Discussion Intel current and future Lakes & Rapids thread

[DrMrLordX]

I kind of expected the performance. Power usage is higher than I expected.

I didn't. P-cores on desktop become far more power-efficient at lower clocks/voltages. Something is wrong with Sapphire Rapids. They should be getting better performance at that power (or lower power for that performance).

 

[A///]

Saying something is wrong with sapphire rapids is the understatement of the decade.

 

[semiman]

Stock voltage is very high, Raptor Lake's P-cores doesn't need 1.2v to hit 5 GHz, and can do 4 GHz at 0.95v or less. It'll be interesting to see what undervolting does, SkatterBencher's 24-core 2495X uses 1.15v at 4.4 GHz:

It simply points out that Intel had no choice but to use a higher voltage than what Raptor Lake used. We 'could' try undervolting, but we can't be sure that it won't break any critical functionality of the chip. Reliability, data integrity, CPU operation in certain high-demand situations...etc

 

[mderbarimdiger]

Intel's main server team, and I can't believe I'm saying this, actually seems to be in a decent enough state now, from what I've been hearing. If nothing else, far better than the GPU team, low bar though that may be.

I am glad to see this post from you. i think you missed to note Sandra Rivera's pervious role in managing network and edge, its Intel's best business segment (at least by growth/decline measures).

 

[nicalandia]

I didn't. P-cores on desktop become far more power-efficient at lower clocks/voltages. Something is wrong with Sapphire Rapids. They should be getting better performance at that power (or lower power for that performance).

I keep saying this, but nobody seems to get it. The very high idle power consumption points to this. The Intel Mesh Interconnect and the IO Design are too power hungry

 

[A///]

Arrow should be using a new interconnect no?

 

[Geddagod]

Arrow should be using a new interconnect no?

Don't know. Don't think so tbh. The new dual ringbus Intel created for Tiger Lake and beyond supposedly doubled? the bandwidth of the original ringbus found in comet lake chips which scaled up to 10 cores. Intel in the past had ringbus that went beyond even that IIRC. Even an 8+32 Arrow Lake with 16 CPU ringstops might be fine, but since new rumors about ARL claim it just goes up to 8+16, there's even less reason to move past Intel's ringbus.

 

[Edrick]

In Blender the 3995WX gets 1093 points, the W9 3495X gets 1078, so the 3995WX has 1.4% advantage

In V-Ray the 3995WX gets 48,300 points and the 3495X gets 50,800 so thats a 5% advantage that goes to the 3495X.

So comparing a 64 core system to a 56 core system, and the numbers are fairly close. I do not see that as a huge disappointment at all from a performance perspective. Power, on the other hand, is disappointing for SPR. I am not claiming SPR is a competitive product in 2023 (it may have been in 2020-21), but looking at what Intel did have before this launch in the HEDT market (Cascade Lake/Ice Lake), SBR is a well needed step in the right direction.

 

[Edrick]

Even if we put all these things into account, 3495X is still a huge disappointment. Their workstation counterpart, 5995WX also has 8 channels MC, and 128 PCIe lanes. 3495X was supposed to win 5995WX in legacy programs and roflstomp 5995WX in simulation and AI applications. But they're even losing ST performance!
SPR is obviously suffering from problems which don't exist in Alder lake and its descendants. If we look into Raptor lake vs Zen 4 comparison, then high core count + big cores should be much better than this.

I don't know where you read that the 3495X was supposed to best a 5995WX, but everything I read was that Intel was not expecting to pass AMD until Diamond Rapids in 2025/2026. SPR and EMR were going to try to compete with the 3995WX. No argument that the process node they are on is causing some crazy power numbers which I did not expect.

 

[poke01]

Saying something is wrong with sapphire rapids is the understatement of the decade.

It's the worse than Rocket Lake. but but its got lots of PCIe lanes. So??

The SPR is crap. But busniesses like to buy crap. So Intel will survive.

 

[igor_kavinski]

But busniesses like to buy crap.

QFT

 

[nicalandia]

This is the Fastest 32 Core Sapphire rapids based Xeon All core boost to 3.1 Ghz, 350W TDP.

2S 64C/128T gets 91,000 points in Cinebench R23, total power is using more than 700 Watts.

闲鱼宝贝详情

h5.m.goofish.com

 

[A///]

And people were losing their minds when Zen 5 Epycs were open to pulling 600 watts.

 

[Shmee]

Saying something is wrong with sapphire rapids is the understatement of the decade.

What is wrong with it? I for one am excited for a comeback to HEDT from both Intel and AMD.

 

[Markfw]

What is wrong with it? I for one am excited for a comeback to HEDT from both Intel and AMD.

its the same as 2 generations ago for dpeed, and sucks power like no tomorrow.

 

[Shmee]

Ah, that is too bad. Maybe too many P cores just adds up for too much power usage. I wonder if the sweet spot will be a 12 or 16 core unlocked CPU. I am interested in it, not only for HEDT features of memory and PCIe lanes, but also because 8 P cores doesn't seem like enough to me, and I don't care about E cores.

 

[Edrick]

This is the Fastest 32 Core Sapphire rapids based Xeon All core boost to 3.1 Ghz, 350W TDP.

2S 64C/128T gets 91,000 points in Cinebench R23, total power is using more than 700 Watts.

Were you expecting Intel to produce CPUs (on a bad node) that could reach 90,000+ Cinebench score while only magically pulling 300 watts?

 

[Hulk]

Were you expecting Intel to produce CPUs (on a bad node) that could reach 90,000+ Cinebench score while only magically pulling 300 watts?

It could be done but would require an unrealistic amount of silicon. My 13900K could do 372 CB points/Watt (at 45w). That would be 90,000 CB score at less than 250 Watts. But you'd need five 13900 dies to get it done.

 

[Markfw]

It could be done but would require an unrealistic amount of silicon. My 13900K could do 372 CB points/Watt (at 45w). That would be 90,000 CB score at less than 250 Watts. But you'd need five 13900 dies to get it done.

But the problem is design. Its a server CPU, and e-cores don't fit the design, and no avx-512 which this has (and a couple other things yours does not have) You could not just paste 5 13900k cores and call it a cpu, as it would be totally different in function.

 

[Edrick]

But the problem is design. Its a server CPU, and e-cores don't fit the design, and no avx-512 which this has (and a couple other things yours does not have) You could not just paste 5 13900k cores and call it a cpu, as it would be totally different in function.

Exactly. Server CPUs are not designed to have mobile level power requirements due to a lot of other die area needed for extensions like AVX512 and AMX. And to be honest, the people who buy these types of Xeons are not concerned about 300W vs. 500W.

 

[Joe NYC]

Exactly. Server CPUs are not designed to have mobile level power requirements due to a lot of other die area needed for extensions like AVX512 and AMX. And to be honest, the people who buy these types of Xeons are not concerned about 300W vs. 500W.

Hyperscalers are concerned about power consumption. Those located in Europe even more so.

 

[Markfw]

Exactly. Server CPUs are not designed to have mobile level power requirements due to a lot of other die area needed for extensions like AVX512 and AMX. And to be honest, the people who buy these types of Xeons are not concerned about 300W vs. 500W.

I would have to disagree. I personally know some of our people in the data centers (square MILES of data centers) and that difference in power is not the total picture. Then you have the UPS capability and the AC capability, so it really makes a difference in the power budget. And in this case, the performance is certainly not worth it, as it performs below the competition. The days of "nobody gets fired for buying Intel" are over.

 

[Hulk]

But the problem is design. Its a server CPU, and e-cores don't fit the design, and no avx-512 which this has (and a couple other things yours does not have) You could not just paste 5 13900k cores and call it a cpu, as it would be totally different in function.

Of course. Just posted it as an interesting test result. It's not really practical.

 

[itsmydamnation]

Exactly. Server CPUs are not designed to have mobile level power requirements due to a lot of other die area needed for extensions like AVX512 and AMX. And to be honest, the people who buy these types of Xeons are not concerned about 300W vs. 500W.

What a load of crap.
in a compute dense rack you can go 4 sockets an RU , people most certainly care about 88kw of CPU vs 52kw*. That is significate power and cooling works and that's just CPU budget .
Even if you go for something more sane like 2 P an RU 44kw vs 26kw is still stupid bad.

i buy lots of servers for NFV purpose i can about maximum number of cores (HT off) with maximum clock rate with scalar / narrow SIMD widths , biggest SIMD user will be TLS decrypt/ encrypt .

Most server workload will never care about AVX-512 or AMX. So unless your telling orgs not to buy intel unless they need AMX really because Zen4 does AVX-512 just fine.

* 48ru with 4ru of switching 44ru of scale out compute , 6 x2.5inch drives a server , 1/2 width 2P server.

 

[ashFTW]

AnandTech Forums: Technology, Hardware, Software, and Deals

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

www.anandtech.com

Lenovo ThinkStation PX Dual SPR and the Xeon W Duo P7 and P5 Launched

The Lenovo ThinkStation PX workstation houses dual 4th Gen Intel Xeon processors while the P7 uses a Xeon W-3400 and the P5 has a Xeon W-2400

www.servethehome.com

Sapphire Rapids based 3 workstations announced by Lenovo. Very tempted to get a quote for the PX, but I think I’ll wait for Granite Rapids.