News [Anand] Intel's Enterprise Extravaganza, Cascade Lake Launches

formulav8

Diamond Member
Sep 18, 2000
6,998
11
126
#2
Intel looks to be using their own glue now.
 
Jan 31, 2019
108
15
51
#3
400W TDP is good for a 56C 112T CPU. But i dont think this will work inside rack server without extreme liquid cooling.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
17,580
1,034
136
#4
400 wattTDP for 56 cores in a server ? who would want that when the EPYC 64 core is like 250 watt. And probably cost less.
 
Jan 31, 2019
108
15
51
#5
32 core 2990wx is also 250W, that is 7.81w per core. You can expect 3.9w per core for 64 core epyc at 250W. Which can cause poor single thread performance due to power limit.
 

Thunder 57

Senior member
Aug 19, 2007
653
126
136
#6
Intel's 10nm failure bites them yet again. 400W??! That's insane.

32 core 2990wx is also 250W, that is 7.81w per core. You can expect 3.9w per core for 64 core epyc at 250W. Which can cause poor single thread performance due to power limit.
I don't see it hurting single thread performance, not that anyone using these would care about that.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
17,580
1,034
136
#7
32 core 2990wx is also 250W, that is 7.81w per core. You can expect 3.9w per core for 64 core epyc at 250W. Which can cause poor single thread performance due to power limit.
2990wx is not 7nm, and it runs faster than EPYC, so yes, I am pretty sure its going to be 180-250 watt TDP
 
Jan 31, 2019
108
15
51
#8
Intel's 10nm failure bites them yet again. 400W??! That's insane.



I don't see it hurting single thread performance, not that anyone using these would care about that.
It will hurt single thread performance, when used in running many VMs at same time. Base clock is 1.4ghz and maximum is 2.2ghz for this 64c 128t rome cpu.
 

Mopetar

Diamond Member
Jan 31, 2011
4,465
390
126
#10
I think that this table was my favorite part of the AT write-up:

Table.png

Those prices :tearsofjoy:
 
Jan 31, 2019
108
15
51
#11
I think that this table was my favorite part of the AT write-up:

View attachment 4817

Those prices :tearsofjoy:
3.8ghz is good enough for this core packed cpu, but these are not everyone's cpus. They might cost like 25k usd. Lol. Only scientists, universities and military might use it. But at this cost, they might consider dual epyc system.
So, a big LOL.
 

Thunder 57

Senior member
Aug 19, 2007
653
126
136
#12
It will hurt single thread performance, when used in running many VMs at same time. Base clock is 1.4ghz and maximum is 2.2ghz for this 64c 128t rome cpu.
I'm not going to speculate on final clocks. I doubt many expected 3.6/4GHz out of the original Zen at launch for an 8 core. Even if they are, maybe AMD decides to make a high TDP version to allow for higher speeds seeing as now 350-400W is now apparently acceptable.
 
Jan 31, 2019
108
15
51
#13
I'm not going to speculate on final clocks. I doubt many expected 3.6/4GHz out of the original Zen at launch for an 8 core. Even if they are, maybe AMD decides to make a high TDP version to allow for higher speeds seeing as now 350-400W is now apparently acceptable.
You're right man, maybe these have configurable TDP like xeons have
 
Mar 10, 2004
28,493
224
126
#14
400W TDP is good for a 56C 112T CPU. But i dont think this will work inside rack server without extreme liquid cooling.
It's apparently only a package deal? So cooling will be included?

https://www.servethehome.com/intel-xeon-platinum-9200-formerly-cascade-lake-ap-launched/

Since Intel is selling the Platinum 9200 series affixed to PCB, the company has three primary S9200WK compute modules. A 1U half-width liquid cooled sled along with two 2U modules, one liquid cooled and one air cooled. The entire premise of the Intel Xeon Platinum 9200 series is to effectively take a four-socket server and put it in a half-width form factor so one can effectively get eight Xeon Platinum 8200 series chips per U of rack space.
 

Thunder 57

Senior member
Aug 19, 2007
653
126
136
#15
You're right man, maybe these have configurable TDP like xeons have
Also, didn't we just see a quad core Zen 2 ES at 3.4/3.8GHz? I wouldn't read too much into clocks yet. That would be slower than the 4 core Ryzen 3's and the 4 core Raven Ridges that currently exist.
 

Zucker2k

Senior member
Feb 15, 2006
751
59
136
#16
56 cores crunching on AVX512 code and consuming 400watts is okay in my book.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
17,580
1,034
136
#18
3.8ghz is good enough for this core packed cpu, but these are not everyone's cpus. They might cost like 25k usd. Lol. Only scientists, universities and military might use it. But at this cost, they might consider dual epyc system.
So, a big LOL.
You don't need a dual EPYC to beat this thing. Just one 64 core would beat it. The 3,8 is only one core.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
17,580
1,034
136
#19
56 cores crunching on AVX512 code and consuming 400watts is okay in my book.
They won't be running anywhere close to the speeds in that chart running AVX512 due to heat. It remains to be seen how well Rome does on AVX512 or even at what speed, but at 7 nm, it will surely beat the 9200 series in heat and power draw,.
 

jpiniero

Diamond Member
Oct 1, 2010
6,319
220
126
#20
You don't need a dual EPYC to beat this thing. Just one 64 core would beat it. The 3,8 is only one core.
Given the monster TDP and AVX-512, it is going to be faster on anything that heavily uses AVX-512 even compared to Rome. Which was the point, it was supposed to be a replacement for the Phi. Funny thing is, the raw throughput isn't that much better. The Phi 7290's theoretical DP perf is 3.5 DP GFlops, roughly same as the 48 core XP 9242 (at it's base), and has a lower TDP to boot. Of course the whole reason they dumped the Phi was because they weren't getting anywhere near that peak theoretical perf.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
17,580
1,034
136
#21
Given the monster TDP and AVX-512, it is going to be faster on anything that heavily uses AVX-512 even compared to Rome. Which was the point, it was supposed to be a replacement for the Phi. Funny thing is, the raw throughput isn't that much better. The Phi 7290's theoretical DP perf is 3.5 DP GFlops, roughly same as the 48 core XP 9242 (at it's base), and has a lower TDP to boot. Of course the whole reason they dumped the Phi was because they weren't getting anywhere near that peak theoretical perf.
Lets wait until Rome benchmarks come out. The bottom line is that 14nm can't compare to 7 nm in power usage or heat. You can't beat physics,.
 

jpiniero

Diamond Member
Oct 1, 2010
6,319
220
126
#22
Lets wait until Rome benchmarks come out. The bottom line is that 14nm can't compare to 7 nm in power usage or heat. You can't beat physics,.
I didn't say it would be competitive in power usage or heat though, I was talking raw performance in DPFP. Which is the market aimed with the AP line.
 
Jan 31, 2019
108
15
51
#23
Lets wait until Rome benchmarks come out. The bottom line is that 14nm can't compare to 7 nm in power usage or heat. You can't beat physics,.
Yes, we cant beat physics, but there is not much large difference between 22nm E5 2696V3 (unlocked @240W TDP) vs 12nm 2950X (@180W TDP). (when both cpus compared @ same 3.4GHz all core boost).
So, i don't think there will be very large performance and power difference between 12nm vs 7nm
 

rainy

Senior member
Jul 17, 2013
394
19
116
#24
32 core 2990wx is also 250W, that is 7.81w per core. You can expect 3.9w per core for 64 core epyc at 250W. Which can cause poor single thread performance due to power limit.
This is a very strange comparison: you have Epyc 7601 which is 32 cores CPU with 2.2 GHz base clock and 180 W TDP and made of course in 14nm.
It would be no problem for AMD to have 64 cores Rome at similar clock, higher IPC and comparable TDP (about 200 W) in 7nm.
https://en.wikipedia.org/wiki/Epyc
 
Last edited:

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
17,580
1,034
136
#25
Yes, we cant beat physics, but there is not much large difference between 22nm E5 2696V3 (unlocked @240W TDP) vs 12nm 2950X (@180W TDP). (when both cpus compared @ same 3.4GHz all core boost).
So, i don't think there will be very large performance and power difference between 12nm vs 7nm
So you handicap the 2950x, just so the Xeon can get close ? And 16 cores vs 22 ? I have 4 Xeons, but they are 14 core@2.5 ghz. My 2700x beats them with only 8 cores@stock ! A Ryzen 3xxx series or Rome will wipe the floor with them.
 


ASK THE COMMUNITY

TRENDING THREADS