HEDT Wars - These chips are HOT

File: nigger.gif (613KB, 498x498px) Image search: [Google]

613KB, 498x498px

>>60402623
Shintel's server misfits
or AyyMD's server misfits

Party like it's ryzen 1999

>>60402702
Do you really want to pay for an Opteron/Xeon + board, though?

>>60402623
>Intel 4-core: 112W
>AMD 16-core: 155W

>>60402729
Hell fuckin no, shit's expensive

2011-1 looks better every new Shintel release

>>60402623
More cores. Less Watts. Same no. of lanes. Probably cheaper.

AMD wins.

>>60402796
AMD measures TDP differently. Add 10-15% when comparing to Intel.

File: 1494767421045_0.jpg (43KB, 630x630px) Image search: [Google]

43KB, 630x630px

>>60402814

Looks like the AM4 platform is already deprecated.

>>60402932
>HEDT platform deprecates consumer one
Wut.

.001% of /g/ could even afford any of this. Better argue over it anyways

It's Holocaust™ 2 for Intel

>core iPanic

Lmao

>>60402623
i'll consider both but leaning amd (especially if they do the same thing they did with the r7 line and massively undercut the blue jew on price)

was eyeing an 1800x before the hedt rumours surfaced

>>60403297
Due to how Intel produces these chips, there's less room for failure than there is with AMD.
Intel requires a perfect die, and even then, they need it to be able to hit those clocks. So that means shit isn't cheap due to the level of perfection required, and to account for the loss of those less than perfect chips.
Meanwhile AMD just glues a bunch of CCXs together, much simpler to produce, much more room for failure, just laser away any failed cores.

>>60402623
> 1998
> 2017

>>60403351
They glue a bunch of Zeppelins. One Zeppelin contains 2 CCX.

>>60403356
1998 was when they started designing K8 and AMD64.
They seem to be pretty confident.

>>60402702
>AyyMD's server misfits
Wat? HEDT is 2 ryzen modules slapped together, while server ones are 4 modules. I don't really think that they're somehow separating best binned chips between HEDT or server.

>>60403444
They'll probably have 16c Opterons.

>>60402814
If that's true, how can you explain overheating of "91W" stock 7700k, while 95W stock ryzen 7 can work with stock cooler?

>>60403463
Then they are going to be 4 1500 together instead of 2 1700.

>>60403483
while I agree tdps between amd and intel are comparable, temperature doesn't only depend on power consumption, and the reasons the 7700k is hotter are that it is a smaller die with higher power/area and that it uses a shitty thermal interface between die and heatspreader.

>>60403525
No, I mean 7700k consumes waaaaay more than its rated TDP.

>>60403297
I'm also considering both, but leaning heavily towards AMD after seeing that the 6 and 8-core i9 chips are only going to have 28 PCIe lanes. I'd rather not spend $1,000+ for their 10-core chip just to have 44 PCIe lanes.

I guess my main worry with AMD is single-threaded performance and potential bugs. Also not sure what to expect in terms of motherboards since I've never paid much attention to AMD.

1 Intel core is as good as 2 AMD cores, Intel has nothing to worry as usual

>>60403839
Did you just time travel here from 2011 to say that, or are you genuinely retarded?

>>60403839
It might be true for uncommon AVX 256 workloads.

File: 1494107396001.jpg (1MB, 1505x4363px) Image search: [Google]

1MB, 1505x4363px

>>60403839
That explains why Intel's 8 cores lose out in 80% situations while using more power.

>>60405263
Name one real application you use that uses AVX-512.

>>60405383
>a feature so useless that even intelshills don't know what it's for

AVX-512 is useless garbage that has no use whatsoever in the real world.

File: muhsinglecore.png (454KB, 1920x1200px) Image search: [Google]

454KB, 1920x1200px

>>60405231
>can you change that to vs 5960X and 6900k, 6850k and so on ?
CPUs that cost twice as much are pretty much out of the running already unless you're a retard that likes pissing away money for no reason, but here you go.

Note this also shows how embarrasingly small the single core advantage Intel currently has over AMD is. That teeny little bump on i7-7700K's single core score is all they have left.

>>60405263
>>60405383
>>60405437
This is beyond desperation at this point.

>>60405437
All I'm hearing is that you can't name a single application that uses it.

>>60405231
There's a 6900k there.

>>60403483
jewtel bundles 65w capable coolers, while amd gives proper 95w coolers to every cpu.

>>60405500
You're the one fapping to an instruction set you have no use for on a CPU you don't even own.

those Ryzen 9 names have to be a joke

>>60405524
Ryzen spire is 125W rated

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

4KB, 452x523px

16c/32t 155W. DELET PLS

>>60402623
>AMD don't cuck thei customers like intel with those sweet PCIe lanes

i like it

>>60405383
So it's useless trash, just like Intel chips.

>>60405477
The i7-6900K is right there are you blind

File: Bomb has been delid.jpg (43KB, 679x367px) Image search: [Google]

43KB, 679x367px

>>60402734
>filename

>>60405621
TDP ≠ POWER CONSUMPTION

>>60405533
>not buying a 'Suez Crisis' Ryzen

>>60402623
POOTEL STUTTERING HOUSEFIRES

intels making house fires now lol

>>60405643
No, power consumption is often higher than the TDP, which is merely its rating for cooling.
Rating, not a measurement.

>>60405663
>TDP ≠ POWER CONSUMPTION
≠? Are you stupid?

>>60403444
>wat is Snowy Owl
Naples isn't everything

>>60405676
I quoted the wrong post.

>>60405621
They might have the entire power saving server features enabled, unlike the R7's and lower
Also, apparently they will be a B2 revision
Naples 32c/64t 8 channel memory tops out at 180W TDP

File: AMDRyzen1800XIntel6950X1.png (29KB, 617x837px) Image search: [Google]

29KB, 617x837px

>>60405231
The 6900k is in that chart, shekelchaser. Oh and here is a comparison of the 6950X to the 1800x.

The real hilarious part is that the 16c/32t AMD HEDT CPU is ALSO going to be cheaper than the 6950X while utterly pushing its shit in.

The second shoah has begun, kiketel shekelchaser. Time to get in the oven.

>>60405858
16/32 is an MCM, realistically it shouldn't go over $1100 for the top chip.
It's not a monolithic chip that has shit yields, it's literally two 1800X glued together with a fancy interconnect

File: ss.png (230KB, 1449x852px) Image search: [Google]

230KB, 1449x852px

>>60405858
The Intel i7-6950X is $1,700, you absolute mong. AMD can sell a 16c/32t HEDT CPU for $1,100 and it will eat the 6950x alive.

Oh, and the 6950x is still faster than the new Skylake-X 10-core (Pic related).

TIME TO GET IN THE OVEN.

File: 1430236296498.gif (112KB, 255x231px) Image search: [Google]

112KB, 255x231px

netburst 2.0

>>60405947
Userbenchmark is garbage and that Skylake-X doesn't have turbo enabled

>>60405903
>>60405947
I don't expect AMD to price the 16c part that aggressively since I don't think there's any way Intel can come close to touching that price point with anything relevant. I'm not sure what price AMD will choose, but it's likely to be based on Intel's pricing rather than what can realistically be achieved with the Zen CCX and Infinity Fabric.

I do agree that AMD could go as low as $1100, or possibly even $1000, though, and remain profitable. But if they can charge $1400, they probably will.

>>60405728
Stay mad with your inferior housefires.

>>60406022
>16c32t eats 8c alice and brags about it
>I won't buy AMD products
Well, you're clearly not buying a 6950X either, since you don't even know what it actually is.

But for the record, that comparison was made because the 16c32t AMD chip is going to be cheaper than a 6950X. Screenshot this, if you want.

>>60405779
Holy shit Intel is a fucking joke.

>>60405947
AMD 16c/32t would beat even two 6950X considering the fact that the 1800X is suprerior to the 6950X

>>60405997
>>60406022
Cry more, kike shekelchaser faggot.

Intel has no answer whatsoever for AMD's 16c/32t HEDT cpu. It's going to be cheaper than the 7920X, It's going to have more cores and threads, it's going to have more L3 cache, it's going to have more PCIe lanes, and it's going to be MUCH faster than the 7920X.

Skylake-X is D.O.A.

>intel fans right now

https://www.youtube.com/watch?v=OEmO-v2R4G0

>>60402623
>Ryzen
>14 cores
literally impossible
CCX units must have same number of cores disabled
you can also disable a whole CCX
you cannot, however, get 14 cores.
not even sure if 12 cores is possible since it's multi die.
I don't even believe AMD is going to release a HEDT platform, 1P server platform is the best you'll get and it will not have the 'ryzen' name
fake news
let me guess, you got that from wccftech, right?

>>60402814
It's Intel that under rates their TDP.

>>60406084
>the lid
DELID DELID DELID

>>60406085
>I don't even believe AMD is going to release a HEDT platform
What? It's X390/X399 and it's been known to be coming for quite awhile, as well as their Naples server platform.

>>60406085
8 core R7 -> 6 core R5
16 core -> 14 core

>>60406141
Shill switches to desperation insult mode! It's not very effective.

>>60406141
the only thing on my that smells like the third world is my dick from face fucking your nigger mother

>>60406160
IT'S FUCKING MAGIC

>>60406141
Please, do explain which Intel HEDT CPU has the remotest chance of being faster than a 16c/32t Zen HEDT CPU?

Because the Skylake-X 12-core isn't going to be it.

File: 1446810325458.jpg (161KB, 1462x1462px) Image search: [Google]

161KB, 1462x1462px

>>60406141
>being this mad over intel's inferior housefires
Holy shit this is hilarious

>>60406158
>it's been known to be coming for quite awhile
wccf-tier rumors.
AMD has never even hinted at a HEDT platform.
16 core and 12 core benchmark leaks are snowy owl 1P server platform

>>60406160
>8 core R7 -> 6 core R5
>16 core -> 14 core
did you see
>CCX units must have same number of cores disabled
8 core = 2 CCX
6 core = 2 CCX with 1 core disabled each
16 core = 4 CCX
14 core = impossible
12 core = 4 CCX with 1 core disabled each
10 core = maybe possible with you can disable a whole CCX in a multi die configuration

>>60406190
THREE
BING
BUSES

>>60406133
yeah
the hand on fire at the end is also a nice touch
I'm surprised noone has tried to do a meme edit with it

>>60406207
>10 core = maybe possible with you can disable a whole CCX in a multi die configuration
no wait I'm dumb, it's also impossible.
even with 1 CCX disabled and other three had 1 core disabled it would be a 9 core.
10 core is also impossible.

Welp I guess another thread derailed by fucktards spamming pajeet and shekel memes. Guess this board is a useless shithole.

>>60406212
no, i just had to demonstrate to you proper shit posting. just calling people indian is lazy and boring. if you're going to troll, at least put in some effort, ya fuckin useless slacker

>>60406237
It's literally just one anally devastated pootel retard.

>>60406239
Ah, so you are conceding defeat and retreating to pure shitposting.

Thanks for playing, shekelchaser :^)

>>60402623
>4c/4t
>112W TDP

This piece of shit of a company needs to be burnt to the ground.

File: AYY.jpg (273KB, 1032x774px) Image search: [Google]

273KB, 1032x774px

>>60406281
Looks like they are working on it themselves.

>>60406217
6 cores enabled on one die, 4 cores enabled on the other. There's your 10c processor. 14C would be 1 fully enabled die, 1 6c die.

We don't know how the Dies handle having cores disabled on different dies but in the same package yet, but it could be possible.

File: megatasking_intensifies.jpg (77KB, 493x493px) Image search: [Google]

77KB, 493x493px

>>60406347
You're like a fucking bot, you get BTFO trying to shill Intel housefires then resort to calling people shreetshitters, pajeet etc. You really should just consider ending your own life.

>>60406370
666W MEGA-TASKING

>>60406298
true, they haven't specified how multi die configuration work, but I just don't imagine them creating such difficult configurations for a niche market.

>>60406347
You are seriously just like a mentally damaged little kid.

>>60402623
>the i9 7920X has AVX512
This shit is going to be the new Sandy Bridge.

>>60406549
what would you use AVX512 for?

>>60406549
>most software doesn't even use AVX256
>This shit is going to be the new Sandy Bridge.
not sure what you'd even use it for in the first place but okay

AVX512 has been limited to Xeon Phi's, literally no consumer chip out supports it, even Broadwell Xeons.

There is nothing using it.

>>60406674
[Citation Needed]

>>60406674
AMD will lower the prices of their old CPUs.
Expect the R7 1700 to be less than $200 at the end of the year.

An engineering sample surfaces, people lose their shit and call it crap because it's apparently slower than the 6950X.

It's an engineering sample. If you would have spent more attention with the past examples you would have seen the fact that on all generations, Intel has first come up with engineering samples running much lower clock. Final product vs engineering samples at the same clock with the engineering samples (ranging from 0.8 GHz to 3.2GHz) performing around 3-5% better than the engineering sample and ultimately managing to hit much higher clocks (over 1-1.5Ghz over the engineering samples while running cooler/much cooler and much more stable).

Can we stop pulling the baby card cryfest over how "shitty" Intel is right now? Remember that the RX 1800X engineering samples were as much as 10-15% more potent than the final 1800X is. No 1800X on the market is as potent as those engineering samples.

Basically wait, the new core i9 CPUs aren't even for your average Joe. You Should actually care for those CPUs only if you are going to use all cores. Unless you are going to use all cores, stop right now and look at the upcoming 8870K (or how the fuck the top quad core i7 will be called).

>>60406674
>On every website I see Intel outselling AMD by atleast 10:1
Based on what, exactly?

>>60406719
I highly doubt it.

If 16c32t is $500 and the 1800X is sold at the same price as the 6700K, then AMD will actually be competitive for once.

>>60406719
[Citation Needed]

>>60406750
Doesn't matter. It doesn't stand a chance against a Zen based 16c/32t HEDT processor. Skylake-X is DOA.

>>60406846
sorry, but isn't the 1800x running at roughly the same performance as the 6950k for half the price already

What is the house fire meme? Can /g/ consistently name a max TDP they would use? Why not go any higher? Why some arbitrary limit?

>>60406085
>CCX units must have same number of cores disabled
Just because that's true for the current SKUs, doesn't mean it has to be true for every SKU.

>>60406750
What's the point if 1800X beats 6950X? This shit is DOA.

>>60402623
>Source: My ass

>>60406893
Yeah this retarded shit is literally Netburst 2.0 fiasco

>>60407056
Doesn't change the fact that it's a shitty DOA housefire abortion.

>>60406946
Anyone?

>>60407125
So basically you're ignoring reality >>60405779

Stay mad.

>>60407056
>made AMD jump twice in core count within months
They had this shit planned out years ago. Zen is designed to scale. Making a 16 core CPU is easy for them. Intel's monolithic design scales like shit (lol ringbuses) and has terrible yields.

>>60405535
wraith - 125w
spire - 95w
stealth - 65w
most common is 95w

>>60407307
What SKUs go with what? Wraith goes with 1700 and stealth goes with 1400.

Where is the spire and that huge Wraith Max one?

File: 1489122339460.png (495KB, 1070x601px) Image search: [Google]

495KB, 1070x601px

>>60407297
>Intel's monolithic design scales like shit (lol ringbuses) and has terrible yields.
FUCK ZHEEEEEEEEEGHHHHHN

>>60407344
Make more bingbus? Yuo of likes more bingbus?

>>60407332
Keep telling yourself that.

>>60406963
>Just because that's true for the current SKUs, doesn't mean it has to be true for every SKU.
except snowy owl is literally just 2 ZP dies in MCM. It's not new silicon.

>>60402623
>AMD's completely bumblefucked cryptographic naming scheme from their unpaid 3-man marketing team

The AMD part is almost surely fake.

>>60407397
>Source: My ass

>>60407412
Do you think we're gonna get 4 die MCM with 16 cores? Dat 8 channel memory and cache would be nice to have.

>>60402814
BS. Lots of Intels, mostly the low-clocked ones, take less power than their TDP.

AMD usually takes preciselly their TDP under all-core load. But they don't cross that line, unless you tell them to by disabling the limits in the BIOS. But that's your call, the CPU itself checks its power draw and adjusts clocks to fit in it, by default.

And yeah, I have that tested by 12V line measurements.

>>60407412
I think he means that there is no technical reason that they would have to use full CCXs or have an equal number of cores disabled on each CCX, i.e. there is no technical reason AMD couldn't have released a nine core version of Ryzen with one core disabled on one CCX only.

File: xeon_e5_v4_hcc_rings.jpg (183KB, 1200x706px) Image search: [Google]

183KB, 1200x706px

>>60407357
I want to get off Mr.Intel's ringbus ride.

>>60407503
You dont like bingbus? Bingbus is future.

>>60407412
So that means they could use one 6C die and one 8C die for 14 cores, then.

>>60407885
we have no idea how CCX communication works with multiple dies
if it's the same as ryzen then all CCX units must be symmetrical

>>60402623
Some programs run significantly faster on one of the systems than another

4ghz 1800X with 4 cores and 8MB ram turned off performs slightly better in cinebench MT than the 7700K at 4ghz
The 7700K does slightly better in winrar MT
16-15-15-35 1T vs 16-16-16-35 2T

AVX-512 might find its way into x265 since its already got AVX256

>>60408162
>Source: My ass

>>.60407503
Stack 12 cores onto the other 12 cores with EMIB?

>>60402734
>TDP scales linear with cores

hurr

>>60406217
>impossible.
You keep throwing that word around, but I for one haven't seen any basis for it save for the current R5 CPUs not using that kind of configuration. It's entirely possible that AMD can do that and just haven't yet because reasons.

>>60409408
IIRC arent there boards that allow for core disabling in an unbalanced manner (ex 3c +2c)?

>>60407339
>Where is the spire and that huge Wraith Max one?
They're all called wraith. Spire is the 95w that comes with 1600 and 1600. Stealth is the 65w one that comes with 1400. And max is OEM only, for 1600x and 1800x.

>>60402623
Neither because they don't fit my use case. I just bought a 1700x and I don't plan to change it in the next 5 years unless there are substantial across the board improvements (just did the same stint with my 1100T).

>>60406912
>roughly
An 1800X has 6900K-tier performance. Also, unlike any of the Ryzen chips, Intel's HEDT series can actually overclock to well past 4 GHz without a problem, and at that point Ryzen can't compete with those chips.

As I said earlier, if the 16c32t chips have better performance than the 7920X even after an overclock and AMD drops the price of the 1800X to that of the 6700K and prices its 16c32t at the old 1800X price, then yes, AMD will finally be king of the CPU industry.

>>60409539
>Also, unlike any of the Ryzen chips, Intel's HEDT series can actually overclock to well past 4 GHz without a problem, and at that point they STILL don't even come close to competing with similarly priced Ryzen in multicore
FTFY

>>60409539
>AMD drops the price of the 1800X to that of the 6700K
This is the stupidest thing I've read on /g/ today.

>>60409578
It also can't overclock past 4.1ghz without a problem because it gets really toasty.

>>60408668
>>60409408
source: the stilt
https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/
>The synchronization of the data fabric dictates that each of the enabled CCXs have identical number of cores enabled at all times. The available configurations are 1 (1:0), 2 (2:0 or 1:1), 3 (3:0), 4 (4:0 or 2:2), 6 (3:3), 8 (4:4).

>>60402623
I want more than 28 lanes, so that leaves AMD and Intel's 10 & 12C chips, which are likely to be ludicrously overpriced. That essentially leaves an AMD 10C as my only option, assuming it's going to be ~$500 or less since I'm not willing to pay more. Probably going to be more expensive though since a 1800X is already $500.

Maybe an i9-7800X and a PLX mobo if somebody makes one, or a Coffee Lel 6C with a PLX mobo. I honestly find the selection pretty shit. Everything that isn't ridiculously expensive is PCIe castrated.

>>60409683
On a single die, sure. But on two or more dies per processor we dont know if that is a requirement for ALL dies, or limited to a die to die basis (ex 4c+4c on die 1 and 3c+3c on die 2 for a 14c processor)

>>60409539
The only way for Intel so compete is to price their 10 cores at $500

>>60409600
You are retarded m8. The 6900k is $1,000 vs the 1800x's $490 and you just admitted the 1800x is ok par with it. Why the fuck would they lower the price of the 1800x?

Plus AMD can sell a 12-core Zen HEDT processor for 6900k prices and BTFO the shit out of it, AND they can sell a 16-core for less than the 7920X.

Sorry, shekelchaser. Broadswell-E getting BTFO'd and Skylake-X is DOA.

>>60409824
AMD can sell the 16core for $2000 and it would still be king since in destroys anything Intel has to offer.

>>60409815
Intel doesn't need to compete because it has an unbelievably loyal fanbase that is skeptical of AMD since it was second place for more than ten years now.

Why do you think not too many people are going crazy about Ryzen (other than those on /g/)?

>>60409594
If AMD does this, it WILL win.

File: 1488420066835.jpg (42KB, 898x886px) Image search: [Google]

42KB, 898x886px

>>60406549
Amazing ruse.
Shame it went over so many's heads.

>>60409897
>Source: My ass.

>>60409897
>Intel doesn't need to compete because it has an unbelievably loyal fanbase that is skeptical of AMD since it was second place for more than ten years now
Well yeah retarded people will always exist in this world, there's not much we can do about this fact. A world without retarded people would be an utopia.

>>60409948
>an utopia
Retard spotted.

>>60409923
It's not anything I'm pulling out of my ass. It's a fact. It's a simple plan AMD needs to follow if it's serious about taking back the CPU market.

File: 1491958654249.png (84KB, 653x726px) Image search: [Google]

84KB, 653x726px

>>60409824
>Broadswell-E getting BTFO'd and Skylake-X is DOA

Wait for Cannonlake.

>>60410016
>slower than 14nm
Ha ha well memed my blue friend.

>>60409983
Utopia = ideally perfect world
Sorry but you just confirmed that you're retarded.

>>60410065
>an utopia
>an
I'd tell you to deprive your brain of oxygen, but you can't deprive something you don't have of oxygen.

>>60409706
10C is two dies in MCM so at minimum the cost of 2x R7 1700

>>60410077
It's pronounced youtopia, faggot.

>>60410119
The y would be pronounced as a consonant, so you're still wrong

>>60410077
In early texts, when an indefinite article immediately preceded the English
noun utopia or utopian or the adjective utopian (or the eu- counterparts of such
words), the article was nearly always an, even though (during much of that interval,
at least) those "utopian" words were commonly pronounced with an initial consonant
/y/, as they are today. Only in the middle of the nineteenth century did constructions
like a utopia and a utopian start to appear with any regularity in the written record.
In the seventeenth and eighteenth centuries, thirty-three recorded instances of an occur before "utopian" words, only two of a.

File: 1489103005792.png (162KB, 633x900px) Image search: [Google]

162KB, 633x900px

>>60410036
S-SHUT UP

MODS = GODS

>>60410202
>The y would be pronounced as a consonant
Not a native speaker, but holy shit this is fucking dumb.

>>60402814
Both AMD and Intel measures TDP "differently". You have to add 10-15% to both.

Someone explain AMD's retarded naming scheme

>>60410482
>91w TDP 7700k puts out over 130w during regular usage
>95w TDP 1800X put outs 115 rarely
>add 10-15% to both
>both

>>60410536
It's not AMD's. It's made up by someone dreaming and thinking they're clever and only idiots believe this rumor to be true.

>>60410536
I dunno where that came from, does anyone have a link to the supposed Snowy Owl leak?

File: IMG_5991.jpg (19KB, 261x273px) Image search: [Google]

19KB, 261x273px

>>60409983
>It's a fact
>That I made up
>To shill for kiketel

>>60410642
Stupid anti-semitic goy

File: apprentice and second firing door in foreground.jpg (331KB, 1067x1600px) Image search: [Google]

331KB, 1067x1600px

>>60410687
>MFW all the kiketel shekelchasers start leaping into the oven voluntarily after AMD's 16c Whitehaven BTFO's Skylake-X

>>60410566
>It's not AMD's. It's made up by someone dreaming and thinking they're clever and only idiots believe this rumor to be true.
youknowhesright.jaypeg

>>60402814
>>60410482
>>60410566

Retards, the lot of you. All cpus that have some sort of boost can - and will - exceed TDP for a short time because (as shocking as it sounds) it takes time for a heatsink to heat up. In this totally made up for illustrative purposes example a 95w TDP chip can output more than (again, random example) 120w of heat until the heatsink can no longer dissipate it causing the chip to heat up and its thermal protection measures (i.e downclocking) kick in.

If you have a enough cooling and the cpu/motherboard are allowed to do it you can get 95w TDP chips to spike to nearly 200w if you can pump enough juice and keep the run time short enough. AVX2 is a great way to get Intel chips to exceed TDP for a while before they throttle down.