: Screenshot_20170304_195410.png (54KB, 1480x878px) Image search: [Google]

54KB, 1480x878px

>>59237315
The cutoff point is 2.1GHz, anything lower than that and you can't scale down anymore.

I wanna see if improvements to GloFo's process or the next Zen core iteration will fix this, this doesn't seem like a uarch issue but a metal layer issue.

>actual tech discussion
Where's shitposting? I though that was a Ryzen thread.

>>59237417
Sometimes people who aren't shitposters, crossboarders or normalfags searching for tech support make threads.
Surprising, I know.
Some people actually like hardware design and talking about it.

>>59237474

Those sorts of people don't post here after /v/ drove them out. Its why it is impossdible to have a cpu discussion on /g/ as they are inherently very complex pieces of kit (more so than gpus given the limitations of cpu workloads).

>>59237527
Software side of things GPU should be more difficult to juggle than CPUs.
CPUs on the hardware front are in fact more complicated, but that's comparing someone with a 150IQ to a 140IQ.
They're both some of the most complicated human designed projects on the planet.

>>59237594

Yeah but its harder to pin down the nuances of cpu behaviour in shiny graphs (especially gaming) than it is gpu nuances.

>>59237614
For the layman, CPUs do have more debugging tools for finding out what's wrong.

>>59227118
It's above dual E5-2620v1 in all tests, but it's not that impressive if we consider that's 24t*2GHz vs. 16t*4GHz and the difference isn't that big anyway. Looking at compilation, 7Zip, OpenSSL tests. Unsure if c-ray is widely used in servers. It will be better than dual Xeons in a dual-socket configuration, but how much it's going to cost is another question.

>>59237883
Wut? the 1700x turbos all cores to 3.5GHz and the dual xeons are 10 core 20 thread parts each.
This is certainly impressive if you ask me.

File: kermit sipping tea.jpg (111KB, 600x600px) Image search: [Google]

111KB, 600x600px

>>59237883
>>59237905
the OP image features two E5-2670 v1 which are $50 each now and the Ryzen barely edges them out. The point though is that the Ryzen is just a single chip and it's new whereas you ahve to buy the E5-2670 v1 off-lease on eBay.

Also I bet if you scale down the Ryzen chip to lower clocks and stick two of them on a board it will run circles around the x2 E5-2670v1 while clocking in at similar or lower power draw, especially at idle.

but that's none of my business

>>59237905
>1700x turbos all cores to 3.5GHz
I see. Thanks for pointing that out.
E5-2620 is 6 cores 12 threads. Again, not looking at c-ray since that not the type of software I'm going to use.I guess, 7-Zip test is outdated, though, because 2630v3 is above dual 2630v3. That means the only meaningful tests there are OpenSSL and compilation. For OpenSSL, dual E5-2620 v3 (2,4GHz, 12 threads) are slightly above a single Ryzen.
For compilation, two 2670 (2,6GHz, 16threads per CPU) are winning by 1/4, and that is something to consider. Yet again, this is a consumer CPU, AMD may throttle the frequency to 2,5 GHz to fit into a desired TDP with 16 cores per CPU.

I don't consider synthetic bencmanks because that's not what people run on servers.

>>59238123
Well, Ryzen has proven to scale extremely well with voltage into that range, and do keep in mind they plan on cramming 4 Ryzen dies into each Naples processor. Scale it down so each die pulls 50W x4 per package, thats 200W per processor, with 2 per package, thats about 400W for the CPUs. Intel's high-end quad-sockets hit I think around that figure depending on how they're set up.

>>59238208
I thought quad socket Intel SKUs are all over 160W

>>59238440
There should be some smaller lower core count SKUs that can run in quad socket configuration.

>>59238440
Wouldn't make sense, not everyone needs top performance, some just want plenty of memory and I/O with adequate performance at low power.

People tend to forget Naples is 8 channel, which is a significant bandwidth upgrade over the upcoming Skylakes.

>>59238676
It already has pretty good bandwidth in dual channel but I heard those AIDA tests are bugged, so I'll keep this opinion under leash for a week or two.