Intel Xeon W-3400/2400 "Sapphire Rapids" Processors Run First Benchmarks

[lemonadesoda]

Maybe we need to start including Mac processors in the benchmark comparisons. I bet the M2 Pro found in the Mac mini would beat the new Xeons in Performance per core and Performance per watt.

i find the new Xeons so underwhelming I’m even considering coming out of the x86 closet.

 

[AnotherReader]

Maybe we need to start including Mac processors in the benchmark comparisons. I bet the M2 Pro found in the Mac mini would beat the new Xeons in Performance per core and Performance per watt.

Let's wait for the Ryzen 7040 before writing off x86. As far as comparing the Xeons to the M2, Apple enjoys the advantage of a better process than Intel. The soldered RAM and limited IO also helps with the system's power consumption.

 

[Wirko]

A 13900K and 7950X score exactly the same in CB23, one has 24 cores (8+16) and the other 16, meaning an e-core is quite literally half of a Zen4 core performance wise. Seems pretty horribly underpowered to me.

That's only horrible if you're somehow assuming that an E-core should somehow be equally powerful as a P-core.

 

[Minus Infinity]

Wow getting beat down by ancient TR is embarrassingly bad. MAybe if it used half the TR's power it would be fine. AMD is really dropping the ball though with Zen 4 TR not out until Q4 at the earliest, possibly Q1 2024.

 

[Chaitanya]

der8bauer had posted a video showing overclocked 56core Xeon W beating TR in Geekbench while sucking 650W+ of power from wall. Just like that Ice lake SP, these new CPUs too little too late while barely managing to catch up with last Gen from competition.

 

[Crackong]

To beat top TR, intel will probably need to exchange some P-cores with E-cores.
Going 32p+96e (160 threads, you trade 1p for 4e) or so will probably take the lead in most situations.

Mixing P&E cores is a big NONO for the intended use of these CPUs.
These CPUs are meant for Server type workloads, it needs wide instruction support, and heavily virtualized environments.

E cores lacks some instruction sets , and virtualization software don't like Hybrid architectures.

So P core only or E core only.
Just don't mix them in a single CPU.

 

[Patriot]

Maybe we need to start including Mac processors in the benchmark comparisons. I bet the M2 Pro found in the Mac mini would beat the new Xeons in Performance per core and Performance per watt.

i find the new Xeons so underwhelming I’m even considering coming out of the x86 closet.

They are strong competition on laptops, but there is a reason the Mac Pro keeps getting delayed, their current design doesn't scale to enough cores to compete with workstation/enterprise class.
M2 Max is about on par with AMD's 7700x, also known as 1 Zen4 CCD, Genoa has... 12.
Ampere on the other hand... is competative, and Qualcomm with Nuvia cores will be as well.

https://www.servethehome.com/wp-content/uploads/2022/11/AMD-EPYC-9004-Genoa-c-ray-8K-rendering-Performance.jpg

Processors - Ampere Altra and AmpereOne

Cloud Native Processors reduce data center power consumption and physical space requirements, while delivering unprecedented performance and cost savings.

amperecomputing.com

 

[usiname]

der8bauer had posted a video showing overclocked 56core Xeon W beating TR in Geekbench while sucking 650W+ of power from wall. Just like that Ice lake SP, these new CPUs too little too late while barely managing to catch up with last Gen from competition.

If you talk for the post in videocardz, overclocked intel eat 1100W and is actually faster only in geektrash, in real world benchmarks it is around 10% faster than overclocked TR with double power (1100w vs ~600w for TR). This is not win for me. Do you remember how the intel fanboyzz dreamed for 8+ golden cove cores cpu that will be efficient beast, here we are

 

[Patriot]

If you talk for the post in videocardz, overclocked intel eat 1100W and is actually faster only in geektrash, in real world benchmarks it is around 10% faster than overclocked TR with double power (1100w vs ~600w for TR). This is not win for me. Do you remember how the intel fanboyzz dreamed for 8+ golden cove cores cpu that will be efficient beast, here we are

In memory sensitive things I expect SR workstations to pull out some wins, also PCIE Gen5 will give it a storage boon on the workstation side of things.
I expect for audio work in particular SR will do quite well. Intel's main advantage is also it's main disadvantage, owning its own fabs. It is lagging behind on efficiency, but it can sure make a lot of them.
It is very clearly not the top dog in rendering though.

 

[usiname]

In memory sensitive things I expect SR workstations to pull out some wins, also PCIE Gen5 will give it a storage boon on the workstation side of things.
I expect for audio work in particular SR will do quite well. Intel's main advantage is also it's main disadvantage, owning its own fabs. It is lagging behind on efficiency, but it can sure make a lot of them.
It is very clearly not the top dog in rendering though.

Almost always slower and always with double power and later, its shame. TR4 will be on different level. The next gen xeon is at most raptor lake, literally same IPC with same power consuption for multi core and only faster in single core. Granite Rapids is planned for the end of 2024, but this is Intel and they already delayed their next gen desktop so in the best intel will have answer to TR4 in the end of the 2025, 2 years later when AMD will release their next gen TR

 

[Vya Domus]

That's only horrible if you're somehow assuming that an E-core should somehow be equally powerful as a P-core.

I am not assuming anything, e-cores have objectively trash performance.

 

[BoboOOZ]

der8bauer had posted a video showing overclocked 56core Xeon W beating TR in Geekbench while sucking 650W+ of power from wall. Just like that Ice lake SP, these new CPUs too little too late while barely managing to catch up with last Gen from competition.

Too little too late was what was expected for them anyway, unfortunately.

To be fair to Intel, maybe they can still compete on price and availability. It won't be good for their already slim margins, but it will be good for the users.

 

[Wirko]

I am not assuming anything, e-cores have objectively trash performance.

Trash or not, Intel and AMD are developing their E and 4c cores because they fear each other a bit, and because they fear companies selling small Arm cores a lot.

 

[Avlin]

the issue here is e core purpose to be used as efficient core but instead used to save silicone. then e core are clocked too much and you then loose the efficiency…. arm core are efficient because of process + low clocks.

 

[Vya Domus]

Trash or not, Intel and AMD are developing their E and 4c cores because they fear each other a bit, and because they fear companies selling small Arm cores a lot.

Regardless they have no place in anything HEDT related, e-cores are designed for several purposes and performance is not one of them, clearly.

 

[Minus Infinity]

Trash or not, Intel and AMD are developing their E and 4c cores because they fear each other a bit, and because they fear companies selling small Arm cores a lot.

To be fair, AMD's E core's are going to be much stronger, 4c cores are the same core as Zen 4, minus some cache and clocks and said to be only 10-30% weaker. Genoa might see a 128 Zen 4c core model and Turin gets a 256 core version with 5c cores.

 

[Dirt Chip]

A 13900K and 7950X score exactly the same in CB23, one has 24 cores (8+16) and the other 16, meaning an e-core is quite literally half of a Zen4 core performance wise. Seems pretty horribly underpowered to me.

I think the 13900k has some advantage in cb23. Anyway, both have 32 threds and in the end score the same.
So the e cores are weaker but come at much higher numbers. 4e are faster than 1p so I speculate that 160 threds of mix p+e will match and go above the 128 of top TR.
But if AVX and high IO is needed, nix the e cores. No questions ask.

 

[ypsylon]

Wow. That's Bad. Like really with capital B bad considering the platform cost.

Cinebench (Cinema4D) was always done for Intel (be it Mac or PC) and newest and greatest has nothing to show there. Zen3 TRP stomps SR-X, like its not even funny. (single core is only for benchmark junkies)

Unless you have workflow which will specifically benefits from Intel architecture or firmware updates do some extra magic internally, this release is DOA. No matter how Intel PR will try to spin it or rename it. SR-X can't go toe to toe with old Zen3 and it sucks 50% more power to be in 2nd place. Like whaaat? I was hyped, but I'm fully cured already. Worst part is that AMD may completely ignore new TRP now.

 

[Vya Domus]

I think the 13900k has some advantage in cb23. Anyway, both have 32 threds and in the end score the same.
So the e cores are weaker but come at much higher numbers. 4e are faster than 1p so I speculate that 160 threds of mix p+e will match and go above the 128 of top TR.
But if AVX and high IO is needed, nix the e cores. No questions ask.

Number of threads do not matter for things like rendering. You can't just increase the number of cores thinking that using more lower performance cores is the same as having less higher performance cores, that's not how it works, applications do not scale perfectly. 32 e-cores would categorially be slower than 16 p-cores for example, there is no doubt about it.

 

[Dirt Chip]

Number of threads do not matter for things like rendering. You can't just increase the number of cores thinking that using more lower performance cores is the same as having less higher performance cores, that's not how it works, applications do not scale perfectly. 32 e-cores would categorially be slower than 16 p-cores for example, there is no doubt about it.

I agree, but 64e may beat 16p. 1p take the same dia space of 4e, give or take. And remember, you have base p cores anywayh so you don't really on e cores exclusively.

Maybe some day we can have 56p vs 32+96e.

 

[bl4C3y3]

Number of threads do not matter for things like rendering.

I agree, but 64e may beat 16p. 1p take the same dia space of 4e, give or take.

Wasn't this the reason why they added e-cores ?
To have have better multi-threading performance ... p-cores are faster then e-cores, but for the same die area, e-cores add more multi-threading performance (at least I think that's how it was).

When when only using p-cores, the same die area would have resulted in a 12c/24t CPU ... they 'sacrificed' 4 p-cores to add 16 e-cores, resulting in a 24c(8p+16e)/32t(16p+16e) CPU.
So the assumption is that 12p cores have lower multi-threading performance then 8p+16e.

 

[TTTT]

all the test done here are not really for professionals are they? Cinabench.. video editing ..looks like the benchmark is made for YouTubers
Would be interesting to see some benchmarks in software like abaqus explicit or LSdyna or other software where the large cost is the licence per core…. I think per core performance is in advantage of intel is it not ?