Intel Xeon Scalable "Sapphire Rapids" with HBM2E Beaten by Older AMD EPYC "Milan-X" in Leaked Benchmarks

[btarunr]

Intel's Xeon Scalable "Sapphire Rapids" processor may have a tough time getting to market, as leaked benchmarks suggest that even its premium HPC models with on-package HBM2E memory are outperformed by AMD's older-generation "Zen 3" EPYC processors. The 64-core/128-thread EPYC "Milan-X" processor based on older "Zen 3" microarchitecture with 3D Vertical Cache (3DV cache) chiplets, allegedly outperforms 52-core/104-thread Xeon Platinum 8472C and 60-core/120-thread Xeon Platinum 8490H "Sapphire Rapids" engineering samples in CPU-Z Bench and V-ray tests that scale across cores. These benchmark scores were compared with those of the EPYC "Milan-X" by Tom's Hardware, in which they well woefully short of the AMD chips.



View at TechPowerUp Main Site | Source

 

[Fouquin]

If Sapphire Rapids had launched anywhere close to its original release window Milan-X would be the newer chips with performance to prove. Reminder that SPR samples have been out in the wild since 2020. SPR is yet another bullet point under, "List of examples for why missing your deadlines is bad."

 

[Crackong]

Xeon Users:

 

[Dirt Chip]

It would be nice to see Epyc level of product (that is dominance in performance compare to competition) in AMD`s ZEN and RADEON line.
AMD advantage in servers is nothing short of amazing but for us (4-8 (16??) cores, mortal gamer beings) is of no real help.

Also, The chiplet design really show it`s strength in the very high core count.
Shame it`s no so good use (read: Epyc level) in desktop.
Maybe the next RADEON chiplet design will prospect a movement from the somewhat boring status-quo we are in (consumers segment).

 

[Chaitanya]

Ice Lake SP itself was a day late and dollar short now if Sapphire Rapids is also similar then high time Gelsinger gets the boot.

 

[ZoneDymo]

Ice Lake SP itself was a day late and dollar short now if Sapphire Rapids is also similar then high time Gelsinger gets the boot.

I really dont think Gelsinger can do anything about this, these things are planned years in advance, so he joined as they were happening, you can only start blaming him for stuff realistically in maybe 5 years? maybe more.

 

[bug]

CPU having more cores wins in heavily multithreaded test. Imagine that...

For servers, the #1 benchmark is how much energy does it take to complete a given task. That's harder to test for tech websites (systems will be different, mobos aren't widely available), but that is where Zen wins most of the time. Benchmarks like these are secondary concerns, at best.

 

[ZetZet]

Ice Lake SP itself was a day late and dollar short now if Sapphire Rapids is also similar then high time Gelsinger gets the boot.

Gelsinger is the one trying to save this sinking ship why would you boot him.

 

[to001]

somehow his score jump from 70k to 100k from Feb https://www.tomshardware.com/news/sapphire-rapids-milan-x-epic-fight.

and those 100k are from overclocked chips.

 

[Bomby569]

I really dont think Gelsinger can do anything about this, these things are planned years in advance, so he joined as they were happening, you can only start blaming him for stuff realistically in maybe 5 years? maybe more.

i doubt investors will be that patient. The problem is not the legacy stuff, the problem is you don't see a change in course.

 

[Minus Infinity]

Gelsinger's hubris of only a few months ago claiming AMD were in Intel's rear vision mirror is now a sad joke. Arc discrete cancelled, Sapphire Rapids delayed until 2023, Meteor Lake delayed, Ponte Vecchio barely shipped. What next other than more grandiose announcements about future products.

 

[Bomby569]

Arc discrete cancelled,

he dismissed that rumour

 

[1d10t]

Came late to the party, with HBM2E, yet manage to made home early and drunk.

 

[DeathtoGnomes]

Just goes to show how Core counts matter. Any real reviews/tests would include app speeds.

 

[Robin Seina]

he dismissed that rumour

Actually, Koduri did not. He said:
"we are shrugging about these rumors as well. They don't help the team working hard to bring these to market, they don't help the PC graphics community..one must wonder, who do they help?..we are still in first gen and yes we had more obstacles than planned to overcome, but we persisted."

That is neither yes or no. Just a politician's empty words statement.

What is interesting, that when you recalculate the two results with roughly same base clock (no. 1 and 4; 8472C has +300MHz in upper range of clock) to the perfomance for a single thread. You get results 424 and 426.

That says that IPC is slightly better for AMD's Milan-X. When the Bergamo/Genova will come to the market, it will probably leave Intel in the dust.

 

[Wirko]

Gelsinger's hubris of only a few months ago claiming AMD were in Intel's rear vision mirror is now a sad joke. Arc discrete cancelled, Sapphire Rapids delayed until 2023, Meteor Lake delayed, Ponte Vecchio barely shipped. What next other than more grandiose announcements about future products.

What they see in the rearview mirror right now looks like a "v", which is the central part of the middle letter in AMD.

 

[Minus Infinity]

he dismissed that rumour

Who's he? Gelsinger has said nothing, Raja is the only one denying it, and numerous internal sources are saying it's effectively over for Arc as of 2-3 days ago. Battelmage will be reduced to a few monolithic (chiplets are dead) lower end cards, and Intel admits it doesn't have the resources to compete with RDNA4 and Blackwell late 2024. It cannot put money into cards that will be smashed by RDNA3 and Lovelace, let alone hope to compete against next gen offerings.

 

[Bomby569]

Who's he? Gelsinger has said nothing, Raja is the only one denying it, and numerous internal sources are saying it's effectively over for Arc as of 2-3 days ago. Battelmage will be reduced to a few monolithic (chiplets are dead) lower end cards, and Intel admits it doesn't have the resources to compete with RDNA4 and Blackwell late 2024. It cannot put money into cards that will be smashed by RDNA3 and Lovelace, let alone hope to compete against next gen offerings.

you want a CEO responding to rumours? The right person replied, now how cynical are you is up to you

 

[Jism]

If Sapphire Rapids had launched anywhere close to its original release window Milan-X would be the newer chips with performance to prove. Reminder that SPR samples have been out in the wild since 2020. SPR is yet another bullet point under, "List of examples for why missing your deadlines is bad."

View attachment 261665

If im correct the SR is bugged or plagued with bugs.

2 years of delay.

 

[Lianna]

What they see in the rearview mirror right now looks like a "v", which is the central part of the middle letter in AMD.

They did see AMD in the rearview mirror, they just did not admit going the wrong way.