Epyc Rome 7F52 Upgrade from 7302P?

[TXAG26]

Current machine for Davinci Resolve and photoshop has a 7302P (16c; 3.0ghz/3.3ghz, 128MB L3, 155w) with 8x 32gb DDR4-3200 to populate all 8x memory channels on a Supermicro H11SSL v2.0 series board.

Recently saw that Rome 7F52 processors (16c; 3.5ghz/3.9ghz; 256MB L3; 240w) were going for about $300 and wondered if that would be a useful upgrade to wring some more life out of this board/ram for multimedia/content creation? Also any info on idle power consumption between the 7F52 and 7302P?

 

[erock]

My machine with a H11DSi mobo and two 7302’s runs idle at around 96W. My other machine with a H11DSi mobo and two 7f52‘s runs idle at around 114-135W. For my scientific calculations, which involve shared memory parallelization, I see around a 15-20% boost in performance with the 7f52 relative to the 7302. I also see around a 10% difference between the 32 core threadripper 3970x ($2000) and the 75f2x2 ($600). All memory slots are populated with 16GB sticks.

 

[TXAG26]

My machine with a H11DSi mobo and two 7302’s runs idle at around 96W. My other machine with a H11DSi mobo and two 7f52‘s runs idle at around 114-135W. For my scientific calculations, which involve shared memory parallelization, I see around a 15-20% boost in performance with the 7f52 relative to the 7302. I also see around a 10% difference between the 32 core threadripper 3970x ($2000) and the 75f2x2 ($600). All memory slots are populated with 16GB sticks.

Nice systems and thx for the numbers. I'm thinking a 15-20% boost would be enough to make the upgrade worthwhile and get more time out of this workstation.

 

[erock]

Nice systems and thx for the numbers. I'm thinking a 15-20% boost would be enough to make the upgrade worthwhile and get more time out of this workstation.

The actual performance improvement magnitude will depend on the specifics of your calculation but you can‘t go wrong with a higher clock rate and more cache as long as the higher TDP is worth it for you.

If you do go with the higher TDP 7f52 make sure you have plenty of airflow across the board before you test a moderate to heavy cpu load. I made the mistake of only installing the heatsink and heatsink fan and then running a stress-ng cpu test. This was not a good idea and led to weird spontaneous restarts. Only bench test to setup/check BIOS then get some good fans blowing across the board.

 

[erock]

…and at $300 the 7f52 has a very good performance over price ratio (so does the 7302 at $120).

 

[TXAG26]

The actual performance improvement magnitude will depend on the specifics of your calculation but you can‘t go wrong with a higher clock rate and more cache as long as the higher TDP is worth it for you.

If you do go with the higher TDP 7f52 make sure you have plenty of airflow across the board before you test a moderate to heavy cpu load. I made the mistake of only installing the heatsink and heatsink fan and then running a stress-ng cpu test. This was not a good idea and led to weird spontaneous restarts. Only bench test to setup/check BIOS then get some good fans blowing across the board.

Good tip. If I get the 7F52 it will go in a CSE-743 that has a 4x fan wall and with IPMI I can keep an eye on VRM temps and such.

 

[TXAG26]

I received a 7F52 Epyc CPU and the upgrade went well! In Passmark Memtest86 v10.6 I noticed that the cache speeds and memory bandwidths are noticeably higher/faster on the 7F52 vs the 7302P, which I was not expecting. I ran Memtest86 a couple of times with cold reboots on each config and these numbers stayed within the same range.

—————7302P : 7F53
CLK Spd: 3000 : 3500
L1 Cache 64K: 183.9 GB/s : 216.6 GB/s
L2 Cache 512K: 72.1 GB/s : 77.4 GB/s
L3 Cache 262,144K: 14.6 GB/s : 16.4 GB/s
Memory 256G: 14.5 GB/s : 16.4 GB/s
Ram Config: DDR4 ECC 2934MT/s 16x Channel / Samsung M393A4K40DB3-CWE

While my DDR4 ram is 3200, I’ve always run it at 2933 so that it is in-sync with the CPU’s Infinity Fabric speed to reduce system latency.
Nice little unexpected performance bump with the upgrade!