[VR-Zone] Intel Haswell-EP Brings DDR4, to Eat up to 160W, 190 Amps

[zalbard]

As Nova elaborated in his exhaustive and insightful article with Ivy Bridge-EP in 2013, Intel will bring out higher TDP-draw processors, and that trend is set to continue with up to 16-core Haswell-EP monster which targets the introduction in the second half of 2014. The main competitor for Haswell-EP will not be AMD with their unannounced successor of the yet unannounced Abu Dhabi Server CPU (confused yet?), it will be a combination of NVIDIA Maxwell GPU paired with an Project Denver-based Server CPU.

The processor will be available at different clock speeds and with the different amount of enabled CPU cores, but the main power characteristics are 120W, 135W, 145W and 160W. There will not be any sub-100W TDP parts, as the company was told that it has to compete against NVIDIA Tesla GPUs, which all come in 225 and 300W range. With Tesla K10 (Dual-GK104, server version of GTX 690 with ECC and 8GB of GDDR5 memory) coming at 375W, and GK110-based K20 drawing 300W with 12GB of GDDR5 memory, it is obvious what Haswell-EP has to do.

Power draw characteristics require dedicated four and five rail solutions to feed the CPU. We were shown a prototype of the platform which draws less than 100 Amps in typical working conditions, with Turbo Mode jacking that up to 120 Amps. What will make overclockers wet is the fact that Haswell silicon is designed to accept draws of up to 190 Amps on AIR, meaning we might get another overclocking king if its able to sustain 190A when you freeze it and increase the voltage to let's say, 1.575V i.e. 300W.

As far as the platform goes, The 2013/14 server and workstation platform is called Grantley. It will support DDR3 or DDR4, with extensive memory options. We were a bit intrigued by the amount of DIMMs/SO-DIMMs which Intel supports with the single 72-bit (64+8) lane (times four and six), bringing this author back into 1999 and the 8-channel RDRAM DEC Alpha… funny to see history repeating itself with the amount of supported DIMMs per channel.

Source: http://vr-zone.com/articles/

[qcmadness]

Dual memory support...
but this is server version.

[PatRaceTin]

yessssssss if they don't use paste inside cpu

[K404]

nVidia led the way with massive GPU power consumptions, now they're forcing Intel to increase envelopes too. BOOOOOOOO nVidia, BOOOOOOOOOOOO

[paulbagz]

Believe it when I see it.

-PB

[zanzabar]

isnt ddr4 just ultra low volt ddr3 with higher latency, it seams like ddr3 would be a draw back since the cpu has to be running at less than 1V

nVidia led the way with massive GPU power consumptions, now they're forcing Intel to increase envelopes too. BOOOOOOOO nVidia, BOOOOOOOOOOOO

why boo, this is not the only platform intel has, and the only thing keeping the higher end chips higher end right now is NV. sure amd has a beast for VM farms but intel owns that with marketing and IT stupidity so the only thing that they have now is to keep server loads in x86 instead of cuda/opencl. this is also the base of multi cpu servers so why would they go for 4 slow parts when you could have 2 fast parts do the same work load.

[Boogerlad]

please please use paste instead of soldering, I love removing the ihs. I hope that the high power envelope doesn't mean just high clocks, but high ipc in this architecture.

[K404]

why boo, this is not the only platform intel has, and the only thing keeping the higher end chips higher end right now is NV. sure amd has a beast for VM farms but intel owns that with marketing and IT stupidity so the only thing that they have now is to keep server loads in x86 instead of cuda/opencl. this is also the base of multi cpu servers so why would they go for 4 slow parts when you could have 2 fast parts do the same work load.

I imagine it would still be a selling point "cut electricity running costs by 50%" It still puts pressure on the opposition. The move would take 1-2 generations, then performance starts increasing again with successive releases.

[sholvaco]

You know, they could start by offering a fully functional and fully unlocked SB-E die (8 cores, 20 megs of L3, QPI, unlocked multipliers). Ensuring this within 160W envelope should be a breeze.

[zanzabar]

I imagine it would still be a selling point "cut electricity running costs by 50%" It still puts pressure on the opposition. The move would take 1-2 generations, then performance starts increasing again with successive releases.

typically you can force a lower pstate on servers to get the wanted TDP, i do not see why they should aim at lower total wattage when the wattage per flop or operation will be lower. if you look at the NV cuda cards now they use about half the power of the sb-e to do the same operation.

[Tenknics]

I missed a lot of other stuff posted about Ivy-EP and EX...Is vr-zone reliable?

I just read a few of their articles, and its making a lot of sense. Why SB-E turned out the way it is, the difference's going forward with -E versions, specifically starting with Ivy. Also why SB-EX was skipped, going strait to Ivy-EX.

Also this leads me to believe Intel ISN'T skipping Ivy-E for Haswell-E. Same misinformed and confused people did this with Sandy. They all forget that the mainstream 4 core versions come out first, followed much later by the server version/high end desktop versions. People saw only the Ivy 4cores that just came out and freaked/assumed the worst.

Business as usual for Intel. I would love to see a 10-core ivy-e on the desktop but I will be happy with a fully enabled 8 core/Fixed/enhanced Patsburg Ivy setup!

[zalbard]

I missed a lot of other stuff posted about Ivy-EP and EX...Is vr-zone reliable?

I'd say so, yes!

I just read a few of their articles, and its making a lot of sense. Why SB-E turned out the way it is, the difference's going forward with -E versions, specifically starting with Ivy. Also why SB-EX was skipped, going strait to Ivy-EX.

Also this leads me to believe Intel ISN'T skipping Ivy-E for Haswell-E. Same misinformed and confused people did this with Sandy. They all forget that the mainstream 4 core versions come out first, followed much later by the server version/high end desktop versions. People saw only the Ivy 4cores that just came out and freaked/assumed the worst.

Yeah, I think that's the case. Their high-end platform direction has changed, but that's about it.

[Tenknics]

There is no more like button..nice find zalbard.

This news is the only type I care about these days really...high end CPU news...Its the only thing that excites me these days. SoC's and mobile computing and all that jazz is just so blaaaah

[iddqd]

This news is the only type I care about these days really...high end CPU news...Its the only thing that excites me these days.

unf oh god intel don't stop

[Olivon]

http://www.chiphell.com/thread-497401-1-1.html

Bigpao007 strikes again

[FlanK3r]

19031660k8sv66fzxwdxx8.jpg

[RPGWiZaRD]

Looks like I'll wait until P87 b4 upgrading, Ivy was too disappointing to me to bother invest my money. I used to swap CPU once a year or so, A64 3700+ -> Opteron 165 -> E6750 -> E8400 -> i7-860. Now looks like I'll skip both Sandy and Ivy. ^^ Sandy upgrade had made sense at launch, Ivy didn't to me.

[Tenknics]

19031660k8sv66fzxwdxx8.jpg

so nothing new for a year on the chipset front?

[god_43]

isnt ddr4 just ultra low volt ddr3 with higher latency, it seams like ddr3 would be a draw back since the cpu has to be running at less than 1V



why boo, this is not the only platform intel has, and the only thing keeping the higher end chips higher end right now is NV. sure amd has a beast for VM farms but intel owns that with marketing and IT stupidity so the only thing that they have now is to keep server loads in x86 instead of cuda/opencl. this is also the base of multi cpu servers so why would they go for 4 slow parts when you could have 2 fast parts do the same work load.

source? i thought they performed almost the same as the intel parts?

[SoF]

Looks like I'll wait until P87 b4 upgrading, Ivy was too disappointing to me to bother invest my money. I used to swap CPU once a year or so, A64 3700+ -> Opteron 165 -> E6750 -> E8400 -> i7-860. Now looks like I'll skip both Sandy and Ivy. ^^ Sandy upgrade had made sense at launch, Ivy didn't to me.

think about it after new stepping - everything seems better now...had chips (actually the first week 18 I had) that did 1.3V 32m @5GHZ...Intel seems to improve 22nm..


....remarkable!

[kuroikenshi]

Looks like I'll wait until P87 b4 upgrading, Ivy was too disappointing to me to bother invest my money. I used to swap CPU once a year or so, A64 3700+ -> Opteron 165 -> E6750 -> E8400 -> i7-860. Now looks like I'll skip both Sandy and Ivy. ^^ Sandy upgrade had made sense at launch, Ivy didn't to me.

Well, after my experience in the last 4 years...

PII -> PII -> I5 750 -> 2600k -> Brazos e-350n -> E5300 -> IB 3570K, I can defetinely attest that my newly acquired IB is a much better bang for back than any of the preceeding systems, it outperforms my 2600k and it did cost me 30% less. god the RAM alone cost me half for twice the bandwith and size. I was thinking of waiting for Haswell and DDR4 awesomeness, improved processed, etc, but afteer the Bulldozer fiasco I refuse to sell my system in hope that the next platform's performance and price exceed the current's.

[paulbagz]

Here's hoping for a new stepping IB that has the TIM issue fixed.

-PB

[kuroikenshi]

mine is running at 4.2 Ghz only doing 38-42 Celsius.
and the fact that I am running ram at 2400MHz straight out of the box is also something of an advantage when I was messing around with 2600k

snip

Thanks for that my helpful Frenchman.

[zanzabar]

source? i thought they performed almost the same as the intel parts?

a "cheap" $1500 per chip socket g34 setup will outperform an extremely expensive 4p 1567 that are $3k-4k each chip and the amd will use less power, so per watt and cost amd is a beast.

[BeepBeep2]

I wish people would understand TDP does not mean power consumption.