Discussion Intel current and future Lakes & Rapids thread

Page 769 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Markfw

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,258
16,115
136
Geddagod said:
No idea what happened to SPR.
I would love to see the power allocation on these SPR parts vs Milan or Rome though. If there was a non-hardware related reason why SPR is weaker than Milan in some cases, this would be my best guess. Power allocation becomes hugely important in these parts, with IO and L3 cache power consumption tweaks being able to create drastic changes in performance- so much so that Milan vs Rome was originally a 15% perf/watt regression, but then with shifts in power to the core versus IO and L3 cache in a different system, it obviously became a perf/watt uplift over rome (in Anandtech's testing).
What makes this even more believable is that CB23 isn't exactly a memory intensive workload, so shifting power away from the IO towards the cores themselves could show benefit.
However this is only if there is a non-hwardware related issue. It could just be Intel 7+GLC is way worse at lower voltages/clocks than Zen 3 (which is also true I think). Also, the difference is so small between what I personally expected (slightly faster than Milan) versus reality (slightly slower than Milan) that I don't think there really is an issue per se, just SPR performing worse than expected.
Click to expand...
I am not sure exactly what your question is about Milan, just let me say, my DOUBLE 7763 Milan system with all 128 cores and 256 threads @ 100% use, for the entire system, not the CPUs, with 384 GIG of DDR4 3200 and 6 case fans, is less than 589 watts, They are rated at 225-280, and not sure what my motherboard is doing, but they are certainly using less than 280 watts each for 128 threads each. Does that answer your question ?

So, 64c/128t Milan uses less than 24c/48t SPR. for sure. I don't think you want to compare to Genoa....

Edit: and pretty sure they run faster than 2.5 ghz.... ALL CORE. Base clock is 2.45
 
Last edited:
  • Like
Reactions: Tlh97, moinmoin and lightmanek
G

Geddagod

Golden Member
Dec 28, 2021
1,541
1,627
106
Markfw said:
I am not sure exactly what your question is about Milan, just let me say, my DOUBLE 7763 Milan system with all 128 cores and 256 threads @ 100% use, for the entire system, not the CPUs, with 384 GIG of DDR4 3200 and 6 case fans, is less than 589 watts, They are rated at 225-280, and not sure what my motherboard is doing, but they are certainly using less than 280 watts each for 128 threads each. Does that answer your question ?

So, 64c/128t Milan uses less than 24c/48t SPR. for sure. I don't think you want to compare to Genoa....
Click to expand...
It wasn't a question. There were literarily 0 question marks in that comment.
Infact nothing about those numbers bear any relevance to my comment in question.
My point was that, like Milan did at first, Sapphire Rapids may be pushing excessive power to the L3 cache or IO, and less to the cores themselves, resulting in lower scores in compute heavy tasks that don't benefit much from L3/Memory (like Cinebench). Milan doing in a faulty system resulted in Anandtech testing showing a regression in perf/watt over Rome.
And this would be my best guess of what's wrong with SPR if infact something was wrong with SPR beyond just the hardware not performing as well as I and some others expected it would. It most likely is just SPR just being marginally worse than expected, but the idea of driver issues and related problems maybe hindering performance was an idea that was floated by not just me but the reviewers over at Puget Systems. I'm guessing a definitive answer will show up in a month or so.
 
Markfw

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,258
16,115
136
Geddagod said:
It wasn't a question. There were literarily 0 question marks in that comment.
Infact nothing about those numbers bear any relevance to my comment in question.
My point was that, like Milan did at first, Sapphire Rapids may be pushing excessive power to the L3 cache or IO, and less to the cores themselves, resulting in lower scores in compute heavy tasks that don't benefit much from L3/Memory (like Cinebench). Milan doing in a faulty system resulted in Anandtech testing showing a regression in perf/watt over Rome.
And this would be my best guess of what's wrong with SPR if infact something was wrong with SPR beyond just the hardware not performing as well as I and some others expected it would. It most likely is just SPR just being marginally worse than expected, but the idea of driver issues and related problems maybe hindering performance was an idea that was floated by not just me but the reviewers over at Puget Systems. I'm guessing a definitive answer will show up in a month or so.
Click to expand...
Sorry it did not answer your question. But seeing as it uses considerably more than twice the power/core than Milan, and scores worse than Zen 2 threadripper, I don't see how a bios fix will do anything to make it come even close to Milan, let alone Genoa. (its real competitor)
But I will certainly look forward to any answers we find in the coming months.
 
BorisTheBlade82

BorisTheBlade82

Senior member
May 1, 2020
710
1,132
136
Exist50 said:
This question got me thinking. Is throwing more and more silicon at their server platforms a sustainable long-term solution for Intel, AMD, etc? At what point is the cost so high that it alienates too much of the market? Or do they just keep further subdividing the server platform into more socket tiers? Eventually they'll either need new memory tech, or they're not going to be able to fit enough physical memory channels in a standard width rack... That, or things plateau for a while.
Click to expand...
Well, there is always an option for disruptive technology advancements. But for the foreseeable future there can only be more silicon. And that is the beauty about AMDs chiplets. They can keep cost increasements on a nearly linear scale, while monolithic approaches are not only nonlinear (more exponential), but also have a hard limit in the reticle size.
Then there is consumption for compute as well as data transfer. IMHO Lisa's ISSCC presentation is a good overview about challenges for the near future.
 
  • Like
Reactions: Tlh97, Vattila, Elfear and 4 others
K

Kocicak

Golden Member
Jan 17, 2019
1,177
1,232
136
What is the drama all about? They just built a CPU from the cores they have, on the process they have with expected power to performance ratio, with added features compared to desktop models as memory support and more PCIe lanes, and are getting expected performance.

I am not sure that Cinebench can use that extra memory support and more PCIe lanes.

The fact that the CPUs for the new Intel workstation platform may not be very competitive now may explain why we have not seen many reviews of these CPUs yet, I am not even sure if any "official review" bas been published yet.

In the future they surely can bring more competitive products on this new platform... Surely.
 
G

Geddagod

Golden Member
Dec 28, 2021
1,541
1,627
106
Kocicak said:
What is the drama all about? They just built a CPU from the cores they have, on the process they have with expected power to performance ratio, with added features compared to desktop models as memory support and more PCIe lanes, and are getting expected performance.

I am not sure that Cinebench can use that extra memory support and more PCIe lanes.

The fact that the CPUs for the new Intel workstation platform may not be very competitive now may explain why we have not seen many reviews of these CPUs yet, I am not even sure if any "official review" bas been published yet.

In the future they surely can bring more competitive products on this new platform... Surely.
Click to expand...
No drama. Also SPR was not expected to perform as bad as it is, though to be honest it did not miss the expected mark by a wide margin. Let's not pretend like everyone was not surprised by SPR's performance as early reviews started rolling in, you can go back in this thread and check yourself a couple weeks ago.
SPR, in theory, should perform a bit better. Intel 7 should be on par with TSMC 7nm, GLC client has better IPC than zen 3, not to mention GLC server, all while the GLC client parts, such as the 12400, have essentially the same efficiency as Zen 3 5600x in Cinebench and SuperPi.
Whether it's because of mesh, power distribution, or Intel 7 just being drastically worse at extremely low voltages compared to TSMC 7nm compared to their performance at higher voltages, something is making the efficiency gap seen in client parts just not really match up in server.
 
  • Like
Reactions: Tlh97, lightmanek and Elfear
poke01

poke01

Diamond Member
Mar 8, 2022
4,274
5,601
106
nicalandia said:
Elmore labs World Record attemp

View attachment 77828

Look at that efficiency, 1300 Watts
Click to expand...
Man. What has become of Intel?

Truly, a fall from grace. SPR is just sad. Does Intel not know if you target efficiency first then they benefit from laptop to server???

This is what AMD and Apple do. They use core designs that scale well across their product lines. Intels designs seem to not make sense.
No consistentcy.
 
  • Like
Reactions: Tlh97 and scannall
poke01

poke01

Diamond Member
Mar 8, 2022
4,274
5,601
106
Geddagod said:
TSMC 7nm compared to their performance at higher voltages, something is making the efficiency gap seen in client parts just not really match up in server.
Click to expand...
It's Golden Cove. It's a big core that is not meant for Servers. More like desktop. Zen 4 is smaller than Golden Cove.
 
  • Like
Reactions: Tlh97 and Edrick
I

igor_kavinski

Lifer
Jul 27, 2020
28,175
19,192
146
moinmoin said:
Yeah, just need something that is both ever more harder and expensiver to achieve. Intel really has its work cut of for it there. Easy peasy. 😁
Click to expand...
Time for them to poach some AMD engineers and demote a few of their senior CPU designers on the server team.
 
K

Kocicak

Golden Member
Jan 17, 2019
1,177
1,232
136
86K points at 800W is in line with what desktop Raptor lake CPUs do - 40K points at 300W.

The only problem here is the manufacturing process.
 
K

Kocicak

Golden Member
Jan 17, 2019
1,177
1,232
136
nicalandia said:
Which is bad, Zen2 based Threadripper 3990X can do better than that with much less power
Click to expand...
Yes. The question is, why Intel decided to use this inefficient process to make these CPUs, are there technical reasons (this design is for example impossible to make on any TSMC process), financial (it would be too expensive to use any other process than their own), strategic (perhaps Intel does not even want or need to sell a lot of these CPUs, because they know the next gen will be much more competitive), purely image building exercise so that they can present their workstation lineup to somebody, contractual (they are oblidged by some contract to make and deliver such CPUs), or what else there could be?
 
Last edited:
Markfw

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,258
16,115
136
Kocicak said:
86K points at 800W is in line with what desktop Raptor lake CPUs do - 40K points at 300W.

The only problem here is the manufacturing process.
Click to expand...
I am sure a better MFG process would help, but thats not the only problem. Look at Genoa vs Milan !
 
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom