That's the most important.
Because old ver. MCU? Full clocks 3.8GHz is crazy, but that's may not possible on supermicro.
ES Xeon Discussion
- Thread starter britinpdx
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
Core ratio Auto to -Core ratio by core usage mod,dont work with these processors?
this is the ES, the accelerator and the UPI is different from prod. unit, we never confirmed Q03J could work normally 4 or 8 way. And don't expect E0's accelerator will work good. it's just a 60 core E0 CPU, far from 8490H. also Q0KG/Q0KL is just a 52core E0 cpu but can work with all C741 board.
Not all cores 3.8. But before no one core went over 3.0GHz, just never, not even with single thread benchmark. In HWinfor history max clockspeed history was 3.0GHz, never anything over it. Tried many bios settings, did not get it to work. But with other bios 1.02, and cmos reset, i get it working normally, but of course only 3.8GHz if other cores not active. Still have to try if disabling half of cores gives 3.8 on 26 cores. Maybe tonight or tomorrow.
After i got it working normally Q0KG is quite good, 65K CB R23. And also silent noctual 140mm on 1000 rpm. Still need to make some better cooling for the vrm heatsink and ram. Memory gets very hot under load.
On Asus w790 i am quite confident you can change BCLK for that cpu. But how much depends on the silicon lottery, and if you want to run it with a gpu or not. On the forum heared of factors 102-104. In my opinion not really worth it. Especially since 13 and 14 gen intels (with the same/similar cores) are in the news for getting unstable when high PL settings are used. Now Mainboard manufacturers should offer intel profile, with nominal (lower) thermal settings, reducing performance. See for example the link here or here.
yes on ASUS W790, but your GPU is the funblocker. some don't work above 103mhz.
i got 109mhz with ASUS W790 Ace & GT630 and QYFQ (TDP270W - turbo 3300mhz)
but you will not get 109 with a CPU 350W - and turbo 4000mhz
you can not realy compare this with SPR. above 5ghz you can think about.
gen 13,14 8P-cores 253Watts
vs
SPR 48 P-cores 350W
I dont like the Z790 platform.
-400W with chilled,but 100 celsius.
I am go to W790.
Thank you everything!
Some testing with Q0KG 8470 E0, on Asrock Rack SP2C741D16X
I know these are not the best memory test. More will follow later one day.
These were the best settings that i could find. As Million Miles Mentioned near 250GB/s. But i did not quite manage to get there. This is 1 cpu 8x32GB 2Rx8 memory.
However some ebay sellers on Gigabyte MS73 Did better (more than double) see the screenshot below. I dont know if it is the mainboard difference. Or maybe Single Rank vs Dual Rank difference. Another factor could be that this was on windows 11 Home (to go) that only supports 128gb, so i dont know in what way they block the other 128gb.
I know these are not the best memory test. More will follow later one day.
These were the best settings that i could find. As Million Miles Mentioned near 250GB/s. But i did not quite manage to get there. This is 1 cpu 8x32GB 2Rx8 memory.
However some ebay sellers on Gigabyte MS73 Did better (more than double) see the screenshot below. I dont know if it is the mainboard difference. Or maybe Single Rank vs Dual Rank difference. Another factor could be that this was on windows 11 Home (to go) that only supports 128gb, so i dont know in what way they block the other 128gb.
Last edited:
dual CPU should be double the bandwidth thought.
Yes, i understand but they got more than double 550GB/s read vs my 234GB/s read.
I am not complaining, it was just interesting for me. with Q03J i got 270GB/s read, on Asus w790 sage. Same memory, also on 4800MT. 234-270GB/s still a bit of a gap.
Interestingly enough dual Q0KG is reported to have around 90K CB R23, while a single one already does 66K.
AIDA64 is a guess meter. multithread bandwidth report.
you can have lower band but more power. only this value form AIDA64 says NOTHING,
there a loss in dual socket config. cbr23 can not utilize good over sockets, cbr24 is better.
and there is windows.....
and why do you think AMD EPYCs have so much L3 ??? they will have much lower scores with less.
OK, but it is the only comparison i have against another Q0KG. The screenshot of the seller.
I also included the passmark score. In the next week i will try to do this stream memory test on Linux that Civiloid was speaking about.
This was also a good point.
if you want to compare your Q0KG with them you need same board/RAM/AIDA64(wlicense)
Probably best would be to visit them, and plug my Q0KG in their board, keeping everything else the same. But seriously i just wanted to compare my Q0KG setup against anything that also had a Q0KG to see if the performance is about right. Yes, true. maybe also the AIDA64 license part makes a difference.