Intel's Ice Lake Xeon Processor Details Leaked: LGA 4189, 8-Channel Memory

[Raevenlord]

The Power Stamp Alliance (PSA) has posted some details on Intel's upcoming high-performance, 10 nm architecture. Code-named Ice Lake, the Xeon parts of this design will apparently usher in yet another new socket (socket LGA 4189, compared to the socket LGA 3647 solution for Kaby lake and upcoming Cascade Lake designs). TDP is being shown as increased with Intel's Ice Lake designs, with an "up to" 230 W TDp - more than the Skylake or Cascade Lake-based platforms, which just screams at higher core counts (and other features such as OmniPath or on-package FPGAs).

Digging a little deeper into the documentation released by the PSA shows Intel's Ice Lake natively supporting 8-channel memory as well, which makes sense, considering the growing needs in both available memory capacity, and actual throughput, that just keeps rising. More than an interesting, unexpected development, it's a sign of the times.



View at TechPowerUp Main Site

 

[Hood]

Last gen had 6 channel memory, now this with 8 channel - just like Epyc. What a surprise.

 

[voltage]

Looking forward to these, what iv been waiting for.

 

[Valantar]

Last gen had 6 channel memory, now this with 8 channel - just like Epyc. What a surprise.

Also, the socket now has (95) more pins than TR4. Even if it's highly unlikely that there's any causal relation, I do get a feeling of oneupmanship from this. It'll be interesting to see the PCIe lane count.

 

[TheGuruStud]

LOLtel and their TDPs.

 

[bug]

LOLtel and their TDPs.

Eh, I remember the days when AMD mocked Intel about their TDP numbers: Intel used to cite average consumption, while AMD took pride in advertising peak consumption instead. That's how a K7 used to beat a Prescott, despite both being nominally advertised as ~70W parts. These days it seems that has been somewhat reversed.

 

[Valantar]

LOLtel and their TDPs.

To quote AnandTech:

Also in this image (and verified at Power Stamp) are the power ranges for Cascade Lake (165-205W, similar to Skylake Xeons) and for Ice Lake (set to go to 230W). Should the new Ice Lake Xeon platform incorporate features such as OmniPath or on-package FPGAs, which Intel has teased future Xeon platforms to be, then 230W is well within the [realm] of possibility.

 

[voltage]

espite both being nominally advertised

Eh, I remember the days when AMD mocked Intel about their TDP numbers: Intel used to cite average consumption, while AMD took pride in advertising peak consumption instead. That's how a K7 used to beat a Prescott, despite both being nominally advertised as ~70W parts. These days it seems that has been somewhat reversed.

Spot on with facts. It wasn't long ago when I lived in very cold winter climate, I had two PC's one was an AMD, it would heat my room in a few minutes of use.

 

[SAL9000]

I saw this coming when AMD released the Threadripper. I had a feeling that Intel was going to have a new socket fairly quick to one-up AMD and presto here it is. I bet it will have more than 48 PCIe lanes and I'm also guessing it will have PCI 4.0 instead of 3.0 if they're smart and looking towards the future. I wonder if we are at the beginning of a pin count war. How many pins can a CPU have? Is 5000+ the next target?

 

[bug]

I saw this coming when AMD released the Threadripper. I had a feeling that Intel was going to have a new socket fairly quick to one-up AMD and presto here it is. I bet it has more the 48 lanes on PCIe and I'm guessing it will have PCI 4.0 instead of 3.0 if they're smart and looking towards the future. I wonder if we are at the beginning of a pin count war. How many pins can a CPU have? Is 5000+ the next target?

Well, with the memory controller on the CPU die, you can't really add more memory channels without fiddling with the pins.
Still, in that market, the cost of a motherboard is peanuts anyway (some high-end desktop motherboards can be more expensive than a server one, while the CPU is signitifcantly more expensive).

 

[efikkan]

I saw this coming when AMD released the Threadripper. I had a feeling that Intel was going to have a new socket fairly quick to one-up AMD and presto here it is. I bet it will have more than 48 PCIe lanes and I'm also guessing it will have PCI 4.0 instead of 3.0 if they're smart and looking towards the future. I wonder if we are at the beginning of a pin count war. How many pins can a CPU have? Is 5000+ the next target?

To be fair, development of new chips usually take 2-3 years, so we wouldn't see any changes resulting from Zen until late 2019 or later. Any hardware changes arriving before that will be the result of decisions prior to the launch of Zen.

 

[TheGuruStud]

Eh, I remember the days when AMD mocked Intel about their TDP numbers: Intel used to cite average consumption, while AMD took pride in advertising peak consumption instead. That's how a K7 used to beat a Prescott, despite both being nominally advertised as ~70W parts. These days it seems that has been somewhat reversed.

Intel lied (and for so long) that AMD had to sink to their level. Not that TDP is the same as consumption, but should be close.

 

[bug]

Intel lied (and for so long) that AMD had to sink to their level.

Ah, the "two wrongs make a right" argument.

 

[TheGuruStud]

Ah, the "two wrongs make a right" argument.

No, but that's what happens when the competitor uses it to sell junk products for higher prices. Bulldozer made it easy to switch over lol

Victim blaming is why AMD is in the position they are now. Intel dindunuffin!

If they do everything ethically or honest with consumers = looks bad compared to competitor. If they do even a fraction of the same tactics as the others = bad AMD. There's no winning with a rigged market and idiots.

 

[bug]

No, but that's what happens when the competitor uses it to sell junk products for higher prices. Bulldozer made it easy to switch over lol

Victim blaming is why AMD is in the position they are now. Intel dindunuffin!

If they do everything ethically or honest with consumers = looks bad compared to competitor. If they do even a fraction of the same tactics as the others = bad AMD. There's no winning with a rigged market and idiots.

Idk, they did the right thing back in AthlonXP days and it did them a lot of good. Those were the days people started to take AMD seriously (i.e. they stopped being just a cheap alternative for "real" processors).
What AMD did wrong was to believe that because they had a better product, Intel would simply lay down and die. When that didn't happen, AMD had no contingency plan. TDP really had nothing to do with it. Still, I'm not going to fault AMD for switching tactics simply because there's no stadard way to measure TDP anyway.

 

[TheGuruStud]

Idk, they did the right thing back in AthlonXP days and it did them a lot of good. Those were the days people started to take AMD seriously (i.e. they stopped being just a cheap alternative for "real" processors).
What AMD did wrong was to believe that because they had a better product, Intel would simply lay down and die. When that didn't happen, AMD had no contingency plan. TDP really had nothing to do with it. Still, I'm not going to fault AMD for switching tactics simply because there's no stadard way to measure TDP anyway.

Athlon XP was when Intel went ballistic and strongarmed every OEM and large store on the planet. Intel was finished in the long run until that happened. AMD marketshare tanked and continued to do so even after the best CPU ever made was released (athlon 64). That was the beginning of the end. Even their paper marketing was just as sinister. Lie after lie in publications, owning a benchmark to show in pretty little graphs how netburst is better...

 

[lexluthermiester]

Looking forward to these, what iv been waiting for.

And what would you do with it?

 

[bug]

Athlon XP was when Intel went ballistic and strongarmed every OEM and large store on the planet. Intel was finished in the long run until that happened. AMD marketshare tanked and continued to do so even after the best CPU ever made was released (athlon 64). That was the beginning of the end. Even their paper marketing was just as sinister. Lie after lie in publications, owning a benchmark to show in pretty little graphs how netburst is better...

Yeah, like AMD didn't use Cinebench to show the world how Ryzen will crush Kaby Lake...

 

[TheGuruStud]

Yeah, like AMD didn't use Cinebench to show the world how Ryzen will crush Kaby Lake...

Cinebench is historically heavily Intel favored. Were you just born? Their website used to be plastered with 10 Intel logos. It would be smart marketing to use Intel fanboys' favorite bench and turn it on them. You mad, bro?

However, I remember AMD using Blender.

And how is that a rebuttal to controlling a prominent synthetic bench to completely defraud consumers?

You Intel shills really have to work on your arguments. They're not even passable as trolling.

 

[Prima.Vera]

Looking forward to these, what iv been waiting for.

How many VMs planing to run on the same time??

 

[lexluthermiester]

Cinebench is historically heavily Intel favored. Were you just born? Their website used to be plastered with 10 Intel logos. It would be smart marketing to use Intel fanboys' favorite bench and turn it on them. You mad, bro?

Hey, be nice!

 

[SAL9000]

To be fair, development of new chips usually take 2-3 years, so we wouldn't see any changes resulting from Zen until late 2019 or later. Any hardware changes arriving before that will be the result of decisions prior to the launch of Zen.

You may be right but so I am since I have no inside track to processor. I can speculate all I want about upcoming tech and in this case I was correct. I saw the I/O advantage of the Threadripper/Epyc in terms of PCIe lane count and how nice it was not to have PCIe switches on the motherboard. Having direct PCIe connectivity to the CPU is quite helpful. So I guessed that Intel was working on something to surpass the Threadripper/Epyc CPU but we still don't know a lot of the specs of this CPU socket and this CPU is probably not going to be used in consumer PC's so AMD will still have an advantage in lane count for now.

 

[THU31]

My CPU only has 1151 pins.