32 Core AMD Zen "Rome" Processor Leaked on SiSoft

[fangxp]

Here's the link:

Details for Result ID AMD Eng Sample: ZS1711E3VIVG5_24/17_N (32C 64T 2.4GHz/1.7GHz 40% OC, 733MHz IMC, 32x 512kB L2, 8x 16MB L3)

ranker.sisoftware.co.uk

Here's the decoding chart:

Here's an explanation:

32 Core/64 Thread AMD Zen 2 Epyc "Rome" Processor Leaked on SiSoftware

Twitter user, momomo_us has shared the SiSoft benchmark of a Zen 2 based Epyc "Rome" processor, packing 32 cores and 64 threads, running at a base clock of 1.7GHz and a boost of 2.4GHz.

www.techquila.co.in

Cheers

 

[er557]

 

[FordGT90Concept]

Has a really low clockspeed…
Z = QS
S = Server
171 = 1.71 GHz base clock
1 = Standard SKU
E3 = Unkown TDP & VRM
V = Unkown Physical Package
I = 32 cores
V = Unknown Cache Configuration
G5 = Rome
24 = 2.4 GHz boost clock
17 = 1.7 GHz base clock
N = ???

 

[NdMk2o1o]

Has a really low clockspeed…
Z = QS
S = Server
171 = 1.71 GHz base clock
1 = Standard SKU
E3 = Unkown TDP & VRM
V = Unkown Physical Package
I = 32 cores
V = Unknown Cache Configuration
G5 = Rome
24 = 2.4 GHz boost clock
17 = 1.7 GHz base clock
N = ???

Probably to do with the QS? not a production/final release sample? but it is Epyc so expect lower clock speeds than Ryzen as with previous Epyc chips.

 

[Bones]

View attachment 123225

What are the specs for the system ran with this for comparison?
That was given with the AMD run.

What version of SiSandra was used for this test?

Just results from someone's test run, unless actual details of BOTH setups are given it doesn't mean much concerning comparison between the two because there really isn't much to go on for such here. Also bear in mind the AMD chip is still under development and the results/specs of the chip could change with the final release version of the chip.

 

[Shatun_Bear]

These low clocks line up with previous leaks of ROME. I wonder whether putting 32-cores on a 7nm chip is causing heat problems so the clocks have to be low.

 

[er557]

What are the specs for the system ran with this for comparison?
That was given with the AMD run.

What version of SiSandra was used for this test?

Just results from someone's test run, unless actual details of BOTH setups are given it doesn't mean much concerning comparison between the two because there really isn't much to go on for such here. Also bear in mind the AMD chip is still under development and the results/specs of the chip could change with the final release version of the chip.

these are the specs, the test was run with an asrock rack ep2c612-ws motherboard, unofficially boosted xeons(all 72 threads@3.5ghz), software: sisoft sandra business edition, processor multi media benchmark

 

[NdMk2o1o]

these are the specs, the test was run with an asrock rack ep2c612-ws motherboard, unofficially boosted xeons(all 72 threads@3.5ghz), software: sisoft sandra business edition, processor multi media benchmark

View attachment 123230

So your'e saying that the AMD Epyc QS/ES W/E it is @ 32c/64t whilst being 1.1ghz lower (boost 2.4ghz) clocked came close to your Xeon 36c/72t @3.5ghz? or am I reading it wrong, cause that's not too bad IMH and I'm sure this is something like a 120/150w part taking into consideration previous Epyc power consumption.

 

[Divide Overflow]

Early, low frequency engineering sample.

 

[er557]

I am not sure how to interpret the exact results, as the ranking system is a bit different in regard to avx/fma extensions being used, but we need to keep in mind that AMD rome is 2019 architecture, while xeon e5 v3 is from 2015, albeit has an efi boost driver.

 

[FordGT90Concept]

These low clocks line up with previous leaks of ROME. I wonder whether putting 32-cores on a 7nm chip is causing heat problems so the clocks have to be low.

I think they're aimed at driving thin clients so you need lots of threads but each client doesn't require much performance. They are keep the clocks low to keep power draw to a minimum.

These processors wouldn't end up in high performance compute scenarios, for sure.