E5-2697v2 Performance

[No_one]

Hi,

i've recently ugraded my DL380p Gen8 (128 GB DDR3, 1600MHz), 1 GB SAS-Controller) from Dual E5-2670 to Dual E5-2697v2, but the Linpack (Ubuntu) results do not impress me very much. Leaving Linpack Parameters untouched, i get as a maximum 250 GFlops (HT off, for both CPUs!), whereas my Haswell I5-4570 (16GB, 1600MHz) with 4C alone does 150 GFlops. Means, one E5 with 12C is weaker than the I5-4C .

NUMA doesnt affect this result very much and i know that the Xeons do only offer AVX, whereas the 4570 supports AVX2. But could it be that 12 Cores deliver way less performance than a 4 Core i5, only because of the iteration from Ivy to Haswell - cant believe that. I know that AVX2 adds Performance, but AVX also does "some" Job. And yes, i use the most recent MKL-Linpack.

All cores are fully loaded during measurement.

Any numbers for comparison would be very helpful - thank you !

Best Regards.

[NormanS_Intel]

Hello No_one,

Thank you for posting in the community!

To ensure you receive the most specialized assistance, we have a dedicated forum that addresses these specific concerns. Therefore, I will be moving this discussion to our Server Forum. This will allow our knowledgeable community and experts to provide you with timely and accurate solutions.

Best regards,

Norman S.

Intel Customer Support Engineer

[Fikri_Intel]

Hi No_one,

Thank you for contacting Intel.

We appreciate you reaching out to us regarding the reported issue. Upon checking, Intel® Xeon® Processor E5-2697 v2 and the associated spare parts and accessories have been End of Interactive Support (EOIS). This email is to inform you that the support for the said products has been switched to self-service support.

https://www.intel.com/content/www/us/en/support/articles/000022396/processors.html

The available self-service support options include:

# Technical support content available online

# Discontinued products could be under warranty. Submit warranty service through the Warranty Center (supporttickets.intel.com).

# Support and services for the Intel® Partner Alliance partners through the Intel® Partner Alliance support site (https://www.intel.com/content/www/us/en/secure/partner-alliance/overview.html)

# Help from other users in the Intel Support Community (https://community.intel.com/)

We no longer respond to telephone, chat, community support forums, or email inquiries for self-service products.

For CPU comparison you can refer to the below link:

https://www.intel.com/content/www/us/en/support/articles/000027710/processors.html

Since there is no further assistance we can provide for this, we will proceed with the case closure at our end. 

It is our pleasure to continue to serve you with the next generation of Intel innovation at Intel.com. Thank you for using Intel products and services.

Regards,

Fikri O.

[No_one]

Well, the issues seems to be related to LinPack, since older Versions seem to use FP32-Arithmetics, while the newer Version use FP64, which isnt precisely the strongest point of those Processors. Using the 2022 Linpack Version, i get around 460 GFLOPS (still not the best performance for two XEONs; thats a number i would rather expect from one), but its more than 250 GFLOPS for both.

LLMs, which i use on them in combination with six external grpahics cards, rather need at max (very seldom) FP32, where FP64 does not count at all. I would rather tend to say, that it would be better to give out a Linpack Version which offers Benchmarks down to INT8 for better comparison.