Share your AIDA 64 cache and memory benchmark here

[Marcoz]

Slightly better without some background programs I managed to get 66Gb/s Read, 67Gb/s Write, almost 61Gb/s Copy and 48.7ns. Do you think I can improve further and do you think it's a good result?

 

[damric]

Slightly better without some background programs I managed to get 66Gb/s Read, 67Gb/s Write, almost 61Gb/s Copy and 48.7ns. Do you think I can improve further and do you think it's a good result?

You can probably go higher with loosened timings. Cezanne is very interesting, not only because it is a unified 7nm SOC, but because when AMD designed them for laptops they came with advertised DDR4L-4266 support out of the box. I doubt anyone has actually ever seen such a configuration on a laptop, but supposedly it was purposely built for such.

 

[bersy]

Slightly better without some background programs I managed to get 66Gb/s Read, 67Gb/s Write, almost 61Gb/s Copy and 48.7ns. Do you think I can improve further and do you think it's a good result?

It's great.
It's required to set GDM disable to get evel lower latency. Another option is to rise MHz, take a look at Reous 4400CL16 GDM https://abload.de/image.php?img=440016-16-16-1610000kigkmb.png

 

[Sitkam]

My c-die sticks. 25 cycles in TM5 1usmus v3 and y-cruncher 4 interactions passed without errors.

 

[bersy]

My c-die sticks. 25 cycles in TM5 1usmus v3 and y-cruncher 4 interactions passed without errors.

I believe you could try CL17 + CR1 mode and RFC 260-270 ns.

My OEM H8C SR was like this:

 

[Sitkam]

I believe you could try CL17 + CR1 mode and RFC 260-270 ns.

My OEM H8C SR was like this:

I didn't tight RFC yet, anyway this is c-die...so i cannot tight it too much like those b-die sticks . maybe a little bit. about primaries, i'll test C17 and run TM5 for errors. thanks a lot!

 

[bersy]

anyway this is c-die...

Ah, is it Samsung or Hynix C-Die?

I initially assumed it's Hynix C-Die (CJR, H8C).

 

[HD64G]

Finally my current profile is stable and 100% working at 1.50v

I think that the SOC Voltage is unsafe if you want to keep the CPU for years.

 

[Sitkam]

Ah, is it Samsung or Hynix C-Die?

I initially assumed it's Hynix C D-Die (CJR, H8C).

yes, it's Samsung C-Die

 

[bersy]

I think that the SOC Voltage is unsafe if you want to keep the CPU for years.

1.25V is OK for APUs.

yes, it's Samsung C-Die

My mistake. That's quite mediocre ICs and you config on the contrary looks good to me.

 

[Calmmo]

1.45v, Stable at sub 60c

 

[Canned Noodles]

My other results with this CPU featured an FCLK of 1866 but that proved to be unstable, so I am now testing with 1833 and it is much more stable. The latency is a tiny bit lower than before.

 

[Sitkam]

1.25V is OK for APUs.


My mistake. That's quite mediocre ICs and you config on the contrary looks good to me.

I'd say that b-die sticks, it's my dream

 

[Marcoz]

Hello everyone, I wanted to increase the mem clock of my 4333mhz cl16 ram, perhaps towards 4533mhz, do you think I should loosen the timing? I'm on 16-16-16-32 without mentioning the already extensively modified secondary timings I'm interested in reaching 70Gb/s since I'm on 66-67Gb/s any advice on how to act? Thanks in advance to everyone.

 

[bersy]

Hello everyone, I wanted to increase the mem clock of my 4333mhz cl16 ram, perhaps towards 4533mhz, do you think I should loosen the timing? I'm on 16-16-16-32 without mentioning the already extensively modified secondary timings I'm interested in reaching 70Gb/s since I'm on 66-67Gb/s any advice on how to act? Thanks in advance to everyone.

Hi,
4533 isn't optimal for APUs because of an effectiveness hole at all **33 MHz except 3333 and 4333 ones.

https://imgur.com/mnbEmiX

So your best bet is 4400 and 4600.

 

[Tatili]

Mine at the moment

View attachment 289419

Hi enthusiast there what kind of memory kit you have, did you check event viewer for Wheas19 and run TM5 stability test 25 cycles ?
I found tones of them at 4000 IF 2000 and I had to roll back my previous rock stable 3944Mhz (102 FSB included) to get rid of Wheas

 

[bersy]

I found tones of them at 4000 IF 2000 and I had to roll back my previous rock stable 3944Mhz (102 FSB included) to get rid of Wheas

Sure, 3733-3800 MHz is a common frequensy for ZEN2-3 non APU CPUs after whitch point WHEA usually occur.

There is also small amount of Ryzens with two CCDs that are capable to hold ~4000 MHz without any WHEA.

Hi enthusiast there what kind of memory kit you have, did you check event viewer for Wheas19 and run TM5 stability test 25 cycles ?

Cinebench R23 or LinX is a faster and more reliable way detecting WHEA errors afaik.

 

[Athlonite]

Hi enthusiast there what kind of memory kit you have, did you check event viewer for Wheas19 and run TM5 stability test 25 cycles ?
I found tones of them at 4000 IF 2000 and I had to roll back my previous rock stable 3944Mhz (102 FSB included) to get rid of Wheas

Funnily enough AIDA64 doesn't get WHEA errors but but the likes of CB23 or even watching YT videos will do so had to back mine off to 3800MTs/1900MHz IF, which is still plenty fast for me as the only AAA game of recent times I play is CP2077 which runs perfectly fine with those settings

 

[freeagent]

Mine atm..

 

[Marcoz]

I improved further reaching 4400Mhz C16 at 1.52v

 

[bersy]

I improved further reaching 4400Mhz C16 at 1.52v

Is it stable?

 

[Marcoz]

Is it stable?

I passed tm5, occt, aida stress test with cpu fpu and memory, other tests and games of different types from warzone 2 to battlefield 2042 in 128 player conquest or history games like Red dead 2 or Elden Ring or Dead Space Remake. So yes it is stable. ;D

 

[bersy]

I passed tm5, occt, aida stress test with cpu fpu and memory, other tests and games of different types from warzone 2 to battlefield 2042 in 128 player conquest or history games like Red dead 2 or Elden Ring or Dead Space Remake. So yes it is stable. ;D

Great, that's truly top bin ICs.

 

[Marcoz]

Great, that's truly top bin ICs.

It is true

 

[Robius]

My new 2x16GB stock Samsung B-Die rams (1.35v)