AMD Zen Microarchitecture: Dual Schedulers, Micro-Op Cache and

>>56142310
P O O
O
O

>>56142310
>QUANTUM LEAP
I hate this.

>>56142310
i3 > Zen

Who gives a shit, if it performs like an i5 from last year.

>>56142363
That would make it 5% behind Intel's current stuff. Huge jump from Bulldozer.

File: boku.jpg (340KB, 1607x2535px) Image search: [Google]

340KB, 1607x2535px

>>56142310
>quantum leap
>slower than old intels
lmao

>>56142363
Me, because I could get an affordable last year i5 with 8c/16t

>>56142363
So what? I'd be happy to spend less on a reasonable processor so I can drop some more cash on a new graphics card.

>>56142392
I am now gay.

> AMD says Zen CPU will outperform Intel Broadwell-E, delays release to 2017
http://arstechnica.com/gadgets/2016/08/amd-zen-performance-details-release-date/

>>56142363
I'd probably even buy it if it performed the same as Sandy Bridge, just to spite Intel and their disgusting practices.

>>56142412
For real. Unless I fire up a video to encode I rarely use 15% of my 4770k.

CPUs are rarely a bottleneck.

File: Road-to-Zen_Architecture-Brief-11-980x552.png (133KB, 980x552px) Image search: [Google]

133KB, 980x552px

AMD is back, believe the hype.

>>56142421
It's ok, be becamea bimbo girl now.
You're just half gay.

>>56142392
muh dik

>>56142442
I do emulation so I need intel.

File: FuckingTriggered.gif (2MB, 421x338px) Image search: [Google]

2MB, 421x338px

>>56142428
>2017
REEEEE, I WANT TO UPGRADE NOW. IM SICK OF MY 2500K

going to love the wait meme on this one

>>56142446
>unlabeled axis
literally what

>>56142632
It's wait time.

>All those retards in this thread think Summit Ridge will be cheaper than Intel's i7.

Loving every laugh. Expect 8c/16t starting at $300 and going up.

>>56142386
Problem is 14nm LPP, they won't clock as high as Intel. Maybe Zen+ will use 14nm (or 10nm) SOI FINFET

>>56142924
Stop regurgitating this stupidity. The Stilt is a smart guy, but he understands literally nothing about process.
Samsung's 14nm clocks higher than 28nm HPP, and we have 28nm HPP chips on the market right now with 3.8ghz turbos and a 3.5ghz base clock.

Global Foundries will not offer a 10nm process either as its only a half node from 14nm. They will move to IBM's 7nm process after 14nm for high performance chips.

File: 1471178850324.jpg (44KB, 419x328px) Image search: [Google]

44KB, 419x328px

>>56142393
>It only has 4 physical cores

>>56142363
>higher core count at same price
>priced like an i3 from current year
>unlocked multiplier on every chip always
>hardware virt on every chip always
>no IME (PSP won't be on high end desktop chips)
>way way longer motherboard lifespan

I'll trade a nominal decrease in power for all of the above. Any time. You don't belong on /g/ if you wouldn't.

>>56142962
If they can hit 4 GHz I'll be content

>>56143039
>>no IME (PSP won't be on high end desktop chips)
[citation needed]

File: amd zen.gif (2MB, 164x275px) Image search: [Google]

2MB, 164x275px

>>56142310
>Zen logo
>that smear

Found AMD's inspiration

>>56143057
With a stock turbo of 3.2ghz on an ES inside of a 95w TDP thats a virtual guarantee.

>Early 2017

Fuck, I want to do a zen build for my next system, but there have been so many delays.

What the fuck, I hate intel now.

>>56143115
it officially announced to come out in Q4 2016 and it's delayed to Q1 2016, there were no other delays officially.

>>56143115
There has not been a single delay of any Zen based part.
The only thing AMD has delayed is K12, and it might be shitcanned all together.

>>56143039
>im poor

>>56143087
https://libreboot.org/faq/#amdpsp
>Early anecdotal reports indicate that AMD's boot guard counterpart will be used on most OEM hardware, disabled only on so-called "enthusiast" CPUs.

I choose to believe

>>56143115
Same. My 8370 is starting to gimp me hard.

I'll buy a Zen CPU as early as possible.
I have preferred AMD since my first computer, an AMD 386 with 40 MHz.

>>56143134
Things that never happened: The Post

There was no official release date given. AMD stated that 2017 would be the first full year of revenue for Zen. They also stated that there would be limited availability in 2016.

Stop spreading FUD.

>>56143138
>literally the only intelfag argument left
any adult without kids that works full time can afford X99 quad Titan setups. that doesn't mean it's the smart thing to do.

>>56143151
>They also stated that there would be limited availability in 2016.

For any anons who are slow to catch on to this: 'limited availability' usually means final sample chips in the hands of vendors.

So which is the best i7 for 350-450€ ?

>>56143152
>afford

That's not the word you should use there m8

File: AMD-2015-2016-x86-Zen-Roadmap.jpg (266KB, 2256x1265px) Image search: [Google]

266KB, 2256x1265px

>>56143151
Zen was promised 2016 and of course "full year of revenue for Zen" wont be starting with Q4 of 2016 you dumb bother fucker.

now it wont even be 2017 because full year = ALL THE DAYS.

>>56143185
I have money stop implying I don't reeee

>>56142721
I dont mind. Give me 8c/16t broadwell-e with vt-d for $300, I'll buy it.

Everytime I want to believe and everytime I'm let down. Why should I believe them once again? Note that I actually believe and I'll probably be disappointed as always.

>>56143211
>concern troll
shills are deploying the heavy tactics now

>>56143095
kek

>>56143193
No, it wasn't "promised" in 2016.
If you actually watched that FAD you'd know that. They have said the exact same statement consistently. Only limited availability for 2016 was ever explicitly mentioned. They never, ever, ever stated that Summit Ridge or Zen Opterons would release full volume in 2016.

>>56143222
If only I was a troll. I'm just a poorfag who wants a nice cheap CPU.

>>56143152
>Not spending money on the one item you spend the most of your time on.
You may be willing to compromise on your hardware but don't shill products as superior when they are in fact objectively slower. Sure AMD is the economical choice....it always has been.

>>56143246
just save up, for fuck's sake. do you not have any revenue at all?

>>56143238
Summit Ridge was always aimed at 2017.
Also nice goal post moving there.

>>56143257
I live in a third world shithole. A CPU is like 10 times my monthly wages.

>>56143264
>Summit Ridge was always aimed at 2017.
Yes, indeed it was always aimed at 2017 for full retail availability. That has been AMD's statement from the beginning, and that is exactly what I stated.

>>56143282
And now that statement is FALSE

You dense motherfucker.
Now there is no availability for ZEN in 2016 making their prior statement false.

Go fuck yourself

>>56142518
>RX480
Why, it's still a great CPU. Still rocking an i5-3570k at stock speeds.

>>56142393
>#
Are you serious, core i5s have only 4 physical pleb

>>56143254
there's no "economical choice" in the CPU market. only what's best at your price range. I don't know if you've noticed but Intel has products from 40 bucks to multiple thousands.

in addition, there's more to a product like that than anandtech benchmarks. with Intel, there's a very real tradeoff. you have to pay a premium for unlocked chips. you have to buy a brand new motherboard every time you want to upgrade. you have to deal with the IME and all the restrictions and security issues it inflicts upon you.

you're a massive retard if you don't understand why someone would prefer to trade 5% perf (which would still be way over bottleneck in most vidyo gaems) for all the shit I've mentioned above.

unless zen turns out to be way slower than advertised, or unless Intel drastically changes their business practices, the zen lineup will be the most enticing to me and many others.

>>56143305
see: >>56143169

>>56143305
Yet again, retard. There has been no delay.
AMD has not delayed anything. Sites like Ars are just repeating clickbait from digitimes which was never based on anything except someone misinterpreting statements made by Su and Papermaster.

Summit Ridge wasn't delayed. It is right on track.

>>56143353
>Hey, They promised it would be available in 2016
>They didn't promise it will be available in 2016

I repeat myself.
Go fuck yourself.

>>56143375
Skylake was a paper launch

inb4 next Mac Pro uses Zen
inb4 that's the reason for the delay

>>56143395
>Irrelevant information the post

I repeat myself.
Go fuck yourself.

File: 74b.jpg (29KB, 252x291px) Image search: [Google]

29KB, 252x291px

>>56143415

>>56142990
>8 core CPU
>only has 4 cores
u wot

>>56143375
it will be available in 2016. just not for you.

>>56143403
God that'd be great, Apple using Intel for the last 10 years is the main reason I stay away from Macs

>>56143339
Im confused what is the cause of your hatred for IME? What restrictions are you talking about.

>>56143403
That would be hilarious

>>56143375
I'm pretty sure they said limited availability in 2016.

File: AMD-Computing-Roadmap-2016-FX-CPUs-APUs-GPUs.jpg (561KB, 2256x1264px) Image search: [Google]

561KB, 2256x1264px

>>56143369
Didn't realize that this roadmap says 2017 on it.

Also where is Bristol Ridge. it also was supposed to come out 2016.

>>56143497
7th gen APUs are Carrizo which is out. There have been GPUs with HBM for months. Zen is limited availability.

>>56143497
OEMs have had Bristol Ridge chips for weeks.
Summit Ridge has limited availability in 2016.

Nothing was delayed, and once again if you actually watched that FAD you'd know this.

>>56143561
Carrizo is 6th gen, though they're basically the exact same thing.
Some stuff does point to Bristol Ridge being a respin though, they're still more or less the same chip in a different package.

>>56143432
Thats been the idea for amd

>>56143599
They've never done such a thing, dumbass.

>>56143471
do you know what it is? it's a separate core built into your CPU's die. it has access to

>all of your system's RAM via DMA
>your network
>a full fledged JVM with a hefty amount of class libs
>storage on your PC, including the ability to run software off of it

it runs even when your PC is powered off. there's no way to tell what the fuck it's doing except for some network level forensics.

should Intel's signing key be leaked (like Microsoft's secureboot bypass key was not long ago), every single Intel machine from 2006 and later is compromised.

not to mention the built in DRM components.

and on top of that, it will only run signed firmware which means you absolutely can't do shit about it. I'm probably missing half a dozen things too.

now since 2013 IIRC AMD also does the same thing with PSP, which is equally shitty. but it's possible that desktop x86 chips won't feature it.

>>56143589
>Carrizo is 6th gen
Oh woops you seem to be correct.

http://www.anandtech.com/show/10362/amd-7th-generation-apu-bristol-ridge-stoney-ridge-for-notebooks

wait did bulldozer not have a uop cache?
wtf lol

File: 1471527461591.png (125KB, 980x552px) Image search: [Google]

125KB, 980x552px

>>56142446
So Zen's IPC increase over Excavator is equal to Excavator's IPC increase over Bulldozer? What?

File: 4iivjn.jpg (108KB, 1024x724px) Image search: [Google]

108KB, 1024x724px

>>56143589
From the roadmap that leaked last year.
IP means in production. Final silicon being run at the fab.

Summit Ridge has met every milestone AMD set out. It was never delayed.

>>56143678
AMD made that claim months ago anon. This graph is nicer looking though.

>>56143254
how much do you spend on your CPUs?

>>56143678
the total being about 80% is consistent with the leaked Ashes of Singularity benchmark.

>>56143471
hardware backdoor

File: ipc uplift.png (10KB, 582x162px) Image search: [Google]

10KB, 582x162px

>>56143678
Carrizo APUs show a nearly 30% uplift on average over Bulldozer. Total IPC would be slightly higher if they had an L3.

>>56143737
One advantage excavator has though is it uses a fraction of the power while offering comparable performance.

I find it interesting that AMD took their "big" construction cores and turned them into low power cores. Their bobcat cores just disappeared.

>>56143805
Well the Jaguar core arch was ever only a tiny change from Bobcat, and the Bobcat core dates back quite a ways. It was designed around 2007-2008, its an ancient arch. All the Jaguar cores need is frequency, but they aren't designed for it. Excavator can clock high at relatively low power. Thats the one thing you can say about it. Its kind of surprising that Excavator can compete at all given it still has all the cache issues, and performance in some ops actually decreased vs Steamroller from the L2 being cut in half.

Frequency comparison for fun:


XVR with AVFS
3700mhz: 22.5w~
3500mhz: 16w~
3150mhz: 10w~
2800mhz: 7w~
2450mhz: >5w~
2100mhz: 3w~
1750mhz: >2.5w~

STR
3700mhz: 25w~
3500mhz: 20w~
3150mhz: 14w~
2800mhz: 10w~
2450mhz: 7.5w~
2100mhz: 6w~
1750mhz: 4.5w~


If the Excavator core were to be featured in a 14nm design it might really shine, but then again AMD's Zen core allegedly scales better than this.

>>56143497
Official launch 2017, HOW EVER they will have a few chips out Q4 2016 for a few of their pardners and maybe a luck people will get their hands on one.

>>56143613
nothing to hide, nothing to fear

>>56144057
0/10 troll

>>56143613
>every single Intel machine from 2006 and later is compromised

Only the ones with vPro enabled. Intel didn't start putting the thing in CPUs until sandy bridge. Before that it was only on workstation/enterprise mobos.

File: 1448861584965.png (362KB, 700x700px) Image search: [Google]

362KB, 700x700px

>>56142518
>sick of one of the best processors in the decade

>>56142363
>Who gives a shit, if it performs like an i5 from last year.
That's like a 30% improvement over bulldozer, 5% improvement over ivy bridge, and it's only 5% behind skylake. That means that if they price it right and leave them unlocked, there will be no logical reason (yet) to buy the current generation intel processors.

>>56143306

i5-3570k checking in!

>>56142310
Nice, and it even beats Broadwell in IPC, running Blender, if you watch the video. That is probably one of the best cases for the architecture, but still, that is very nice, naysayers were putting it way lower.

>>56144525
note that i7-6900K had bigger L3 cache for that test. Not sure if AMD gave it four-channel RAM, Summit Ridge Zen will only be dual-channel.

File: excite.jpg (159KB, 1013x1092px) Image search: [Google]

159KB, 1013x1092px

>>56144063
Truth is the best troll, anon. - old knowledge

>>56144525
That's great but it's only 1 program. I'm guessing you realize they're gonna choose the program where it edges out intel THE MOST. So that was the best case scenario. Also it's a bit suspicious that they underclock the Intel instead of just "overclocking" their own processor. Let's see what frequencies they hit. That being said it's looking promising.

>>56143902
>Well the Jaguar core arch was ever only a tiny change from Bobcat, and the Bobcat core dates back quite a ways

Not really, IPC was significantly better in jaguar and the FPU/SIMD part was completely overhauled. Jaguar has full 128bit throughput for up to SSE4 and AVX compatibility as a bonus. The SIMD performance of Jaguar is pretty neat, even then-x264 dev Dark Shikari liked it when it launched.

It scaled to higher clock too

>>56144577
>That is probably one of the best cases for the architecture
I said the same thing already.

> Also it's a bit suspicious that they underclock the Intel instead of just "overclocking" their own processor.

Dude, how else do you want to compare IPC of the architecture than on the same frequency. The IPC comparison is actually more interesting to me than raw results for stock clocks.

By all means the final clocks of Zen will probably be lower than desktop Skylake (compared to i7-6900K, dunno, might still come lower even than that). It will also likely have lower OC headroom, if any. I don't OC so that part I don't care about.

File: slide-1-728.jpg (95KB, 600x450px) Image search: [Google]

95KB, 600x450px

>>56144594
Bobcat parts were 40nm, Jaguar parts are all 28nm. The clock scaling comes almost entirely from process, though Jaguar does have 1 more stage in its pipeline, they're almost the same. Some Bobcat based parts hit 1750mhz, might be slightly higher but I can't remember exactly. Early Kabini parts didn't reach anything radically higher. The Athlon 5350 clocked to 2.05ghz, and the highest turbo Beema parts briefly touch 2.4ghz.

Its about akin to K8 to K10. They're not radically different.

>>56142357
you = retard

>>56144688
Yeah, that is a good comparison. Not radically different like K10 and BD, but significantly improved evolutionary change.

K10 actually had lots of architectural tweaks that boosted IPC significantly from K8, it came close to Core 2. And secondly, there was similar very important boost in SIMD/FPU peformance. K10 has much beefier FPU with native 128bit vector ops, so basically it'S SIMD capacity doubled and it had massive speedups for encoding and multimedia compared to K8. Jaguar did a similar job on top of Bobcat, but it did better because it added very useful support for SSE4, whereas K10 stayed at SSE2 which harmed it later on.

>>56142363
Sounds perfect for most people. I only ever use the full power of my processor when emulating and not that many people do that.

i5 + 8 cores/16 threads = killer CPU for affordable price.

>>56143095
That picture is disgusting and does nothing to further conversation or any semblance of a point you might be trying to present.

Fuck off back to /b/.

>>56144057
do you seriously believe the law to be perfect

>>56144094
I appreciate the correction

File: 1357238326120.png (1MB, 1280x650px) Image search: [Google]

1MB, 1280x650px

>>56144924
no u

>>56143095
It's one thing to not being able to shit, but you took a proud picture of it? Dude you need to shape up your life.

File: [FFF] Akatsuki no Yona - 01 [BD][1080p-FLAC][E3F40AC4]_0001_00:09:46_[2016_08_12_14].jpg (279KB, 1920x1080px) Image search: [Google]

279KB, 1920x1080px

Any news is good for me. I'd like to finally get more cores, but refuse to buy the shitshow that is Broadwell-E. The 3570k has served me well, it might be time to say goodbye.

>>56145096
Whats the matter, goyim? Can't you afford $1700? Its less than $2500. Who doesn't have a spare $5000?

>>56142363
They literally did a benchmark where an 8 core zen performed as well as an 8 core Core i7 6900K

>8 core 16 threads
Would this affect Skylake-X prices? Because I was planning on getting it next year.

>>56142310
Where were you when Intel suffered the Second Shoah?

I have a 3570k and just want a >4 core with haswel ipc for <$300 with a Mobo I don't need to junk every gen. Anyone who wants Zen to make all Intel's offerings obsolete is retarded. Ask it needs to do is make more than 4 cores standard while keeping prices down. It costs almost nothing to Intel to make a 4 core over an i3 and they know it and have no problem pocketing the margin.

>>56145242
>I have a 3570k and just want a >4 core with haswel ipc for <$300
You niggas are dreaming. I expect 8cores/16threads to be at least $500.

>>56145264
A 6 core for about 300 isn't crazy, I'd bet you could end up getting something like that for a sale.

File: 1467344145637.jpg (8KB, 493x402px) Image search: [Google]

8KB, 493x402px

>>56144924
>mad

>>56145326
6 core intels are going for ~$380. AMD will most likely sell them for $320 or less. So yeah completely conceivable to think so.

>>56144997
AYYY LMAO

JUST GIVE ME SOMETHING FAST SO I CAN ENCODE CHINESE CARTOONS WEBMS REALLY FAST

>>56144997
>>56145410

really cringey attention whore

>>56144057
>should Intel's signing key be leaked (like Microsoft's secureboot bypass key was not long ago), every single Intel machine from 2006 and later is compromised.

Hello Goebbels

>>56145387
Why the hell would they go that low if they finally had competitive CPUs? They're building up the hype and they'll be able to sell that shit for $400 and drop it down to $360 or so later.

>>56145387
I'm hoping the 6 core HEDT will drop down in price just a little bit because of Zen.

More fake hype by AMD's marketing department and the bald drunk Roy Taylor.

We already know from leaks that Zen is only roughly equivalent to an i7-980 performance wise.

>>56145434
No one is expecting them to win core to core, and they will no matter what price them by performance. They will def not be more expensive than Intel but the gap is yet to be seen.

>>56145518
>good goy, pay more for a chip that uses an old microarchitecture and an outdated platform

>>56145560
paid Intel shills are a thing. they are here right now.

>>56145600
I'm getting the Skylake-X. Kaby Lake is pretty pointless anyway.

>>56145615
>getting skylake when zen already cucked it

My body is ready. I'm waiting for the next gen Apu though because I'm not going to be gaymin

File: i-am-poo.png (38KB, 1971x697px) Image search: [Google]

38KB, 1971x697px

>>56143095

File: 1471513693630.png (170KB, 501x506px) Image search: [Google]

170KB, 501x506px

>>56145653

>>56145560
This.

Anyone with a brain knows this is going to be shit, just like everything AMD has done in the last 5 years.

>>56145649
That's more than a year from now, desktop bristol ridges (current gen, new socket) should however be coming out any moment now

>>56145672
>should however be coming out any moment now

6 months from now. read the article in the OP.

File: browse_0097.png (6KB, 506x87px) Image search: [Google]

6KB, 506x87px

I'm waiting for that upgrade.

>>56145682
No, desktop bristol ridge (not zen) is coming out 2016, zen in limited quantity is coming out in 2016 and zen in full production 2017. Zen based apus are late 2017

>>56142446
>40% more instructions per clock
Did they just increase the issue width...? Super scalar has been a thing for years, I cant see any other way to get an IPC boost that high

>>56145763

no, it was pushed back. read the article. any 2016 launch will be in the last week of the year and will be a full paper launch.

>>56145860
>I cant see any other way to get an IPC boost that high

You barely know anything about even simple high level architecture then. The BD family in its entirety has horrid pipeline stalls constantly, it has a relatively long pipeline to begin with, the memory controller is terribly inadequate, and its cache system is abysmally slow in copies and writes in addition to having poor latency.
The general width of the Zen core isn't whats increasing throughput so dramatically.

>>56145383
That picture, like your post, is literal shit.

>DELAYED TO 2017

AYYMD IS FINISHED & BANKRUPT

>>56143403
This could really happen.

Mac Pro is due for a refresh. Last gen Mac Pro exclusively used AMD Firepro.

If the high priests at Apple were happy with the bending down of AMD, then the CPU might just be a Zen Opteron this time.

>>56145905
>Timeframe and Availability

>At the presentation, it was given that Zen will be available in volume in 2017. As the AM4 platform will share a socket with Bristol Ridge, users are likely to see Bristol Ridge systems from AMD’s main OEM partners, like Dell and others, enter the market before separate Zen CPUs will hit the market for DIY builders. It’s a matter of principle that almost no consumer focused semiconductor company releases a product for the sale season, and Q1 features such events as CES, which gives a pretty clear indication of when we can expect to get our hands on one.

>It’s worth noting that AMD said that as we get closer to launch, further details will come as well as deeper information about the design. It was also mentioned that the marketing strategy is also currently being determined, such that Zen may not actually be the retail product name for the line of processors (we already have Summit Ridge as the platform codename, but that could change for retail as well).

That's the only part in that article that talks about timeframes and does a) not support your bullshit b) is mostly speculation by the writer of that article. Now piss off.

>>56146061
>That's the only part in that article that talks about timeframes and does a) not support your bullshit b) is mostly speculation by the writer of that article. Now piss off.

except for the statement by AMD that it won't be available until 2017

zen is paper launching in the last week of 2016, you won't be able to buy it. the real launch isn't happening until mid-Q1. this is most likely because AMD has to respin to fix the bad usb 3.1 controller and clockspeed problems in the A0 chips.

Or I can just get an i7 during black friday and get the real thing for cheap instead of a knockoff

and no AMD compromises like fried motherboards or PSUs

File: nvidia=dead, intel=dead.jpg (306KB, 1500x773px) Image search: [Google]

306KB, 1500x773px

nvida lol

>>56142363
everyone who does not want to pay $350 for a i7 + $200 for a decent mobo

the main issue with zen is heat, overclock potential and if anyone will even use 8 cores.

current i5s will last maybe a decade, overclock well and dont catch fire.

>>56146194
>reposting this WCCFtech nonsense.
There is no bad USB3.1 controller, child. The on die south bridge doesn't natively support 3.1 at all, support is provided by the Promontory chipset from ASMedia. There was no problem with that either, just clickbait fooling 80 IQ subhumans like you. USB3.1 needs a signal booster to eliminate interference, exactly like the issues with USB3 and wifi signals, just amplified.

How can you managed to be so fucking dumb and wrong all in one post?
Are you that same retarded little faggot trying to say Summit Ridge was only going to have single channel memory? LOL

>>56146239

nigga? if its anything lik the 8350/8370 then expect 5ghz reachable.

and i7 6900k is more expensive that threefiddy no?

>>56142310
Got a Xeon e5 1650 0 and all for 270€. Did I do well?

>>56146272
>The on die south bridge doesn't natively support 3.1 at all,

you are incorrect, zen is build to be an almost complete SoC and has the usb 3.1 controller built into the die. if any promontory chipset has it on the motherboard it's most likely because AMD is shoehorning the excavator APUs (which won't have an on die ddr4 memory controller or usb 3.1 controller) in before zen is available.

>>56146276
No one runs FX chips at 5ghz
Because putting 1.5v into it and drawing 200w+ kills it very quickly

>>56146239
>and if anyone will even use 8 cores
As long as you run 8 or more processes at a time I can see 8 cores(or threads) being useful. I mean windows alone is like 30 processes running at all times. Chromium based browsers spawn like 400 processes just for amusement or something

>>56146276
IBM's 32nm PD-SOI was intentionally designed for a certain amount of leakage in order to achieve high clocks since the POWER family makes up for weak perf/clock by hitting 4-5ghz.
14nm LPP is a very low leakage process that provides moderate clocks. Higher than 28nm HPP, but not anything like the aforementioned 32nm SOI.

An overclocked Zen based chip should be able to hit 4ghz if we're seeing 8 core 95w parts hit north of 3ghz stock. Not going to get near that 5ghz mark by a long shot though. If AMD uses the IBM developed 14nm SOI FinFET process for their Zen+ chips then they'll likely clock to the moon.

>>56146350
Wow, its like you really have no idea what you're saying.
Summit Ridge does not have USB 3.1 on die. At all.
Carrizo/Bristol Ridge has a dual DDR3/DDR4 IMC. Its the exact same 3rd party design that was featured in Kaveri.
There is no off die memory controller on anything made this decade. Jesus Christ you're retarded.

>>56146370

dude stop posting.

>>56146405

Alright, thanks for the info.

>>56146239
>heat
Heat is a meme in 2016.
If it's below 90C it's safe. and it absolutely will be because 14nm power draw is low.

What really is limitin clocks now is safe voltage limit. To go faster you need more volts but you don't want to kill it.

>>56142329
T. Pajeet at Nvidia HQ

>>56146405
>Summit Ridge does not have USB 3.1 on die. At all.

[citation needed], every credible source and rumor mill has already reported that it does and that the controller AMD went with ended up being defective.

>Carrizo/Bristol Ridge has a dual DDR3/DDR4 IMC. Its the exact same 3rd party design that was featured in Kaveri.

DDR4 memory controllers were not even being produced when carrizo was in design/validation.

>There is no off die memory controller on anything made this decade. Jesus Christ you're retarded.

Phenom II.

>>56146405
>If AMD uses the IBM developed 14nm SOI FinFET process for their Zen+ chips then they'll likely clock to the moon.

Once again GF fails to produce a decent process and ends up hamstringing AMD.

Based IBM saves the day AGAIN

>>56146455
pajeet is the leader of amd thoooooo

>>56146482
GloFo's 14nm FinFET is straight from Samsung. GloFo Fab8 is running the same 14nm LPP that Samsung has running at 3 different facilities around the world.

Also GloFo now owns the entirety of IBM's foundry business including all their IP, and several hundred employees. They'll be skipping the 10nm node all together and jumping down to 7nm because of that.

Pajeet has wares if you have coin.

>>56146495
Fuck off Rajesh, we all know your posts are worth only one rupee, Nvidia can be quite stingy at times.

>>56146517
>Samshit

:^)

>>56143039
GOD MODE

>>56146446

Ive swapped a fx8350 for an i7
You see 70'c vs 50'c at stock under load
The 20'c drop in temperature output cools the entire room down, the typical user will notice this.

>>56144694
What does AMD's cock taste like?

File: 1464269945452.jpg (347KB, 1000x982px) Image search: [Google]

347KB, 1000x982px

>>56146591
no idea m8 i was hoping you could tell me

File: wewww.png (2KB, 529x29px) Image search: [Google]

2KB, 529x29px

>>56144094
my Haswell i5 isn't compromised at least phew

>>56143152
> not buying latest and greatest instead of feeding my family

Hahaha fuck off

>>56146446
You gotta be kidding me.

I won't even talk about overclocking headroom but something even more tangible.
Room temperature.

What do you happens in the summer?

>>56146628
The joke's on you, I have an AMD card, but I still recognise AMD CPUs as the shite they are.

>>56146628
>they have the most powerful gaming oriented GPU on the market

what's your point again?

>>56146472
There are no credible sources, subhuman retard. There is WCCFtech rumor bullshit, and the issue they're talking about isn't specific to anything. Its a problem with USB3.1 itself. High frequency signaling in poorly shielded cables and connectors.
A signal booster is used to amplify and cleanup the signal to prevent degradation. Thats it. There is no "problem" beyond that.

Kaveri and Carrizo use the exact same 3rd party memory controller from Cadence. You simply have no idea what you're talking about. At all.
The Phenom II chips have integrated memory controllers. They're not off die.

You just keep proving how fucking stupid you are. Its amazing.

>>56146672
Jokes on you, you are a newfag

>>56146636
Check for Intel "active management technology" (AMT). That's the part you don't want

>>56146702
Fag master calling others fags.

File: 6fc09d5672dea5a557e5ab050a92278e.png (1MB, 1440x1493px) Image search: [Google]

1MB, 1440x1493px

>>56146517
yeah but it's GloFo, yields are going to be worse than Nvidia's GTX 480. They were barely able to crank out 45nm chips when intel already mastered 32nm. They had to falsely advertise in some cases with some chips that were specified as 45nm being 65nm instead.

And you think they've just magically managed to catch up to a company like Samsung? Also i know your post is bullshit because 7nm is not possible with Silicon.

>>56146695
>The Phenom II chips have integrated memory controllers. They're not off die.

AMD didn't introduce IMC until bulldozer.

>There is WCCFtech rumor bullshit,

Digitimes are the ones who reported about the USB3.1 defects originally and are about as credible as it gets for news like this. Your damage control is laughable.

>>56146726
they are one in the same, i wouldn't have it if i didn't have VPro

>>56146756
http://techreport.com/review/5683/amd-athlon-64-processor

Athlon 64s had an integrated memory controller. How fucking retarded can you possibly get?
Digitimes is a giant Korean clickbait blog, exactly like WCCFtech except they aren't Indian.

Have anything else you want to be wrong about today?

File: 1456647016105.jpg (90KB, 477x616px) Image search: [Google]

90KB, 477x616px

>>56146755
>And you think they've just magically managed to catch up to a company like Samsung?

They literally purchased the technology from Samsung.

>>56146756
uhh dude integrated memory controllers were a defining feature of the Athlon 64 series. AMD hasn't made a chip without an INC since then

>>56146756
i think it's time for you to stop posting

>>56146822
That doesn't mean they're going to implement it as well. They're still running the same poo in the loo tier AMD fabs they purchased back in 2009.

>>56146755
>GloFo, yields are going to be worse than Nvidia's GTX 480
GloFo yields on 14nm are actually good, TSMC are the ones with the shit yields

File: 46ec1cb2f9a197c530deb8185fe8ae61.jpg (132KB, 838x734px) Image search: [Google]

132KB, 838x734px

>>56146756
Athlon 64 from 2003 was the first chip to market with a IMC.

>>56146803
>Athlon 64s had an integrated memory controller. How fucking retarded can you possibly get?

that doesn't mean it's derivatives had IMCs, you idiot. the platform switch to AM3 was explicitly to add the ddr3 memory controller, because the previous AM2 chipsets didn't have it and AMD didn't build it into the chip (presumebly to reduce die size and improve yields or something).

>>56146803
>Digitimes is a giant Korean clickbait blog, exactly like WCCFtech except they aren't Indian.

>any source that refutes my bullshit is a clickbait blog!

>>56145159
That means the price will decide whether they succeed. If they get too greedy and price it at $600 for example then intel will drop their prices and outcompete them.

>>56145159

in a closed environment where the people invited were not allowed to verify the test system specs, clock speeds or test results.

maybe if they can pull this off in an independent test then i'll believe it, but after the clearly faked poolaris demos i'm skeptical to say the least.

>>56146871
>If they get too greedy and price it at $600
if they're delivering broadwell-e or even haswell-e performance with their 8c/16t cpu's, $600 is super reasonable and not greedy at all.

>>56145723
>5400+
>tfw I got a 5600+

>>56146866
>that doesn't mean it's derivatives had IMCs
Yes. Yes it does. ALL of AMD's CPUs after it had integrated memory controllers. Every single one. All of K8, K10, K10.5. Everything.
The difference between AM3 and AM2, among other things, is that AM2 chips used a DDR2 controller. Still an integrated memory controller.

Just admit you're trolling right now. It would be unheard of for anyone to actually be this clueless yet so adamant.

>>56146922

You are assuming that AMD was or has been competent in the post-K8 era. It's completely believable that they'd make a step back like that.

>>56146944
I'm not assuming anything, sperg. I know for a fact every single CPU in recent history has an integrated memory controller.

Just give up.

>>56146966

They clearly didn't.

>>56144265
http://www.overclock.net/t/1578480/i5-2500k-4-5ghz-vs-6700k-4-5ghz-in-games

Bottleneck

File: 1471334902806.jpg (33KB, 544x517px) Image search: [Google]

33KB, 544x517px

>>56147089
>bbbbbbbbbbbut best processor of all time

They should kys themselves.

Neat, stock went up again.

>>56146446
>heat is a meme
say this to alle hese companies that moved thier server farms to iceland

>>56147132
Not really a fair comparison. Different types of RAM and i5 vs i7.

>>56142363
I do.

File: 1454952221811.jpg (68KB, 720x552px) Image search: [Google]

68KB, 720x552px

>>56147203
>compare the best processor of all time with a regular 2015 processor
>unfair

>>56147203

this. the lack of the extra l3 cache and SMT can easily be hurting the 2500k in this test, not to mention that sandy vagina can usually clock higher than 4.5ghz.

>>56142962
(10/14)^2 ~ 1/2
10nm is a full node, though it is likely only Intel will do high power on 10

File: 1421931186827.jpg (26KB, 600x375px) Image search: [Google]

26KB, 600x375px

>>56147203
>>56147234

>>56147234
They should also use something other than a 970 because it's possible the vram usage went over 3.5gb for some of the sandy bridge tests. I'm not saying it's likely but anything is possible and you should eliminate as many variables as possible for an accurate comparison.

>>56147250
Only on paper, we all know 10nm is gonna have a good amount of the backend closer to 14nm

>>56147271
that would still be a full node from TSMC/Samsung/GLFlo 14nm/16nm, since those all have 20nm BEOL

>>56142310
Am i the only one curious about what the 4 c/ 8t preformace will be?

or the 6c/12t will be when they release it?

>>56147198
cooling has nothing to do with that, iceland just happens to have stupidly low cost geothermal electricity

>>56142348
Isn't that the same shit that Steve jobs said that basically meant fuck all?
Buzziest of buzzwords.

>>56146405
>There is no off die memory controller on anything made this decade. Jesus Christ you're retarded.

I was literally working with a 775 motherboard this week that had a off die DDR3 memory controller.

>>56147250
I really wish you weren't seriously, but unfortunately I know you are.
The area scaling metrics of a process node have literally nothing to do with your middle school napkin math.
10nm is a half node down from 14nm and 16nm. Front end feature size barely decreases.

GloFo just had a rep speak out on this, they're skipping 10nm entirely because its a half node.

>>56147361
Isn't 775 from 2005~
That isn't this decade now is it?

>>56147361
>775 motherboard
>this decade
wew lad
unless you think this decade DOESN'T mean 2010-2016 ever since we got past 2015

>>56147361
AMD has an IMC on everything since athlon 64, including phenom and bulldozer.

Intel has had it since nehelam, the core2 series came after the a64 but still had to use it's fsb to access the memory controller. It was the last big x86 processor without an imc, it was also on lga775, stop being a twat.

It's always the same with AMD. Promises that eventually have the CPU performing worse than a 5 year old mid tier Intel.

Makes me sick, because I'm beginning to hate paying over the top for Intel.

File: Newest Benches.png (119KB, 1280x720px) Image search: [Google]

119KB, 1280x720px

Can't wait for it to be released and /g/ going from /g/pu to pic related

File: hector-fx.jpg (19KB, 250x318px) Image search: [Google]

19KB, 250x318px

>>56147373
>GloFo rep
lol

Sure foundry labels are somewhat meaningless, but what evidence do you have that 10nm does not offer 0.7x scaling? Glofo not doing a high power process on 10nm has nothing to do with it being a half node. You are probably one of those dumbasses who take Semiwiki seriously.

>>56147528
GloFo has the entirety of IBM's foundry business and IP.

The rest of your post is worthless. CPP, metal pitch, and SRAM cell size are easy enough to find for TSMC and Samsung's 10nm nodes. They've been out there for a long time.

>>56146370
I run mine at 4.5Ghz no problem, I think 1.35V or something.

>>56142446
Oh look, it's AMD making yet another bullshit promise.

>>56146793
Oh interesting, the 3960x seems to be safe too, if zen bombs I'm going to get me one of those.

>>56147700
Enthusiast hardware up to LGA2011 are exempt from it. Xeons might not be though.

>>56147610
You are clearly a moron if you think SRAM scaling has anything to do with logic scaling. Samsung's numbers for CPP and metal pitch show full node scaling, as do TMSC's number in a relative comparison to their 16nm prcoess, so what are you talking about?

>>56147841
>I don't know what I'm talking about so I'll just say things

Thats nice. Anything to save face with your napkin math.

>>56147898
You have yet to say anything of worth or rebut any of my posts. If 10nm is half node like you say then show some evidence. Glofo reasons for skipping 10nm has nothing to with 10nm being a half node.

>>56143238
http://wccftech.com/amd-zen-summit-ridge-launch-q4-2016/

An earnings call is a once a quater call made from the CEO of CFO of a company to the investors of that company. Durring it AMD stated that Zen based Summit Ridge FX CPUs will be launching in late 2016.

Earnings calls arnt just off the cuff remarks, they write them out ahead of time and basically read them, then usually do a question and answer.

anyways you're wrong, nananabooobooo stick your head in doodoo

>>56147972
Thats pretty pathetic. Using a source that proves you wrong and thinking you made a point.
2017, first full year of revenue for Zen. Limited availability in 2016.
You only supplied evidence supporting that. Its the exact same thing restated at Computex not long ago with Su explicitly stating everything is right on track.

All of this falling directly in like with AMD's internal roadmap that leaked last year. Final silicon slated to be in production this October.
No Zen part has ever been delayed.

>>56147972

"AMD CEO Lisa Su responded to questions about the company’s upcoming Zen microprocessors during AMD’s quarterly earnings call. Although it appears some of the Zen CPUs will ship in Q4 2016, most people won’t have access to them until 2017."

From: http://www.tomshardware.com/news/amd-zen-launch-2016-2017,32314.html


I understand you're a fan boy but for come on get some credible shit. You just look retarded right now.

Is it gonna be with a new socket or with AM3+

>>56148492
AM4

>>56148492
It's all on AM3+ chipset and all.
Zen is just Bulldozer on 14nm after all

>>56148492
Why do dumbasses keep asking this?
Yeah, a brand new high performance CPU is going to launch on a 5 year old socket.

>>56144997
>Chelsea what is that?
lyl

>>56142421
>>56142392
I love this picture so much.

File: 7tJEAx6.jpg (138KB, 720x960px) Image search: [Google]

138KB, 720x960px

>>56142329
>>56142357
>>56142363
>>56143175
>>56145653

Look, retards!

https://www.youtube.com/watch?v=Hr-5NM6cO8g

>>56147388
that's like saying PCI-E isn't modern because it's from 2004
775 was only replaced around '09/'10

>>56147361
He's right if he meant AMD only. Intel was pretty late to the on-die memory party.

>>56149381
2004 is 12 years ago. Over a decade old.
Do you have any idea what this decade means?

>>56149511
i'd say "2010" would count

>>56144265
Biggest problem is the motherboard.

>>56143306
i5-3570k here, running at 4.6ghz since 2012. Still a beast, still runs cool. Fuck AMD.

whats "40% IPC over Excavator" equal to? Haswell?

>>56150216

Nobody knows for certain because those excavator chips don't have an L3 cache.

>>56146894
Source?
You better not be talking about AoTS with crossfire. Because 2 480s do perform like that on that game. There was no faking involved. Why do people always have to,spread misinformation because they bought a certain product?

>>56148652
>3ghz
if zen comes out at 3ghz base it will be a laughing stock.

>>56150246

Source is a blender test done on stage.

https://youtu.be/oQS8s7TOXsE?t=87

>>56150175
Nice. Would 3.4ghz on it bottleneck any games in 2016?

>>56150273
Broadwell-E i7 6900k (Intel's highest clocked 8 core part) has a base clock of 3.2GHz. The Zen CPU shown is a engineering sample which are clocked lower than the final product.

>>56150300
4x accelerated video of the test played on stage.
ftfy

both processors clocked at 3ghz (intel underclocked, zen ?) and i suppose both with turbo disabled.

>>56150273
What do you expect? It's a 6 core processor. Everything by intel save the 6700k is 3.4ghz or less, especially haswell/broadwell-E.

For example, the i5-6400, 6500, 6600, i7-6700 are clocked at 2.7, 3.2, 3.3, 3.4 ghz respectively.

The 5820k is 3.3 (0.3ghz boost). If Zen is clocked slightly lower with a comparable or higher boost with a much smaller tdp (5820k is 140W iirc) for about $50-100 off then it'll sell like hot cakes, specially since the motherboards would be a steal too given the exorbitant X99 mobos (with aging features)

>>56150372
>The Zen CPU shown is a engineering sample which are clocked lower than the final product.
Not sure anymore, the 14nm process they're using is hindering clock frequency.

File: 1456681584195.png (223KB, 644x580px) Image search: [Google]

223KB, 644x580px

>>56150432
Actually the Zen engineering sample shown is a 8 core part. Intel's highest clocked 8 core is at 3.2ghz and cost over a thousand burger bux.

>>56143095
This is actually the first thing I thought when I saw the official logo. Quick, someone shop "ZEN" in the center.

>>56148652
Is that the new AOTS?

>>56146370
8320 @ 4.93 1.34v here been running like that since 2014 silicon lottery i guess...

>jim keller left
AMD dead. it's Athlon all over again.

>>56150711

That is one hell of a silicon lottery win.

>>56146370
You are full of shit.
I ran a FX-8150 @ 4.8GHz for 4 years at 100% (rendering) load before replacing it with a Xeon 2670 V3.

>>56150216
In theory? A little behind Skylake.
I have very low expectations. If they can beat Ivy, then I'm switching fast.
>tfw make -j16

>>56143678
What's the x axis? Time?

>>56150362
Probably so

>>56142310
>dual schedulers
>still late

>>56146979
a r e y o u f u c k i n g r e t a r d e d

De-integrating the memory controller would contribute to a literal 70% performance loss immediately. The entire reason it was integrated was because the performance gains from having vastly lower latencies far outweighed the small amount of die yields and die space lost to it.

No modern chip past 2003 has a de-integrated memory controller. The trend of all electronics is integration, not the other way around.

Fuck, you can't even do SoC without having full system integration on chip.

>>56149381
PCIe's not modern. It was superseded by PCIe 1.1 in 2005, That superseded by PCIe 2.0 in 2007, and so on.

>>56153915
>No modern chip past 2003 has a de-integrated memory controller.

Except for DDR3 supporting LGA775 motherboards and Phenom II.

>>56151083
For workloads that are like, -j16, Zen *ought* [i.e. wishful fingers crossed thinking] to get stuff done faster. e.g. that rendering demo is fairly promising

>>56154375
>not still using AGP

Where the fuck are my AGP 1060s?
I'd at least expect AMD to release AGP cards for the poorfags

>>56154574
My laptop uses AGP

>>56153915
>No modern chip past 2003 has a de-integrated memory controller
False.

>>56143712

>AotS

It's sad that AMD is bad they've resorted to using their biased GPU benchmark for even their CPU now.

>>56154565
The rendering demo is completely UNPROMISING.
It's actually pretty fucking doomy. Think about it; AMD are going to cherrypick the best benchmark they can find to make their product look good, so what you expect is that it completely stomps whatever it is up against because they're picking the fight.

The rendering demo is just the first sign of the disappointment yet to come.
If THIS is the best benchmark they can find, they're in trouble.

>>56150432
>tfw was thinkig ablut buying a 5820k
>zen looks promising

The wait never ends

>>56142507
faggot

File: amdumb.jpg (543KB, 1100x1002px) Image search: [Google]

543KB, 1100x1002px

>>56142446
>AMD makes some bold claims.
>Believe the hype anything.

Jesus Christ, you retards deserve to lose all your neet bucks to Pajeet.

>>56143306
>>56144383
>>56150175
3470 checking in ;)

>>56142363
Last years i5 is 6600k

File: 500x1000px-LL-33a7870d_original.jpg (22KB, 500x281px) Image search: [Google]

22KB, 500x281px

>>56142518
Boy are you in for a disappointment.

File: 1465703534819.jpg (31KB, 480x547px) Image search: [Google]

31KB, 480x547px

>>56147203
>>56147234
>excuses