Article Intel 10nm ‘Ice Lake SP’ 14 Core Server CPU Gets First Selfie On SiSoft Sandra, Over 54% IPC Improvement

[happy medium]

Intel 10nm 'Ice Lake SP' 14 Core Server CPU Gets First Selfie On SiSoft Sandra, Over 54% IPC Improvement

Intel's 10nm Ice Lake server CPU has been benchmarked for the first time on Sisfot Sandra and features an insane amount of IPC improvement!

wccftech.com

Will the desktop cpu's be similar?

 

[Markfw]

2 ghz base clock ? thats like Romes 64 core server. As all suspected, clocks are lower. The question is, what will it do in production vs a comparable AMD part ? 54% IPC improvement ? when I see production benchmarks against its competitor, I will believe it. Until then 2 ghz is like AMD is doing with their 64 core for base. I mean for their ES chips its base. I have a 64 core Rome ES, and its 2 ghz base.

Oh, and this is 14 core ? Rome 12 core is 2.9 ghz base, and Rome 16 core is 2.8 base. 2 ghz is so low that even if its reported 54% is true, how do they compare to Rome ? And what about 16,32,48, 64 core ? Do they have anything to compete with Rome ?

 

[vstar]

2 ghz base clock ? thats like Romes 64 core server. As all suspected, clocks are lower. The question is, what will it do in production vs a comparable AMD part ? 54% IPC improvement ? when I see production benchmarks against its competitor, I will believe it. Until then 2 ghz is like AMD is doing.

My impression was that this is an ES, so the clocks on the retail samples should be somewhat higher. In any case, we are a long ways off from finding out (unfortunately).

 

[lobz]

Intel 10nm 'Ice Lake SP' 14 Core Server CPU Gets First Selfie On SiSoft Sandra, Over 54% IPC Improvement

Intel's 10nm Ice Lake server CPU has been benchmarked for the first time on Sisfot Sandra and features an insane amount of IPC improvement!

wccftech.com

Will the desktop cpu's be similar?

Wccftech Just 2 days ago they wrote 100% core to core improvement about the 12 core ICL... =}}}}
Usman also praised Jim Keller for Ice Lake. Best website ever.

 

[Markfw]

Wccftech Just 2 days ago they wrote 100% core to core improvement about the 12 core ICL... =}}}}
Usman also praised Jim Keller for Ice Lake. Best website ever.

But wait, even if it were true, they suck.... See above.

 

[Adonisds]

2 ghz base clock ? thats like Romes 64 core server. As all suspected, clocks are lower. The question is, what will it do in production vs a comparable AMD part ? 54% IPC improvement ? when I see production benchmarks against its competitor, I will believe it. Until then 2 ghz is like AMD is doing with their 64 core for base. I mean for their ES chips its base. I have a 64 core Rome ES, and its 2 ghz base.

Oh, and this is 14 core ? Rome 12 core is 2.9 ghz base, and Rome 16 core is 2.8 base. 2 ghz is so low that even if its reported 54% is true, how do they compare to Rome ? And what about 16,32,48, 64 core ? Do they have anything to compete with Rome ?

Reading the article you see that it's 54% faster in a very specific thing that is totally not representative of real usage.
WCCF is fun and a good source of news, but they are huge clickbaiters, so you have to read the article and parse what makes sense

 

[Markfw]

Reading the article you see that it's 54% faster in a very specific thing that is totally not representative of real usage.
WCCF is fun and a good source of news, but they are huge clickbaiters, so you have to read the article and parse what makes sense

As I said, the new server Ice Lake SP either sucks or its not real. Either way, its a worthless article.

 

[Hitman928]

Reading the article you see that it's 54% faster in a very specific thing that is totally not representative of real usage.
WCCF is fun and a good source of news, but they are huge clickbaiters, so you have to read the article and parse what makes sense

Not just that, but they are assuming that the CPU is actually running at the reported 2 GHz. Not only is Sisoftware notoriously bad at reading frequencies (esp. for ES chips) but even if that were the actual base frequency, they are assuming that the chip doesn't have an all core turbo whatsoever. As an example, if the Ice Lake cores had an all core turbo of 3 GHz, you're looking at a 2% "IPC" increase in their comparison.

I honestly don't know why WCCFtech keeps doing this. I mean, it's not like we don't have actual retail Ice Lake products to test for comparison purposes. Any of those reviews showing a 54% increase in IPC?

 

[DrMrLordX]

Will the desktop cpu's be similar?

We already know from Icelake-U that 54% isn't happening in a non-mesh configuration. The 54% from the WCCFTech leak only obtains if you believe that the chip was stuck @ 2 GHz (I don't).

 

[amrnuke]

Intel 10nm 'Ice Lake SP' 14 Core Server CPU Gets First Selfie On SiSoft Sandra, Over 54% IPC Improvement

Intel's 10nm Ice Lake server CPU has been benchmarked for the first time on Sisfot Sandra and features an insane amount of IPC improvement!

wccftech.com

Will the desktop cpu's be similar?

Of course that depends on your definition of similar. I guess if they lock Ice Lake at 2GHz, sure. If they plan to actually be competitive with overall instructions per second, then no.

 

[happy medium]

We already know from Icelake-U that 54% isn't happening in a non-mesh configuration. The 54% from the WCCFTech leak only obtains if you believe that the chip was stuck @ 2 GHz (I don't).

Makes sense.
Thanks

 

[Kenmitch]

Wake me up when reality sets in. Another one of those " The leak meets my agenda " threads otherwise it's wccftech is a rumor mill that is seldom correct.

 

[nicalandia]

Ice Lake SP will be dealing with Zen 3 Milan servers not with Rome

 

[uzzi38]

Intel 10nm 'Ice Lake SP' 14 Core Server CPU Gets First Selfie On SiSoft Sandra, Over 54% IPC Improvement

Intel's 10nm Ice Lake server CPU has been benchmarked for the first time on Sisfot Sandra and features an insane amount of IPC improvement!

wccftech.com

Will the desktop cpu's be similar?

Don't you just love it when outlets assume that benchmarks of engineering samples must be reporting the correct clocks?

ICL-U should show 54% IPC uplift isn't happening, geez.

 

[killster1]

2 ghz base clock ? thats like Romes 64 core server. As all suspected, clocks are lower. The question is, what will it do in production vs a comparable AMD part ? 54% IPC improvement ? when I see production benchmarks against its competitor, I will believe it. Until then 2 ghz is like AMD is doing with their 64 core for base. I mean for their ES chips its base. I have a 64 core Rome ES, and its 2 ghz base.

Oh, and this is 14 core ? Rome 12 core is 2.9 ghz base, and Rome 16 core is 2.8 base. 2 ghz is so low that even if its reported 54% is true, how do they compare to Rome ? And what about 16,32,48, 64 core ? Do they have anything to compete with Rome ?

feel like you are taking it personal, they only mention amd in the last sentence saying they will have competition, im 100% sure we need competition for progress to continue, amd is the reason intel never bothered to do anything better since they sucked for so long. no need to improve, so better chips from both sides is a win for everyone. Im excited for intel to do some moves up even tho im running new amd chips in my recent desktops but not my recent laptops (never purchased a amd lappy but excited still for ryzen 4000)

 

[Markfw]

feel like you are taking it personal, they only mention amd in the last sentence saying they will have competition, im 100% sure we need competition for progress to continue, amd is the reason intel never bothered to do anything better since they sucked for so long. no need to improve, so better chips from both sides is a win for everyone. Im excited for intel to do some moves up even tho im running new amd chips in my recent desktops but not my recent laptops (never purchased a amd lappy but excited still for ryzen 4000)

Personal ? No... Its just that I am sick of Intel STILL being in the dust. If they don't come out with something soon that can compete with 3000 series AMD chips, including Rome. If they don't get it together soon, AMD may start slacking off.

 

[DrMrLordX]

The other thing that concerns me about these Ice Lake-SP samples is that I haven't yet seen one with more than 14c. Unless I'm missing a leak somewhere. There's supposed to be 38c chips, and those chips should be shipping in QS or early commercial silicon to select customers. Going as far back as at least Skylake-SP, Intel has had silicon ready for limited ODM shipments 6 months (or more) before official launch to the general market. If Skylake-SP is supposed to be ready this year, you would think QS of a significant portion of the lineup would be available by now.

 

[ksec]

Can we please please please stop posting anything from Wccftech, it should have no place on Anandtech.

 

[vstar]

The other thing that concerns me about these Ice Lake-SP samples is that I haven't yet seen one with more than 14c. Unless I'm missing a leak somewhere. There's supposed to be 38c chips, and those chips should be shipping in QS or early commercial silicon to select customers. Going as far back as at least Skylake-SP, Intel has had silicon ready for limited ODM shipments 6 months (or more) before official launch to the general market. If Skylake-SP is supposed to be ready this year, you would think QS of a significant portion of the lineup would be available by now.

I think the XCC Icelake-SP die with 38C is supposed to come out in Q1 2021, with the HCC die coming out in Q4 2020, so IMO it makes sense why we haven't seen leaks for the higher core count chips.

 

[zinfamous]

Can we please please please stop posting anything from Wccftech, it should have no place on Anandtech.

...at least require that "wccftech" be mentioned in every thread title that sources them.

I see that the thread title still uses the highly dubious IPC improvement claims based on wccftech's likely assumption that the sample was locked at 2ghz, for no explicable reason.

 

[DrMrLordX]

I think the XCC Icelake-SP die with 38C is supposed to come out in Q1 2021, with the HCC die coming out in Q4 2020, so IMO it makes sense why we haven't seen leaks for the higher core count chips.

Okay, that makes sense. Also ouch on having to wait a whole quarter just to get the 38c die out.