i9 released amd street shitters BTFO

>>60813744
E-euhm w-what does this mean a-anon?

>>60813768
School's out.

>>60813744
>i9
let's just make MORE cores on chip and represent it like a fucking innovation. Pathetic intel.

>>60813788
Does this mean I can plan minesweeper at 30fps?

>>60813788
im in collage idot

>>60813807
youre saying amd didnt do that before?

File: TB2f4vIvrJmpuFjSZFwXXaE4VXa_2341477237_1.jpg (470KB, 1060x1569px) Image search: [Google]

470KB, 1060x1569px

>>60813744
Even the 18-core Skymeme-X can't crack 3,000 in Cinebench R15, meanwhile Zen scales at 1.99 from 8c/16t to 16c/32t so the top 16-core Ryzen is going to score around 3,100.

Skymeme-X is DOA.

>>60814001
The CPU doesn't go on the GPU
>>60814036
fake

File: dog2.jpg (21KB, 237x233px) Image search: [Google]

21KB, 237x233px

>>60813744

I'll let you into a little secret:

They have been BTFO'd since Pentium 4.

File: ebin.png (2MB, 5098x1500px) Image search: [Google]

2MB, 5098x1500px

>>60813744

File: 1496109435593.png (234KB, 882x758px) Image search: [Google]

234KB, 882x758px

>>60814036

File: pHrnT.jpg (184KB, 980x602px) Image search: [Google]

184KB, 980x602px

>>60814046
L M A O
M
A
O

File: 1460233502623.jpg (38KB, 270x270px) Image search: [Google]

38KB, 270x270px

>>60814097
>vr on the box

>>60813744
>max tdp: 308W
Yeah no thanks

>source
Some says ago the benches were posted here

File: single socket holocaust.jpg (28KB, 550x550px) Image search: [Google]

28KB, 550x550px

>>60814049
lmao.

File: 1496181076435.png (973KB, 801x1500px) Image search: [Google]

973KB, 801x1500px

>>60814197
a pipeline long enough to deliver every jew into the gas chamber.

>>60814174
>max tdp: 308W
its the full system, mr. el retardo

File: 1496500004156.jpg (269KB, 960x686px) Image search: [Google]

269KB, 960x686px

>>60814217
running with integrated graphics?

oh wait
>no igpu

>>60813744
Literal garbage

>>60814239
why do you change the subject now? What has the igpu to do with it now? whats your point?

>>60814256
my point is that it's impossible for an i9 HEDT system to consume merely 300W.

File: CORE_I9_F31652E9_DO_NOT_DISTRIBUTE.webm (975KB, 852x480px) Image search: [Google]

975KB, 852x480px

>>60813744

File: Mysterious Merchant.gif (4KB, 452x523px) Image search: [Google]

4KB, 452x523px

>>60814217
PLS BUY INTEL. PLSSS

>>60814273
Why would it be impossible? The 308W TDP measurement proved your point wrong.

>>60813744
I'm an Intel guy but these i9s aren't competitive.
Intel had these planned but they weren't ready for ryzen to actually be competitive

It's a reaction to a problem they have no solution for.

>>60814299
I said "system", which includes power consumption of additional components vital to running a computer - like cooling solution, storage and some GPU.
also, 300W _TDP_ is impossible, the whole chip would go out with flames if we were talking TDP.

>>60814329
He himself said its not good yet, because you can break some glued SMDs off by deliding.

>300W _TDP_ is impossible, the whole chip would go out with flames if we were talking TDP.
But I said its the whole system, not the chip, are you really that dense?

File: you didn't need that kidney anyway.jpg (33KB, 472x616px) Image search: [Google]

33KB, 472x616px

>>60814317

>>60814335
i7 6950x drains 240W+ overclocked

>>60814335
>trying to save face this hard
topkek mate

>>60814335
>>60814362

>>60814362
now i'm pretty lost, >>60814299 talks about 300W TDP and you're going on about overall consumption. But if we're talking whole system draw then it's probably just set to highest possible clock while not doing any workload at all.

>>60814371
is that on LN2?

>>60814033
Like you said, before, intel its doing that at this moment

>only 44 pcie lanes
toppest kek m8

No one cares about your meme cache drive that takes up precious ram space.

Do these turds even have a soldered on integrated heat spreader?

Also why get this over say a 16-core threadripper that costs $850?

>inb4: "MUH GAYMES!!!111ONEONE"

>>60813807
>amtards complaining about MOAR COARZZZ XXDXDX

File: 1496645419629_copy.jpg (929KB, 5098x1500px) Image search: [Google]

929KB, 5098x1500px

>>60814978
LOL

>>60814978
Intel's attempt at MOAR COARS is a joke. That's why.

>>60814364
Nice normie meme.

Intel should have doubled down on being a gaming only chip
make the coffee lake 700 series still a 6 core though because I want devs to eventually be optimizing their software for it

You can't win a more cores game against a modular system who's performance is consistent at any size

Actually the rumor of the i9 started circa 2009, some websites had a banner showing i5 i7 and i9 with box art and price in euros.

>>60813744
I'll never buy an Intel CPU, you will never, ever get my shekels.

>>60815406
>Intel should have doubled down on being a gaming only chip
Problem there is there's really nowhere for Intel to go. 14nm++ isn't going to be enough of an improvement over 14nm+ to even bother with. They can't afford useless upgrades anymore.

>>60813744
The heat that even the 10 core puts out is unreal. A 240mm AIO has a tough time keeping it from throttling in fucking cinebench.

AMD's JEWRIPPER processor is about to put kikes in their place.

>>60815644
Kaby Lake is 14nm++. Next up is 10nm, which according to Intel's own literature will be a downgrade from 14nm++.

>>60815778
B-but muh BEOL scaling, goyim. Buy our 10nm CPUs.

>>60814316
fuckin kek man

>>60815778
I thought Kaby was 14nm+.

>>60814316
Is that der8auer? The area under the IHS is totally flooded with Thermal Grizzly Kryonaut thermal paste to prevent cracking with extreme (LN2/LHe) overclocking.

>>60815844
Yeah it's some OCer he's even wearing a HWBot t-shirt.

>>60814036
>1.99
Can you explain how does it happen? is epyc going to be the same for the 32 and 64 core variant?

>>60816022
EPYC is 3.90 afaik.

File: amd-zen-core-scaling.jpg (33KB, 597x430px) Image search: [Google]

33KB, 597x430px

>>60816022
Essentially, Zen's MCM design scales much better to large numbers of cores and threads than a bingbus design does.

>>60815926
I hope they benched sinebench :)

File: pissed.jpg (73KB, 600x600px) Image search: [Google]

73KB, 600x600px

>>60816046
>3.96 and 7.90
How in the fuck
DID THEY DO THAT?

>>60814362
he saidthe problem is the glue sometimes gets on the smd.
nothing he can do about it unless he heat it up.

>>60816070
Keep in mind less than 3.96 will be realized as the 32c Epyc parts will have lower clocks. The scaling is virtually linear Ryzen to Threadripper because the rumored clocks are virtually identical (3.5/3.9 vs 3.6/4.0).

>>60816046
>>60816070
>>60816105
Those numbers will be at a fixed clock rate.

File: help.png (6KB, 350x314px) Image search: [Google]

6KB, 350x314px

>>60816105
>these clocks for 16 cores
This kills the kike. It's HEDT platform that's also good for gaymen.

>>60815835
Nah. Haswell was 14nm, Skylake (and maybe Broadwell, idk) was 14nm+. Kaby Lake is 14nm++. Cannon Lake will be 10nm.

>>60816120
will they rise up when they get slapped into water-cooled mobos? ony servre-grade liquid cooling i saw was at linus's techtips and that was pretty much jerryrigged

>>60813744
>i9
i11 when?

>>60816141
what do the '+' mean? an improvement on the die process?

>>60816141
Haswell was 22nm Tri-Gate you dumbfuck. Broadwell and onwards were 14nm.

>>60816161
Yes. It's Intel's marketing meme for something so ordinary as base layer respin.

>>60816164
Fuck me, you're right.

>>60816161
Yes, it's an improvement of the process.

>>60816161
Yeah just process improvement.

>>60816120
Yes, I am aware the chart is at a fixed clock rate, I was pointing out that since Epyc 32c parts will have lower clocks than Ryzen or Threadripper, that it will not perform at 3.96 of an 1800x

>>60816143
There's no technical reason, given that ryzen, threadripper and epyc all use zeppelin, for them to clock any differently assuming they can be kept at the same temperature and supplied with enough power.

>>60816175
There is no reason why 32c Epyc wouldn't run at 4GHz all core as long as you have the cooling and the board can supply the power, though.

>>60816218
>>60816211
I guess whoever wants the most out of it will have to plurge out iquid cooling and perhaps a few hundreds, if not thousands in their electric bill.

>>60816218
It'd have to be watercooled, the TDP would be like a graphics card.

>>60816236
But imagine the power of 32 Zen cores at 3.5-3.9 ghz.

>>60816236
There are 350+w TDP air coolers tho. Should work.
The large IHS should help a lot.

>>60816247
Over 6.5k on cinebench.
Blender render demo in 5 seconds.

>>60816268
You shouldn't have to buy a dual fan CLC just to keep your processor from melting under load at stock clocks.

>>60816287
At least it's still in realm of possibility. You will never clocks 32 Intel coarz to 4ghz without socket fucking melting.

>>60816331
>>60816331
why's that?

>>60816046
The MCM design has nothing to do with performance scaling; it allows AMD to make these processors dirt cheap. The technology that enables them to use MCMs with such good scaling in Infinity Fabric, which we know almost nothing about. MCMs have been around a long time and never really afforded this level of scaling.

I just hope that after Intel throws billions of dollars at reverse engineering IF, kids on /g/ remember that it was AMD that pioneered the tech, just as they've done several times before.

>>60816355
Hey, it took Intel 5 years to launch their HyperTransport clone. Now we have to Wait.

>>60814212
underrated kek

>>60816347
There's a limit on what a pin can deliver.

>>60816218
>>60816236
>>60816247
>>60816268
Processors like that are unlikely to ever happen. While it's possible, the board power delivery and cooling would end up being prohibitive against the cost of the processor itself; that is, it would be more economical to just buy two processors at lower clockspeeds to get the same compute power. Remember that it's overall compute power that counts outside the consumer market, not IPC or clockspeed. Efficiency is also important, and Ryzen is very efficient at lower clockspeeds, which will be compelling for datacenters that may actually have thermal or power limits.

>>60816431
Yes, it was a joke, no one's insane enough to clock 32coarz to 4ghz.

>>60816445
if i had the money i would jsut for shits and giggles.

>>60816451
You could use it to power the ovens for Intel execs.

>>60816445
How about 4.2GHz?

http://hwbot.org/submission/2835735_masterpwnerbob_cinebench___r15_2x_opteron_6272_2296_cb/

File: fag.jpg (20KB, 400x400px) Image search: [Google]

20KB, 400x400px

>>60816451
You know LTT is going to do it eventually anyway.

>>60816468
>faildozer "cores"
Nah.

>>60816347
Big monolithic intel die yields aren't stable enough to clock high with high core counts, and even if they were better their architecture and process isn't as efficient at <=4GHz.

>>60816445
Sure they are. The CPU should be unlocked like all AMD CPUs. Just need a board that supports it.

>>60816501
Nobody is going to make an enthusiast overclocking board for a server chip that three people would buy to make a youtube video about extravagance. Getting the processor to 4ghz isn't the issue; it's preventing the motherboard traces from melting when you start dumping a few hundred amps through them. There's a reason almost all server chips are sub 200W.

>>60814978
Nobody is complaining about MOAR CORES.
They are complaining about MOAR RINGBUS and netburst tier housefires.

File: jimkeller.png (105KB, 255x345px) Image search: [Google]

105KB, 255x345px

>>60813744
Very nice intel

HOWEVER

>>60816233
Electricity is less than 3 cents per kWh. A 300W CPU under full load 24/7 will only cost $6 per month

>>60813744
Threadripper vs Wallet ripper

>>60817140
>netburst tier housefires.
tell me that story

I can only imagine how managers at Intel push these ridiculous fucking changes through and the engineers just die inside.

On the other hand they sold their soul long time ago.

I don't understand how anybody can argue single thread performance on these i9 chips. For the price, you'd be better off building 2 systems, one threadripper, one 7700k.

>>60816141
You missed "coffee lake". Skylake is 14nm, Kaby Lake is 14nm+, Coffee Lake is 14nm++, due out in feb 2018.

>>60814075
What a fucking disgrace

>>60814075
wow.. less than 10x scaling from 10 cores?
My 6 core gets like 7.5x

>>60813744
A binned xeon *yawn*

>>60814299
What is reading comprehension?

I'm sure you're a jewtel goyim

>>60817140
MOAR BORES
MOAR BREADS
MOAR BINGBUS
MOAR BIPELINES
MOAR BIBAHURBS
MOAR TDB
MOAR BOWER DRAW
MOAR EBIN :------DDDDDDDDDDDDDDDDDDDD

>>60817582
Threadripper vs. Socketmelter

>>60818272
Intel's scaling is so abysmal that by the time you hit 18 cores, it can't even break 3k cinebench. Threadripper 16 core is expected to hit 3100-3200.

>>60818661
but what's worse is that it's 20 threads.
You expect at 15-50% gain from hyperthreading.

That's really abyssmal. Intel should not even be attempting over 8 cores. Their arch can't handle it and it's embarrassing to show how bad their multithread scaling is.
Remove bingbus and bring back bonger bibeline. That made more sense than this garbage.

>>60818661
I'm not even sure why people care so much whether the top Threadraper will be a few percent higher or lower than the top SKL-X.

A TR4 setup will be at least roughly comparably fast in integer loads, have more PCIe lanes, have guaranteed forward socket compatibility for the next three or four years, AND cost a fuck-ton less.

The only arguments for Intel HEDT are you absolutely need max AVX crunching or are allergic to money.

>>60813744
Wtf you need it for VR?

>>60814036
Thread ended here.

>>60814075
To be fair here, that one was using a hyper 212 cooler without a fan attached to it. Shit was throttling like a motherfucker

>>60817555
Closer to $.11 average in America

File: we must eliminate the jewish presence in cpu fabrication.jpg (306KB, 634x1223px) Image search: [Google]

306KB, 634x1223px

>>60816070

JIM KELLER

JEW KILLER

>>60821133
No, that was using some dual fan rad AIO, I think it was corsair?

>>60821162
There were 2 different "benchmarks"and both of them throttle like a motherfucker, one because of the hyper 212 not having a fan attached to it while trying to push 4.5 GHz, the other one because you trying to push 4.5 GHz on a fucking to fan radiator, that is using a thermal compound instead of solder. That shit was gonna throttle no matter what. And if you think it 2 fan radiator for 10 core CPU pushing 4.5 GHz is enough Without throttling like a motherfucker, go fuck yourself, you're too fucking stupid to post here.

>>60821156

CONTRACT KELLER
>captcha: henr payback

>>60815675
It's using Intel thermal compound to, if this is going to be in the final product or if they're gonna solder the final product is up in the air. I personally think they're going to solder it in their own using crappy thermal compound for the engineering samples to kill the engineering sample aftermarket.

Granted it's gonna be hilarious if Intel's doesn't solder the CPUs.

>>60816474
Linus will be given one for free.

>>60815675
On a side note they were pushing at the 4.5 GHz, if I remember right there is not a single 6 core plus Intel part that does 4.5 GHz not even there 6 core shit. It is not able they hit those speeds before it hits the thermal wall even if you delid the shit and use bare silicon.

>>60821184
Of course they were throttling, that was the point.
Point is you need cooling that costs more than your motherboard to even stop the throttling if you wanna OC thanks to Intel chimping out on solder.

You didn't need that for a 6950X overclocked to 4.2/4.3

>>60818355
You think the consumer 18 core is going to be been, you are fucking retarded.

>>60821219
Formulate an English sentence.

>>60821213
Again I personally think Intel's only putting the thermal interface and engineering samples to kill the engineering sample aftermarket. If they keep doing this on the consumer version there fucked, no one is that stupid.

And no you dense mother fucker to fans on is 10 core going to 4.5 GHz is not enough this is in the point of AMD's better it's a point of you can't cool it regardless of who makes it. In a revision or 2 AMD might prove us all wrong but they'll also be on 7 nm at that point.

>>60821234
>>Again I personally think Intel's only putting the thermal interface and engineering samples to kill the engineering sample aftermarket. If they keep doing this on the consumer version there fucked, no one is that stupid.

Debauer works with Intel to make the deliding tool specifically for Skylake-X, said tool would no longer be promoted or needed if solder was there.

It's soldered, live with it.

>>60821226
binned, so word got transcribed wrong, is my point in valid or do you honestly think Intel is going to give you a chip that they used to be able to sell for $6000 for 2000? I'm skeptical of the 18 core will ever see the light of day, and this is in just a halo product that they can shine a light on same were better than AMD here and never actually put it out.

>>60821254
Skylake-X are HCC(12-18) and LCC(6-10) core Xeons, with features disabled.
Have been for years, and will remain because there's no reason for Intel to make dedicated HEDT silicon.

>>60821252
People have been deliding soldered CPUs for a long fucking time, there are a lot harder to delid, and there's far less of a reason to delid them. It gives you less than 2% better thermals to delid a soldered CPU but to some people that's more than worth it, there retarded, but they still do it.

Intel actually has competition now, Intel's old E line i7's were soldered, the 6 8 and 10 core were soldered, there's no reason for me to believe that they won't be soldered going forward other than it's funnier that way. Again Intel has competition why would they need Themselves unnecessarily? I fully believe that this is just a way for them to try and end the engineering sample aftermarket, same with the RFID chip, I believe that's a feature that's only in the engineering samples just as a way to more thoroughly track where they're leaking from.

If they make it the consumer products I would be shocked, and on the floor laughing, quite possibly to my death, but I don't believe that to be true.

>>60821303
These chips were made far before they knew they'll get competition again, and it's pretty evident anything over the 10 core is a reactionary knee jerk to AMD, which they didn't believe will offer competition.
Intel has gotten its head so high up in the clouds it hasn't noticed AMD eating its legs and pretty much everyone is saying this

>>60821289
binned means there good, the best of the best. Consumers are going to get that chip.

I also don't believe that the gonna put the 18 core to consumers, they are gonna give us a very limited amount if they give us any at all considering their pushing the release date back as far as they possibly can. I honestly think they're going to cancel it, and maybe only have the 14 or 16 core it the most.

>>60821337
That may be true but I still don't believe that they won't solder it. It would just be so fucking stupid for them to not do that, especially now that they're in reaction mode, you know oh shit AMD actually has something good for change mode.

If there oh shit AMD has something good for change mode is let's use our jizz for thermal compound then they deserve everything they're going to get, but I just don't personally believe that that's the case because no company is that suicidal.

>>60814036
>>60821289
>>60821346

Intel planed up to 12-cores until AMD lit a fire until their ass with threadripper, they and to paperlaunch 14/16/18-core models, even mobo partners had no idea about them and say those chips won't be ready until next year

>>60821363
Intel has saved over 30 million USD by moving to TIM from solder, for a bean counter company those are bigger profits than a few whiny HEDT users who overclock(most don't)
For Intel to consider soldering again, they'd need to retool their assembly line, which I don't see happening for a company that are literal penny pincher.

>>60821366
From my understanding Intel planned 10 cores
then AMD's 8 cores were so close to Intel's 10 core speeds that they added a 12 core
then Threadripper got leak and Intel decided 14 to 18 cores are also necessary

The whole fucking SKU was a cluster fuck at this point, leaving nobody satisfied.

And this is without mentioning the i5 and i7 that are going to be on this chipset to, I don't know where they fit into the equation.

>>60821385
Honestly they just be needing to use the same assembly line they had for every other extreme addition to this point. Unless they completely got rid of it.

>>60818974
Even intel shills know intel's HEDT is DOA

>>60821406
Intel 12 core might do very well against AMD's 8 core.
But Threadripper 16 core is almost surely going to completely shit on Intel's 18 core.

The 18 core Xeon is what.. 2.7GHz base, 3.7GHz boost on 2 cores?
The TDP on them is too bad. The multithreaded scaling is too bad.
A Zen 16 core is going to look like a 20+ core in multithreaded performance.
An Intel 18 core probably looks like a 16 core.
Infinity fabric and AMD's SMT is just too good compared to ringbus and HT when you get into high core counts.
>inb4 CPU-Z gets
>>>>>""""''''fixed''''
>again for this

>>60821474
https://www.youtube.com/watch?v=oNLVWDkx2NY

>>60821487
Intel's 12 core CPU could very well cost the exact same amount is AMD 16.

>>60821500
16/32 ThreadTripper is 850$.
Intel's 8 core is already 1k$.

>>60821523
It was 700 to 800 for their new one, 1000 for the 10 core and I can't remember the rest of the prices.

>>60821487
>The TDP on them is too bad. The multithreaded scaling is too bad.
Now just guessing the bloodbath that Epyc will cause, shits will be wrecked

>>60821500
the 12 core is $1200, the 16c threadripped will probably be less, maybe the same, but definitely not more.

14c threadripper will definitely be under $1000 though.

>>60821487
>The 18 core Xeon is what.. 2.7GHz base, 3.7GHz boost on 2 cores?

No, it's gonna come on a new socket that lets them use 205W TDP chips, it will be clocked at 3.1-3.2 base and probably 3.5 all core turbo.

Too bad it will be 10% faster than a 16 core threadripper and use twice the power while costing 120% more

>>60821533
They mutilated their new 8 core by cutting down PCI lanes, bloody penny pinchers

>>60821562
This, their actual older gen 8 core is now better than their new one if those anemic 1-2% IPC improvements are true, and looking at the leaked benches they seem to be true.

>>60821562
they had to make market segments somehow, and they sure as fuck aren't going to do it by making a compelling product.

>>60821544
Intel's fab doesn't allow for cheap manufacturing, AMDs does.

File: x299-3.png (328KB, 1286x719px) Image search: [Google]

328KB, 1286x719px

>>60821533
>>60821544

>>60821608
100$ less than I thought it was... all honesty, between amd 8 core and intel, I would go intel if I was in the market for the 1800x, but I wasn't so I got the 1700.

File: IMAG2419.jpg (608KB, 2688x1520px) Image search: [Google]

608KB, 2688x1520px

RIP in piss Intel, this is what you created again.

>>60821548
No, I'm saying it's literally a Xeon that they're putting on X299 and slightly adjusting clocks. They likely can't just put them over 20% higher. 3.1-3.3 is more likely but lets say it's 3.5 all-core boost.
I wouldn't be surprised if Threadripper 16c at 3.5-3.7GHz is significantly faster in multithreaded on average than Intel's 18 core at 3.5Ghz. That would be completely expected.

>>60821535
Yeah. A 32core with higher clocks than 28cores at the same or less power usage and thermals.

File: cpuz_2017-06-09_06-35-12.png (78KB, 503x502px) Image search: [Google]

78KB, 503x502px

>>60814075
>only 98.7% multithread scaling
>not 126%
loling at ur life Intel

>>60821535
It's gonna be worse than bloodbath. Three bingbus abortions are gonna scale like shit.

>>60813837
>im
>collage

>>60821734
CPU-z is fucking junk.
So are pissmark, geekbench and userbenchmark
And sunspider, and V8/Octane, and Kraken and Peacekeeper and a bunch of other crap

Seriously fuck this world, what's the point of these shitty microbenchmarks?

>>60821719
>>60821739
Cannonlake-EX was cancelled, will they stuck with Skylake-EX for like 2/3 years now? Intel was tweeting about Tigerlike tape-in recently, dubious that there is something revolutionary about ringbus scalability

>>60821775
yeah you're right. But it's still funny when an Intel-rigged benchmark shows it's awful.

Only benches based on actual applications that people actually use like Cinebench, Blender, Divinci's Resolve, etc, should be used outside of shitposting.

Nothing wrong with V8 as a bench except that it changes so often.

>>60821821
Yes. They are taping out 2nd gen 10nm soon or already did.
Meanwhile AMD is taping out on IBMs 7nm later this year.

>>60821821
>https://twitter.com/intelnews/status/872939120334479361
*icelake- too many lakes anyway

>>60821846
bets on intel will accelerate work on their rumored "brand-new" microarch, supposed to be for 2021

>>60821846
>IBM 7nm
Oh the shoah.

>>60814927
None of these CPUs will be optimal for vidya. And oh yeah, Intel didn't solder their Skykek X CPUs and use mayonnaise TIM again.

x86 is such a fucking joke architecture. Intel and AMD are like two retarded kids fighting over friends. Why can't we just go full ARM?

>>60822258
ARM is junk when over 15 watts

>>60821866
Their 7nm fab probably won't be ready until 2021 to begin with, though.
MAYBE 2020, while AMD is making significant arch updates every year or two apparently.

>>60822258
ARM is totally shit when you try to make something high-performance out of it.
>>60822295
I think GloFo will adopt 5nm GAAFETs by the time Intel opens their 7nm fab.

File: 1495464195552.jpg (6KB, 118x137px) Image search: [Google]

6KB, 118x137px

>>60813744

srsly

>>60822258
ARM is proprietary as fuck.

>>60822331
>ARM is totally shit when you try to make something high-performance out of it.
There isn't much point to RISC now anyway because of the way high-perf microarchs split CISC instructions to microops and CISC still has the advantage of dense instruction encoding. Still Zen lead engineer said he's pretty confident that they could make a high performance ARM cpu, there seems to be no demand for it though

>>60822418
isn't x86 a combo of CISC and RISC at this point anyway?

>>60822425
It was always that.
>>60822418
They have K12 ARM core, which we know nothing about. I guess they shelved it considering how well Zen scales into lower power envelopes.

>>60822258
Yeah why don't desktop users use that awesome line of ARM chips that totally come anywhere close to even AyyMD let alone Intel? What are they called again? I just can't remember the name

>>60821337
>it hasn't noticed AMD eating its legs
Yeah, of course it wouldn't. When you shit on the competition for 18 years, there really isn't much to worry about. AMD just started winning for a couple of months. No need to give Intel all this flak.

File: amd64.png (16KB, 288x288px) Image search: [Google]

16KB, 288x288px

>>60822803
>18 years
*blocks your path*

>>60822833
The last time AMD was king was in 1999. The last time AMD was somewhat competitive was during the Athlon 64 years.

AMD is once again king, and Intel has nothing to blame for that other than itself. If the company had just kept innovating instead of re-releasing hardware, we would've been on 16c32t CPUs a long time ago. However, I think the reason for no innovation is that it's incredibly difficult to break the 10nm barrier as that's smaller than atoms.

>>60822425
Pretty much everything is hybrid since K5/Pentium pro days. Pentium 4 was even more RISCy than current processors, they dialed down after that. Agnes's CPU blog and manuals has excellent high level info on that

>>60822469
They also have A1100 series, if it serves the current limited market they are probably not bothering to spending millions to tape in K12. When it comes to low power Zen it is said to be clock gated as fuck, I would love to see the Zen based bobcat/jaguar successor

>>60822857
>Shitburst was better than K8
Rope yourself fella.

>>60822862
>I would love to see the Zen based bobcat/jaguar successor
Ask Sony to rush out PS5 to see that.
>>60822857
>incredibly difficult to break the 10nm barrier as that's smaller than atoms.
It's piss easy.
t. IBM Nazi Scientist

>>60822803
Everything from K6-II to K8 was AMD being competitive or outright winning, Phenoms weren't bad at all even against Nehalem it only got really bad with BUlldozer.

>>60822877
>Ask Sony to rush out PS5 to see that.
Speaking of that Su talked about a 3rd semi-custom win. don't think it's revealed yet

>>60822912
>Speaking of that Su talked about a 3rd semi-custom win
Link ples.

>>60822930
https://www.linkedin.com/pulse/who-amds-partner-mysterious-arm-based-semi-custom-chip-luskin
It's said to be about ARM

>>60822965
>ARM
I wonder if it'll use their custom K12 cores or reference ARM ones.

>>60822965
Samsung/Mediatek chip with AMD GPU?
Imagine if it was actually Apple, holy shit.

>>60814075
>1.764v
Kek

>>60822833
Ah my motherboard had integrated Nvidia 9400 with Athlon.

>>60822341
This burns me too.

Fucking Intel, their monopoly and fanboys. You could've been the best but now your managers and marketing got
chokehold of progress.

>>60823001
>Apple is going to drop Imagination graphics (PowerVR) from their ARM designs
Really
makes
you
think.

>>60823001
If AMD got a GPU inside Apple's phone/tablet SoCs their stock would go 50% overnight

>>60823036
Considering motherfucking pajeet worked in Apple before. Eh, maybe. Would be very fun to see AMD's stock go to the moon.

>>60823036
Their stock went up something short of 30% in the last few days from nothing particularly big(that Vega in the iMAC is a low volume product)

50%? More like 150% $AMD is horribly undervalued

>>60823058
>50%? More like 150% $AMD is horribly undervalued
EPYC is yet to hit the market. I wonder who's that mysterious company from 7+1 ordering fuckton of EPYCs.

>>60823079
Some cloud provider, all they need is I/O and memory capacity.
AMD can't be beat on that.

>>60823088
>Some cloud provider
Everyone from 7+1 is a gigantic cloud provier. We need to guess which one. I'd say it's microdicks considering Intel kiked them on SKL-EP Xeons.

>>60823088
Hell, Microsoft already told Intel to get on their bikes in regards to MS's "Project Olympus"

>Apple drops PowerVR graphics in favour of something else
Really makes you think.

>>60823058
>50%? More like 150% $AMD is horribly undervalued
Fucking this, AMD's stock goes up and down like a fucking drunk on a acid trip, this kind of high volume high margin product would skyrocket them literally to the moon.

Can someone explain why the i5 and i7 variants in the i9 lineup have a higher tdp than the 7600k and the 7700k despite not having an IGPU? The clock speeds are only 100 mhz higher

>>60823135
Patrick Kenedy already hinted at some EPYC design wins he can't say until NDA is lifted, that I can guarantee would do better on AMD's stock than anything else.

>>60823147
Power consumption increases quadratically with voltage, not linearly

>>60823191
>until NDA is lifted
That's on 20th of June?

>>60823226
Yes.

>>60823245
Sweet.

>>60823226
Check the comment section
https://www.servethehome.com/amd-epyc-june-20-2017-threadripper-vega-computex-2017/

This guy sells servers and is a consultant for a living and is good friends with Ian from anandtech, his words mean quite a lot.

>>60823280
>We are under embargo/ NDA about specific vendor plans and platforms implementing EPYC. We will have a lot of coverage come release day.
Wowzers. It's gonna be a bloodbath for Intel.

>>60823280
AMD has Ryzen from the dead.
They could safely ignore Nvidia for now, they should rape until all those holes are left bleeding.

>>60823320
>They could safely ignore Nvidia for now
They won't, considering Vega is targeted at core nVidia markets.

>>60823320
>should rape Intel*

File: Untitled_d66894_5603393.jpg (26KB, 504x294px) Image search: [Google]

26KB, 504x294px

>>60822258
Fuck outta here with your angry birds processor.

>>60823329
I know, keyword being _could_

CPU market is far larger than GPU one, Intel's 14+ billion USD per quarter tells you everything, this is purely from CPUs

https://youtu.be/TWFzWRoVNnE

>>60823280
AMD 10-15% of the server market by end of 2018 confirmed, and then Zen2 at 7nm comes in while Intel is still on server Skylake.

Snowball effect? More like a falling anvil.

>>60823347
>Intel's 14+ billion USD per quarter
And Huang still tries to push ebin GPU compute and the datacenters meme. Anyway, Vega is aimed at ML/gayming, we'll see how it competes with current NVIDIA lineup.

>>60823351
>2 million views for $800+ CPU alone platform only 2-5% of the market will have

This is gold, Intel must be shitting bricks.

>>60822594
It's called Cortex A73 and it comes close to past generation i3, thus enough for 99% of normies use and enough to run GNU/Linux

>>60821761
kek!

File: JUST krzanich.png (1MB, 1002x1480px) Image search: [Google]

1MB, 1002x1480px

>>60813744
>AMD Eng Sample: 1D3101A8UGAF3_36/31_N 16/32 - 3100/3600
Press S to spit on their grave

>https://videocardz.com/69760/engineering-sample-update

>>60823438
AMD ZD1840A8UGAF4 16/32 3400 3700 AMD Whitehaven

This looks far worse for Intel

File: elves.png (109KB, 414x426px) Image search: [Google]

109KB, 414x426px

>>60823456
>these clocks for 16 coarz

File: 1.jpg (95KB, 921x519px) Image search: [Google]

95KB, 921x519px

>>60816046
>>60816355
If my suspicion is right, this infinity Fabric could potentially make the PS5 incredibly powerful for a really, darn low cost.

You know how Sony poured billions into the PS3 Cell.
But this AMD technology is literally the wet dream they had envisioned back in the day.

And this is also exactly what consoles need.
A custom motherboard, with a shit ton of small, yet affordable chips that scale perfectly. And a custom cooling block that encase the 10-20 small chips, or whatever.

I have a feeling PS5 is going to kick a lot of ass.

>>60823456
I give my left but that that 3400 will be 3600 by the time TR releases.

>>60823533
So will the next xbawks, since both use AMD.

>>60823533
That depends on whether Navi will be what i'll imagine it to be.

File: vega_navi_banner-1068x580.jpg (95KB, 1068x580px) Image search: [Google]

95KB, 1068x580px

>>60823567
Good find.

It looks like scalability is one of its keywords. That's really interesting if they can engineer a shit ton of small graphics chips that interconnect with each other in an efficient way just like the Ryzen.
It would open up for amazing custom design machines like the consoles.

>underpowered xeons
>raid key
>having to pay for a new chipset, even tho the old one was still usable and could do all the same things
>overpriced
>44 pcie lanes
>16 pcie lanes on the rebranded 1151 cpus
>raid key again kek.

Enjoy your overpriced underperforming housefires there intelfag

>>60823599
>It would open up for amazing custom design machines like the consoles
It would simply kill nVidia for a few years until they create something similar to IF. MCMed GPUs means custom design for anything.

>>60821761
No one said you had to be smart to go to college .

>>60823058
I think Intel seeing AMD's stock skyrocketing that high would cause mass suicide in the company.
Not so much seeing the report that their cpus are getting btfo by AMD's but the realization that they can't tap that shekel faucet.

>>60823617
doesn't nvidia have nvlink? I thought it was similar to IF.

>>60823676
NVlink is a handshake point to point protocol replacing PCIe, it's also ran over mezzanine, it's got nothing to do with IF

>>60823676
It's beefier SLI. It's nothing like IF.

>>60823615
>no ECC

>>60823690
>>60823691
my bad. thanks.

File: AMD-Infinity-Fabric-Data-on-a-Chip.jpg (317KB, 1920x1080px) Image search: [Google]

317KB, 1920x1080px

>>60823617
Honestly, if I was the motherboard makers, I would create custom motherboards with implanted CPUs on them, right now.

Imagine a motherboard, with 8 Ryzen chips, welded onto the board, like how the laptop chips usually are.
Sure, you forego the CPU upgrading, becuase there is no more CPU slots.
But really, you now have a beast of a system with 64 cores on motherboard.

You could make a pretty penny out of that.

>>60823676
Infinity fabric is said to be able to interlink all of AMD's Ryzen cores, with their future graphics cores.

>>60823715
>said
It does exactly that in Raven Ridge, replacing custom Onion/Garlic buses from earlier APUs.

>>60823438
>>60823456
RIPtel

File: monitor-spec.png (24KB, 832x1030px) Image search: [Google]

24KB, 832x1030px

New HDR and freesync 2 premium monitors, Vega looks to be doing quite fine.

>>60823837
>1440p
>144hz
>quantum dot
>VA panel (2000+ contrast??)
>125% sRGP
>31" (!!!!!!!!!!!) fuck 27"

Fuck me this would be the most perfect thing ever if it was 4k
I'm sick of 27" monitors, these specs with 4k at 31" when?

>>60823837
Jesus fuck I just bought a 144hz freesync anus monitor a month ago.

>>60823837
How much is that 27 inch going to cost?

>>60824167
Probably not very cheap, quantum dot , FS2 and 125% sRGP are pretty much premium checkboxes

>>60822857
Atoms are far smaller than 10nm anon.

>Apple/Microsoft switches to Ryzen next year
>Macbook Pro/Surface start coming with Raven Ridge APU's
>Macbook Pro/Surface suddenly stop having thermal and throttling issues
>Intel again makes an ass of themselves
I want that to happen.

I wouldn't mind a non-throttling Surface tablet.

Is my 6700k obsolete now? I can barely get 900 on cinebench at 4.7GHz

>>60825176
No, unless you do heavy threaded tasks.

>>60813744

>AMD:64 Bit is cool.
>Intell: U dont need it.


>AMD: Huh, Look at all this ram, hey you can make use of this
>Intell: FUCK QUICK , 64 bit!....... FUUUUUCK.-
>-Just adopth AMD's 64 bit setup.
>-Goddamnit.

--years later

>AMD: Well, we can make the processor cores small right? -
>-Let's put a few more on-there.
Intell: U dont ned it.

And here we are.

>>60825322
It's not just about cores: AMD achieved fucking 7.90 SPECINT scaling across 8 motherfucking dies.

Will Chinknovo release new Thinkpads with Ryzen?

>>60825335
Well yeah, They have more experience dealing with optimizing multi-core system while Intel was like "4 cores is enough, just hyper-thread the rest"
They have more money to throw at R and D, but they wasted both time and money on something that was mostly a soft solution.
Does hyper-threading really do anything?
I mean realistically Is one hyper-threaded "core" faster than a regular core while keeping the same form factor and power draw?
I don't know enough about this stuff to say but it seems that it would be more of a 1.5 core than a single core.
As things shrink we end up in a place where such a thing is pointless.

>>60823615
>jew jizz TIM

>>60825411
It's not about their shitty SMT, it's about Intel being unable to design good interconnect.

File: speccy.jpg (55KB, 659x477px) Image search: [Google]

55KB, 659x477px

>All this cool shit comes out
>I'm still stuck in 2010

>>60825455
Hi! How are those intel stocks going?

File: intel ceo JUST.jpg (119KB, 1000x1000px) Image search: [Google]

119KB, 1000x1000px

>>60825455
>intelcucks are this delusional

>>60825473
Sandy Bridge still is great, but has his days counted.

>>60825455
>Pentium still Machokes
>i3s still not Machamps with fancy hats
We need a more modern edit of this

>>60825608
Was a shitty meme to begin with that only made sense during the FX days. Now it's just posted and edited by retards.

>>60825473
at least your mobo is not catching fire, poorfagbro

File: temps.jpg (61KB, 687x547px) Image search: [Google]

61KB, 687x547px

>>60826857
fug, forgot pic

>>60826881
>>60826857
>tfw mobo runs at 62C every tiem

File: self intoxicated, twisted detonator.png (314KB, 1131x1006px) Image search: [Google]

314KB, 1131x1006px

>>60826881

>he thinks that is hot

STEP ASIDE, LOOK AT A TRUE MASTER AT WORK. This shit is running at 1c below hardlock temps - the second the motherboard hits 71c it will bring the entire system down.

>>60825335
With that kind of scaling Ryzen has, even companies like Asus and Gigabyte can now go into server/HPC business for cheap.
They have the resources to create their own motherboards with all the integrated components they need.

All they would need to do is order a crap million ton of R7 1700 chips by the bulk. And they're in a whole new business.
I get even normal consumers would be interested in a motherboard with a ton of CPU cores.

>>60827376
well, that were only temps on idle
there are temps logged from yesterday, and I was manually switching on fans into max blast mode when I noticed motherboard going over 80

>>60814927
do we have any confirmation that the 16c ThreadRipper will indeed cost $850?

File: temps2.jpg (140KB, 1163x613px) Image search: [Google]

140KB, 1163x613px

>>60827723
and I forgot pic again

>>60827758
>$850
That's actually a huge bargain. You get get 4 Ryzen chips, each will cost you 212 dollars.

That's even cheaper than the R7 1700.

Nevermind>>60827793
>You get get 4 Ryzen chips

That's Epyc, not Threadripper.

>>60817555
I pay tenfold that for my juice.A fucking magnitude more. Fuck you.

File: Untitled 3.jpg (77KB, 1000x600px) Image search: [Google]

77KB, 1000x600px

>>60814046
>graphics card
>GPU

File: GC42GmA.jpg (223KB, 960x720px) Image search: [Google]

223KB, 960x720px

look at it this way, for every 64 bit chip Intel sells they have to pay amd.

>:^)

>>60824167

700€ for the 27"

File: IMG_5803.png (270KB, 750x1334px) Image search: [Google]

270KB, 750x1334px

>>60824167
>>60824198

And 800 for the 32", release date 28th of june

>>60830753
I was expecting a lot more expensive.

Thankfully I'm waiting until a GPU that can do 120Hz on 4k solidly, so after Vega/Volta most likely.
There'll be plenty of high contrast, high refresh rate and high resolution monitors then.

>>60822418
Eh I don't agree.
AMD has a 4w TDP x86-64 2c/4t CPU on their roadmap for next year.
It may perform very favorably to ARM.

They are planning to drop some old instruction paths, like FP16, from what I heard which will make it a lot more efficient in the more common FP32 even if FP16 will run at half speed.

>>60822341
>>60823021
talk about desperate

File: AMD.jpg (363KB, 1079x1115px) Image search: [Google]

363KB, 1079x1115px

>>60813744
Succcccc

>>60818661
Can anyone explain this? Intel has been selling 18-core xeons for years now, pretty successfully-- what's different?

>>60832870
That was before Zen.

>>60832870
Intel's architecture always scaled very poorly past 10-cores, it's simply that until now there was no alternative available so Intel could get away with the multiple ringbus circus on their high core count skus.

Until AMD came out with an architecture that scales almost linearly to 32c/64t. Now Intel has a huge problem because the further you go above 10-cores the larger and larger Zen's scaling advantage becomes.

>>60833055
>7.90 scaling across 8 fucking dies
Somebody stop AMD.

>>60832883
>>60833055
Thanks, that makes sense. Although, it's yet to be proven that IF actually scales that well, I certainly hope that it does.

Teach me something, /g/
I'm reading that intel has only 1.7% yield on their fabs.
Does it mean that for every 1000 chips they try to make, only 17 work to be selleable?

>>60834439
it's a very old alleged figure for their then very experimental 10nm process for unknown dies of unknown size. their kabylake/skylake yields are probably very reasonable.

>>60834439
Sounds awful low. Recently it came out that AMD Zeppelin is yielding 80%, that's probably closer to a "normal" yield figure.

>>60834439

>1.7% woodscrews memes

Unless you do a large complicated chip like a GPU which requires a ton of die space, which Intel's Core series isn't, your yields will never be that low. The main problem is getting yields to a high volume manufacturing profitability-wise. If I were to estimate now on Intel yields based upon rumors it is probably in the 40-50% range right now which means you are wasting more than half the die potentially and it's no wonder Intel will start shipping mobile first on 10 nm with how the figures are.

>>60821204
>>60816474

Surprisingly, he has a recent video that is him basically just shitting on Intel and the i9. Not that that changes his status as a shill...

>>60825455
It's not 2013 anymore.

>>60825322
Delete this, stupid goy

>>60827793
*2 Ryzen chips

>>60832870
No competition so you were pretty much forced to swallow shit.

>>60833071
CAN'T STOP