Quantcast
[ 3 / biz / cgl / ck / diy / fa / g / ic / jp / lit / sci / tg / vr / vt ] [ index / top / reports / report a bug ] [ 4plebs / archived.moe / rbt ]

Due to resource constraints, /g/ and /tg/ will no longer be archived or available. Other archivers continue to archive these boards.Become a Patron!

/g/ - Technology


View post   

[ Toggle deleted replies ]
File: 9 KB, 821x197, AMD.png [View same] [iqdb] [saucenao] [google] [report]
69368488 No.69368488 [Reply] [Original] [archived.moe] [rbt]

It's over AMD is finished.
>14 core 28 threads
>4.0 GHz base clock
>5.0 GHz boost clock

https://www.anandtech.com/show/13804/intel-core-i9-9990xe-up-to-5-ghz-auction-only

>> No.69368515
File: 2.96 MB, 3800x2613, 1385944010278.png [View same] [iqdb] [saucenao] [google] [report]
69368515

>>69368488
Yes, and the 2 horsepower chiller is sold separately. There's another thread up already, fag. This is now a Suzuha butt thread.

>> No.69368523

test

>> No.69368524

What, you have to buy it in an auction?

>> No.69368539
File: 69 KB, 775x837, 1478201807359.jpg [View same] [iqdb] [saucenao] [google] [report]
69368539

>>69368488
>255W TDP
wtf intel

>> No.69368551

>>69368539
it's afraid

>> No.69368570
File: 359 KB, 1300x599, 1497918643725.png [View same] [iqdb] [saucenao] [google] [report]
69368570

>>69368539
and in reality it will double that

>> No.69368576

>>69368539
>TDP is just a suggestion :^)

>> No.69368578

>>69368524

It is because they are enterprise rejects that ate too much power. Intel decided to overclock them in-house and got lucky. There are probably only a dozen or so units if at that.

It is a pure marketing stunt. Not even x86 CPU collectors would want them.

>> No.69368594

Anyone with a workload that needs 28 threads for the probably $2500 these will auction for (also, wtf really intel? An auction?) would probably be better off with 64 threads for nearly $1000 less for a threadripper.

>> No.69368605
File: 16 KB, 480x318, 5bf.jpg [View same] [iqdb] [saucenao] [google] [report]
69368605

>>69368488
>TDP 255 W
>there is no warranty from Intel
lmao what the fuck

>> No.69368619

>>69368488
Apparently this CPU won't even be available to consumers. It's for manufacturers to bid on.

>> No.69368653

>>69368488
>>69368619
Its one of the CPUs Intel sells to stock market traders. Used to be called blackops. Only ever seen one of them myself.

>> No.69368654

>>69368488
>14 core 28 threads
How? Isn't HT dead?

>> No.69368679

>>69368653
I can't see why the hell this CPU would be attractive to HFT traders.

>> No.69368689

>house fire
>Neighbourhood fire
>City Fire
>Country fire <-- you are here

>> No.69368750

>>69368679
higher clock speeds?

>> No.69368837

>>69368488
>14 cores at 5 niggerhertz when a 9900K already gives water coolers a decent workout
Uh...based I guess.

>> No.69369068
File: 19 KB, 1900x900, higher is better.png [View same] [iqdb] [saucenao] [google] [report]
69369068

>>69368488

>> No.69369383
File: 1.13 MB, 1366x768, mfw we started had the fire.png [View same] [iqdb] [saucenao] [google] [report]
69369383

>>69368539

>> No.69369415

>>69368488
>TWO
>HUNDRED
>FIFTY
>FIVE
>WATTS
>TEE
>DEE
>PEE

>> No.69369438

>>69368488
>Intel is listing this processor as 'off roadmap', meaning it won't come to retail. Instead of selling to end users directly, Intel will offer it to select OEMs via a series of auctions, held once a quarter, with the first auction scheduled for the third week of 2019. This means the price of the processor is not fixed, and will depend on how much each system integrator is willing to pay. We also learned that only three system integrators will be at the first auction.

>You can only buy this CPU if you buy a prebuilt
Nice.

>> No.69369439

>>69369068
*Higher is better

>> No.69369445

>>69369415
vs. FX 9590 Power consumption test WHEN

>> No.69369466

It uses more power than my neetbux can afford

>> No.69369528

Intel is going to auction these at 10k$ and there is nothing seething /g/tards can do about it

>> No.69369568

>>69369528
gotta make up for millions in loss somehow kek

>> No.69369682

>>69369445
Difference is you can air cool a 9590 at 5ghz.

>> No.69369705

>>69368488
amd has awakened the dragon

>> No.69369867

>>69368488
Holy kek Intel is fucking hilarious. They're going to get shit on so badly it's not even funny. Brian, oh why Brian must you fuck shit up wherever you go?

>> No.69369891

>>69368488
>only sold at auction to OEMs
>255w (((tdp)))
sure, Intel will take so much market share with this low volume housefire

>> No.69369949

Absolute desperation.

>> No.69369974

>>69368488
$1000000000

>> No.69369981

>>69369867
You can't blame this on Brian. This is Swan's fault. Intel's running headless and this is one of the results.

>> No.69369999

Its over x86 is finished
SPARC T8 32core/256threads
5Ghz base

>> No.69370005

>>69369705
are you calling it the dragon because it's going to burn down my house

>> No.69370009

>>69368539
Big boy CPU has big boy power requirements, 1k PSUs are back baby.

>> No.69370022

>>69368539

oy vey

>> No.69370034
File: 101 KB, 800x1012, 1466983305_intel-vs-amd_o_873261.jpg [View same] [iqdb] [saucenao] [google] [report]
69370034

>>69368488
Things sure are different now.

>> No.69370045

>>69369999 (Checked)
And kek'd.

>> No.69370048

>>69368539
No need for a heater now based.

>> No.69370056

>>69368679
HFT niggers need the lowest possible latencies, ie high clocks
They barely give any fucks about TDP
Still, this should use the trashy mesh which should be quite slow for HFT

>> No.69370060
File: 956 KB, 500x197, hooking-up-the-intel-9990XE-for-the-first-time.gif [View same] [iqdb] [saucenao] [google] [report]
69370060

>>69368488
>255W

>> No.69371595
File: 24 KB, 1712x284, 1547405897008.png [View same] [iqdb] [saucenao] [google] [report]
69371595

>>69368515
/thread

>> No.69371607

>>69368689
INTEL 28 CORE 5GHZ GLOBALFIRES

>> No.69371797
File: 1.86 MB, 5160x2696, Just wait for 10nm.jpg [View same] [iqdb] [saucenao] [google] [report]
69371797

>>69368488
Leaked photo of Intel testing overclocked CPU in closed water loop

>> No.69371844
File: 24 KB, 384x332, 1519227280089.jpg [View same] [iqdb] [saucenao] [google] [report]
69371844

>they likely can only build a few hundred a quarter, hence the one auction per quarter.

>> No.69371861

>>69370060
it's "255W" TDP.

From the article:
>"...which means the peak power will be higher. Motherboard vendors will have to support 420 amps on the power delivery for the chip (which at 1.3 volts would be 546 watts), and up to 30 amps per core."

And the best part:
>no warranty

>> No.69371907

>>69368488
Intel is afraid...

>> No.69372281
File: 107 KB, 631x438, skylake_guy.png [View same] [iqdb] [saucenao] [google] [report]
69372281

>>69368488
>competition doesn't matter until it does but then it's not fair

>> No.69372293

>>69372281
You don't have to beat the competition if you burn the whole planet with 1770W globalfires :)

>> No.69372389

>>69368488
>initially sold at closed auction once per quarter to three system builders with no warranty available
I can't even tell if this is desperation anymore

>> No.69372447

Intel you are drunk. Go home.
When you sober up, you need to chop heads in your marketing department. They are still living like it is late 2000s-2010s where they never dealt with genuine competition.
AMD won the short-term got over it. Go back and address your foundry issues and move beyond a highly evolved Core-2 design.

>> No.69372470

>>69371861
That's literally insane

>> No.69372482
File: 2.74 MB, 640x480, 1533692512381.webm [View same] [iqdb] [saucenao] [google] [report]
69372482

>>69368488
>255W TDP

This guy is using an engineering sample.

>> No.69372488

>>69368539
>255W heater that runs on 400W power

>> No.69372498
File: 112 KB, 255x231, 1430236296498.gif [View same] [iqdb] [saucenao] [google] [report]
69372498

>>69372488
>Motherboard vendors will have to support 420 amps on the power delivery for the chip (which at 1.3 volts would be 546 watts), and up to 30 amps per core.

>> No.69372524

>>69368488
>255W

are we going to see motherboards with beefy cooling for the power vrm?

>> No.69372530
File: 1.34 MB, 3072x2048, 1251640437237.jpg [View same] [iqdb] [saucenao] [google] [report]
69372530

>>69372498
>Buildzoid here
>YouTube explodes

>> No.69372543

>>69368689
Cue in the obligatory "7980XE release in US" meme picture.

>> No.69372548

>>69368570
tdp != power draw

tdp is a measure of heat output by the processor and the amount of cooling required to dissipate heat generated by the cpu.

>> No.69372557

>>69372498
>420 amps
Jesus

>> No.69372577

>>69372498
>420 amps on the power delivery for the chip (which at 1.3 volts would be 546 watts)
FUCK

>> No.69372787

>>69368679
Nigga HFT has been all AI for years, they need all the parallel processing they can get

>> No.69372789

>>69372498
>420 blaze it

>> No.69372797

>>69368488
when you have to use liquid nitrogen just to get 5 more FPS in Fortnite, shit gets retarded even for Intel fans

>> No.69372804

Given this is only gonna be offered in prebuilts, how badly do you think the OEM's are gonna fuck the power/cooling requirements for it?
Think they'll need to roll their own mobos, just for this chip?

>> No.69372850
File: 356 KB, 820x541, 1486177056485.png [View same] [iqdb] [saucenao] [google] [report]
69372850

something very special is happening at intel hq i'm sure, like potato tier special

>> No.69372865

>>69372804
they'll partner up with the usual mobo oems to make their own odm-branded mobos. It'll be interesting to see how they'll cool this beast of a processor.

>> No.69372882
File: 926 KB, 1500x1009, qwe_download.jpg [View same] [iqdb] [saucenao] [google] [report]
69372882

>>69369383

>> No.69372888

>>69368488
>255 W
lol

>> No.69372937

It's just a binned 9940X. With that TDP liquid cooling is already a necessity so you may as well get a 9980XE and OC it yourself.

>> No.69373030
File: 1.09 MB, 498x272, 1541156923068.gif [View same] [iqdb] [saucenao] [google] [report]
69373030

>>69368515

>> No.69373063

>>69368488
Holy kek, it's the FX-9590 all over again except MORE.

>> No.69373151

>255w for the 14x4.0Ghz clocks
>666w for 14x5.0Ghz

>> No.69373237

>>69369439
underrated bost

>> No.69373296

>>69368488
They're only doing this to keep on the top of the benchmark charts right? (as a PR move)

>> No.69373395

>>69373296
Except they won't (zen2 > this bs).

>> No.69373488

>>69371861
>420 fucking amps
you know where else they use this much power?
welding
it's possible the chips will weld themselves to the board

>> No.69373535
File: 17 KB, 346x242, DiabolicalLaughter.jpg [View same] [iqdb] [saucenao] [google] [report]
69373535

>>69372789
that meme has never been more relevant

>> No.69373566

>>69370034
Absolutely based

>> No.69373575

>>69373488
That's at 1.3V, so it's "just" ~550W. Welding uses much higher voltages and therefore much more power.

>> No.69373753

>>69368488
Luckly it is not a real product, but something they will sell for select few people.

>> No.69374171

>>69369981
Intel put Raja in charge of everything.

>> No.69374264

>>69373575
it's mostly amps that heat shit up
thats why transit lines use high voltage/low amps
I wouldnt be suprised if those chips caused some small spotwelds on connectors

>> No.69374265

>>69373575
True, but doesn't really matter.
When it comes to heating, current is the only thing that matters. That's why you can have millions of watts going through power lines at high voltage and low current, then exchange voltage for current with a transformer and melt metal with much lower voltage and higher current

>> No.69374390

>>69374264
its also becasue amps get lost in long wires

>> No.69374409

>>69374390
You sure mean volts, or maybe watts.
The current will be the same on both ends.

>> No.69374439

>>69374409
ohms literally lower amps onii-chan

>> No.69374468

>>69374390
>>69374439
an where do you thing that lost energy is going?
heating is direct result of resistance

>> No.69374495

>>69374468
i wasnt trying to prove you wrong about amps heating

>> No.69374520

>>69373030
kys

>> No.69374542

>>69368488
>auction only

So they couldn't even push out a decent batch of these to compete with Zen 2?

>> No.69374615

> 2950X
> 16 core / 32 threads
> 180W
> 64 PCIe lanes
> up to 4.4GHz
> $880

vs

> i9 9940X
> 14 core / 28 threads
> 165W
> 44 PCIe lanes
> up to 4.5GHz
> $1380

Is Intel even trying?

>> No.69374657

>>69374615
Did someone measure the Rome package yet? If so, does it fit in the TR4 socket? Things could get absolutely crazy in the HEDT space.

>> No.69374690

>>69368539
Note: Intel's TDP calculates only base clock. With AVX and full turbo, its 400+ watts for sure.

>> No.69374757

>>69374657
looks too big

>> No.69375029

>>69374615
>i9 9940X
>165W

>> No.69375042

>>69368488
>limited bin chips for $2500

Gotta jew fast

>> No.69375420

>>69372498
>100 W heat on the pcb traces alone
good thing those are made of FR4 these days

>> No.69375686
File: 1.06 MB, 1846x1923, 1547145956529.jpg [View same] [iqdb] [saucenao] [google] [report]
69375686

>there is no warranty from Intel. This means that system builders will not be able to recoup costs on dead silicon

>> No.69375759
File: 153 KB, 942x569, Ringbus of fire.png [View same] [iqdb] [saucenao] [google] [report]
69375759

This needs to be updated it seems.

>> No.69375813

>>69372548
gee i wonder how much the heat output of a processor corresponds to the power draw of a DC circuit and is wasted on heat over a second. If only there was a way to know this...

>> No.69376031

>>69368488
Pretty impressive e-waste disposal grift, Intel.

>> No.69376040

>>69373488
The quantity you quoted is not in the dimension of power.

>> No.69376095

>>69368488
Does it come with asbestos gloves?

>> No.69376430

>>69368539
They forgot to add the extra zeros. It's actually 25500W.

>> No.69376556
File: 510 KB, 1315x794, 1501239580918.jpg [View same] [iqdb] [saucenao] [google] [report]
69376556

>> No.69376746

>>69373566
>Intel: MAKE MORE TDP

>> No.69376801

>>69374657
EPYC doesn't work for TR4. We might see a 64 core Threadripper at some point much like the 32-core 2990WX, though.

>> No.69376852

>>69376801
I was asking since previous two TRs were the four dies. Maybe they're going 32 cores on 4 chiplets this time around, but this time with less hassle due to the I/O die.

>> No.69377015
File: 156 KB, 800x1012, moar clock.jpg [View same] [iqdb] [saucenao] [google] [report]
69377015

>>69376746

>> No.69377117

>>69377015
kek, saved

>> No.69377151

>>69376801
>>69376852
I wager that unless Intel can give a legitimate answer TR will cap out at something silly like 48 cores (ergo 2 dead cores per chiplet) to keep epyc that little bit ahead precisely because the uniqueness of the 2990WX won't be repeated. If Intel pulls something out suddenly a full fat 64 core TR will appear (still only quad channel).

>> No.69377336

I just came to ask, why wont AMD completely BTFO intel by releasing an gaymer cpu that has first 2 cores clocked much higher than the rest by default and then just add some slower cores for background noise and multicore perf? Seems to me that scaling is shit in gaymen and 2 cores are the most important. No bully pls.

>> No.69377416

>>69377336
This isn't actually a bad idea and it's really what big.LITTLE aims to do in ARM CPUs. Intel is jumping on that bandwagon with Sunny Cope and I'd imagine AMD won't be too far behind with a nice Infinity Fabric implementation.

>> No.69377457

>>69377416
>Sunny Cope
If that was intentional, bravo.

>> No.69377531

>>69377336
>>69377416
Such a setup prevents scalability given the (relatively) large TDP's desktop chips work at given you end up being constrained by the limitations of the slower cores. big.LITTLE is fantastic if your overall goal is low power draw with enough performance to handle the odd spike of utilisation.

>> No.69377572

>>69373575
>>69374264
Brainlet here, whats the advantage of running chips at lower voltages like this instead of 2.5v or 3.3v? It can't be heat if the chip needs that much amperage to make it run.

>> No.69377644

>>69377336
GoFlo's 12/14nm process was developed primarily for mobile processors, AMD was lucky as it is to get 4-4.4GHz on that process. Cutting back 6 cores from the cluster wouldn't help getting the clockspeed up past that barrier, unless they switched to a different process. I guess that's why TSMC's 7nm is going to be interesting, if they can get clocks up past 4.5 then they should be sitting quite comfortably against Intel's offerings.

>> No.69377646

>>69377572
Transistors can't handle all that load.

>> No.69377661

>>69368488
No guarantee and I need nuclear powerplant for that.

>> No.69377663

>>69377646
Is it because of the size of transistors these days?

>> No.69377678

>auction
So they only managed to make one? Lmao.

>> No.69377681

>>69377663
Yep.

>> No.69377973

>>69377572
double the volts, double the current
now its not that easy irl on silicon but should give you an idea

>> No.69378045

>>69377531
>>69377644
i was not meaning that they have to put extremely shitty cores after the 2, just generic server tier things 2-2,6GHz that are slightly slower but still scale.

>> No.69378075

>>69370034
>>69377015
Tbh, they are catering to braindead gaymers "muh clockspeed".

>> No.69378189
File: 34 KB, 550x512, 1539935863660.jpg [View same] [iqdb] [saucenao] [google] [report]
69378189

>>69368539

>> No.69378314

>>69377572
higher voltages are harder to contain and handle
if the voltage is too high it tends to bleed trough isolative components/layers which obviously causes big problems

>> No.69378536

>intel release a higher core count part than their current 8c "95" monstrosity
>still lower core/thread count than AMD's threadripper
>more expensive than threadripper
>draws way more power than threadripper
>has zero warranty because its such a tremendous housefire

Yes, AMD is truly finished this time. Pack it up, boys. Its all over.

>> No.69379577

>>69376746
>We can't compete with higher cores, start marketing that we have higher tdp

>> No.69379613

>>69378536
I think you need to buy another sarcasm meter.

>> No.69379633

>>69378536
Is capture gopnik audience comrade
Is very cold in Russia, heating bill save by invest in housefire.

>> No.69380279

>>69379633
welcome to megehertzegovina, comrade!

>> No.69380337

>>69378536
Still going to perform better than anything AMD has to offer, specially in Adobe products and general shit that relies on AVX. AMD has nothing but Cinebench scores and better prices, if you really want peak performance, then you'll have to pay the Intel tax.

>> No.69381507

>>69372937
>get a 9980XE and OC it yourself
That's gonna get a bit harder now that Intel is pulling any 9980 that OCs well so they can auction it off (lel) as a 9990 instead

>> No.69382342

>>69368488
Didn't we already laugh this chip into the ground? Didn't the Chinese already see it sprout from the earth and start making knockoffs?

>> No.69382364

>>69381507
The auctioned chips are 9940x (housefire edition)

>> No.69382437

>>69370034

So you are using a single core CPU then?

WTF are you paying for if not more cores? Cores are the fucking product. It's like buying bananas and asking for less banana for the same price

>> No.69382489

>>69380337
Zen2 does 16 DP FLOPs per cycle now, up from the 8 of Zen1.
Only a few SKUs from intel with AVX512 support can do 32. All the mainstream Skylake/Kaby derivatives do 16. That means in peak theoretical throughput all the mainstream parts have parity. The Zen core is wide enough and has all the front end necessary to actually facilitate that execution too.

128 and 256bit AVX throughput isn't an issue for AMD any more.

>> No.69382499
File: 234 KB, 882x758, 1501676989383.png [View same] [iqdb] [saucenao] [google] [report]
69382499

>>69368488

>> No.69382518

>>69382437
This
I got memed by buying into the fast dual core over the slower quad core 10 years ago. As a result I had to do a fresh build earlier than I would have needed which as a result has me in a worse position today.
Always go more cores.

>> No.69382578

>>69382364
Oh I just assumed the 9980 was also a high binned 9940, with the 9990 being the same thing but even higher. If that's not the case then never mind

>> No.69382686
File: 62 KB, 491x550, 2qv4x2.jpg [View same] [iqdb] [saucenao] [google] [report]
69382686

amd stops sucking when?

>> No.69382852

>255w
>won't reach past 4.6Ghz on all cores.

>> No.69382893

This proves Intel have nothing to counter AMD's Zen 2 line.

>> No.69383041
File: 81 KB, 580x326, 0fcf82eaa66e123cfdcfc9a2c506d1222009380019.jpg [View same] [iqdb] [saucenao] [google] [report]
69383041

>>69369999
Based sparc poster

Checked

>> No.69383081

>>69374657
>>69374757
tr1, tr2, naples, and rome all share the SP3 socket. you can use the same coolers.

tr and epyc use different mobos obviously but more importantly run different agesa.

tr3 is still a secret but its gonna be sp3 too

>> No.69383091

>>69382686
2 years ago?

>> No.69383127

>>69383091
>continually makes subpar products outclassed by all competition
>not sucking

yah okay

>> No.69383145

>>69373151
Lol holy shit

>> No.69383501
File: 20 KB, 480x360, hqdefault.jpg [View same] [iqdb] [saucenao] [google] [report]
69383501

>>69368488
DID YOU KNOW THE CIA DOESN'T WANT THE PUBLIC TO KNOW THERE'S A XEON? DID YOU KNOW THAT THERE'S A 24 CORE PC? NO YOU DIDN'T KNOW THAT BECAUSE THE CIA DOESN'T WANT YOU TO KNOW. INTEL DOESN'T WANT TO BE SUCCESSFUL.

>> No.69385180

>>69378045
For what purpose?

>> No.69386009

>>69383127
Gotta shill harder than that my Intel marketeer.

>> No.69386333

>>69373151
Move over boys I'm making a new grill!

>> No.69386796

>>69370005
(You)

>> No.69387570

AMD CTO Mark Papermaster promises more Radeon products this year

https://videocardz.com/newz/amd-cto-mark-papermaster-promises-more-radeon-products-this-year

>>
Name (leave empty)
Comment (leave empty)
Name
E-mail
Subject
Comment
Password [?]Password used for file deletion.
Captcha
Action