Leaked 12-core Ice Lake server ES nearly 100 percent faster than Cascade Lake in Geekbench, takes the fight to EPYC Milan

[Redaktion]

A leaked Geekbench listing for Intel's upcoming Ice Lake server parts shows massive performance gains. The 12-core Ice Lake server ES delivers nearly the same multithreaded performance as a 24-core Cascade Lake part.

https://www.notebookcheck.net/Leaked-12-core-Ice-Lake-server-ES-nearly-100-percent-faster-than-Cascade-Lake-in-Geekbench-takes-the-fight-to-EPYC-Milan.453582.0.html

[william blake]

ice lake cores are less energy efficient than zen 2 cores.
i repeat, furure server ice lake is less energy efficient than current server zen 2.
end of story.
on a free market, like put your thing on a shelf, set a price, and call a customer-intel would be dead.

[Kyle_v]

Just for a quick comparision, my 12-core 3900x ( which is admittedly a consumer and not a server part ) scores 48,000 to 50,000 on Geekbench 4.3 for the multi-core test.

This score at 27k, even for a 12-core efficiency-focused / engineering part doesn't seem that good at all. Interesting to see the final results after release.

[JD_2020]

Xeon Gold 6226

That's a 12 core Xeon, not 24 core

[Arjun Krishna Lal]

Xeon Gold 6226

That's a 12 core Xeon, not 24 core

Both of these are dual-socket systems. The Ice Lake ES has 2x6 cores while the Xeon Gold 6226 system haS 2x12 cores.

[Hifihedgehog]

Not buying it. It has the telltale signs of being spoofed. The Pentium II/III entry and the others are fishy.

[Hifihedgehog]

Just for a quick comparision, my 12-core 3900x ( which is admittedly a consumer and not a server part ) scores 48,000 to 50,000 on Geekbench 4.3 for the multi-core test.

This score at 27k, even for a 12-core efficiency-focused / engineering part doesn't seem that good at all. Interesting to see the final results after release.

Nail. Head. This looks worse than first gen EPYC, let alone Rome or Milan. Besides, InstaLatX64 just grabs new results he scraps from Geekbench's database. Plus, it is trivial to fake a result on Geekbench.

[m53]

Just for a quick comparision, my 12-core 3900x ( which is admittedly a consumer and not a server part ) scores 48,000 to 50,000 on Geekbench 4.3 for the multi-core test.

This score at 27k, even for a 12-core efficiency-focused / engineering part doesn't seem that good at all. Interesting to see the final results after release.

This is Geekbench 5, not 4.3. Ryzen 3900X usually scores around 12K in Geekbench 5 while this CPU is scoring 27K for the same core count. That's impressive especially for an Engineering Sample.

[Phataliity]

This score at 27k, even for a 12-core efficiency-focused / engineering part doesn't seem that good at all. Interesting to see the final results after release.

This is Geekbench 5, not 4.3. Ryzen 3900X usually scores around 12K in Geekbench 5 while this CPU is scoring 27K for the same core count. That's impressive especially for an Engineering Sample.

No, this is Geekbench 4.3.3. Look here.
cdn.wccftech[dot]com/wp-content/uploads/2020/02/Intel-Ice-Lake-benchmarks-892x1030.jpg

Here's a comparison too.
browser.geekbench[dot]com/v4/cpu/13516988 - 4.2.3 (4239/30480)
browser.geekbench[dot]com/v4/cpu/14099402 - 4.3.3 (4708/37417)

4708 / 37417 - 3.2/4.2 clock speeds
3427/27926 2.2/2.6 clock speeds
5555 18% IPC improvement (4708 * 1.18) @ 4.2
61% max clocks (5555 * 0.61), 3388.81

It's the same exact Ice Lake. 18% IPC. It's not 200% faster. At the clocks we currently see, its actuallly slower.

[Phataliity]

Here is the actual geekbench listing. 4.3.3

browser.geekbench[dot]com/v4/cpu/15119835