Share your AIDA 64 cache and memory benchmark here

[md2003]

Here is an small update of this patriot cl19 b-die kit.

 

[FilipM]

Already done these A2s are no joke

It's really interesting to see how tRFC behaves in the presence of an iGPU. I can pass endless TM5 runs @ 133ns without moving VDIMM even an inch, but the GPU really doesn't like it. 130ns no POST. 133ns tests stable but crashes instantly opening a game. 135ns test stable, runs for a while until throwing a UNEXPECTED_KERNEL_MODE_TRAP. 138ns (290) works just fine though.

It was the same with the A0s, tRFC went all the way down to 140ns stable but only about 143-145ns was actually iGPU stable.

On the A0s I couldn't do tRRDL 4, tWTRL 8 or tWR 10 without needing more VDIMM, not yet sure if it's the same here. I'll probably play with tRTP next.

@FilipM that reminds me, the old GDM vs 2T debate seems to be rather interesting. 2T is *sometimes* faster in membench by half a second or so. But GDM consistently wins in AIDA read/write by about 500MB/s and copy by about 200MB/s. GDM might also be leading to better GPU performance, but I'll need to bench that.

View attachment 198959

If you have the time to do some cpu benchies and game runs, please share

 

[tabascosauz]

If you have the time to do some cpu benchies and game runs, please share

Unfortunately it's going to be hard to bench Genshin until I can set up RTSS properly, but in the meantime Valley works well.

AIDA GDM vs 2T:



Membench GDM vs 2T:



Valley GDM vs 2T:



So yeah, mostly placebo in gaming but definitely faster than the A0s at 4200 17-18-18:

 

[Cowboystrekk]

Unfortunately it's going to be hard to bench Genshin until I can set up RTSS properly, but in the meantime Valley works well.

AIDA GDM vs 2T:

View attachment 199077View attachment 199078

Membench GDM vs 2T:

View attachment 199138View attachment 199137

Valley GDM vs 2T:

View attachment 199081View attachment 199082

So yeah, mostly placebo in gaming but definitely faster than the A0s at 4200 17-18-18:

View attachment 199083

If I remember correctly gdm makes tCL, tCWL, tWR and tRTP even values. If one of them is odd then 2T would probably give better performance.

 

[dskhury]

That write speed i get is normal? This on an Asus X570 TUF

 

[Cowboystrekk]

That write speed i get is normal? This on an Asus X570 TUF
View attachment 200169

Yes, single chiplets like 5600X and 5800X get half speed vs 2 chiplets. Test dram calc to tweak your scores. 10ns lower latency with just ramtuning and a bit more speed should work

 

[tabascosauz]

Sometimes tWR 10 and tRTP 5 works, that gave me a slight boost without requiring more voltage. tRRDL 6 and tWTRL 10 or 8 might work aswell

Unfortunately anything below 4/6/16 - 4/12/12 requires more VDIMM to stabilize, and it's looking like 0.02V or more at 4/4/16 - 4/8/10. 4/4/16 - 4/8/10, 4/6/16 - 4/8/10, 4/4/16 - 4/10/10, it's all the same: needs more VDIMM. It was a difference of like 300MB/s copy perf, no latency or membench improvement.

I know a lot of ppl run 4/4/16 - 4/8/10 and I remember BZ saying in a more recent video that the tighter timings shouldn't need extra voltage, but I suppose that's not the case.

My A0 B-die, 4Gb E-die and CJR also can't tighten past 4/6/16 - 4/12/12 without more voltage. Was hoping the better B-die could buck that trend somehow, but oh well.

 

[Cowboystrekk]

Unfortunately anything below 4/6/16 - 4/12/12 requires more VDIMM to stabilize, and it's looking like 0.02V or more at 4/4/16 - 4/8/10. 4/4/16 - 4/8/10, 4/6/16 - 4/8/10, 4/4/16 - 4/10/10, it's all the same: needs more VDIMM. It was a difference of like 300MB/s copy perf, no latency or membench improvement.

I know a lot of ppl run 4/4/16 - 4/8/10 and I remember BZ saying in a more recent video that the tighter timings shouldn't need extra voltage, but I suppose that's not the case.

My A0 B-die, 4Gb E-die and CJR also can't tighten past 4/6/16 - 4/12/12 without more voltage. Was hoping the better B-die could buck that trend somehow, but oh well.

Okay, tried 12/6 or 10/5 tWR/tRTP? Those timings affect perf more.

 

[uno.di.voi]

patriot viper steel 4400cl19 -->4266 1.45 vcore

 

[glnn_23]

Running a different Ryzen 4650g here in an Asus ROG Impact.

 

[freeagent]

^^ Noice!

 

[Cowboystrekk]

^^ Noice!

View attachment 200290

Zentimings? Pbo? Safemode?

 

[freeagent]

Zentimings? Pbo? Safemode?

Whoops sorry!

No Pbo, no safe mode..

Next time I will post zentimings.. but really nothing special to see there..

 

[uno.di.voi]

patriot again with 11400..... ceers!

write >83000 is incorrect imho....

 

[freeagent]

Nothing crazy.. just DOCP settings minus the mains..

2T is still pretty quick

 

[uno.di.voi]

again with settings..... patriot 4400 @ 4266

 

[glnn_23]

 

[freeagent]

Jeebus was at my side

 

[uno.di.voi]

under 50 ns .... yep!

 

[tbob22]

Micron E Die dual rank (32gb 3200mhz Ballistix Sport), subtimings seem to need to be pretty loose. Still getting WHEA errors at 2000mhz IF though, tried all kinds of voltages up to 1.2v SOC, no luck.

1900/3800 seems all good though at SOC 1.125v and can set 1T with GD disabled, still have some optimizing to do though.

 

[Cowboystrekk]

Micron E Die dual rank (32gb 3200mhz Ballistix Sport), subtimings seem to need to be pretty loose. Still getting WHEA errors at 2000mhz IF though, tried all kinds of voltages up to 1.2v SOC, no luck.

1900/3800 seems all good though at SOC 1.125v and can set 1T with GD disabled, still have some optimizing to do though.

View attachment 201795View attachment 201793

I have done a lot of rev E tweaking on single rank that is. Post your zentimings and I can try to help you if you want?

under 50 ns .... yep!

View attachment 201444

Set tRAS to 32, tWR to 12 or 14, tFAW to 16 or 20 and try tRFC to 282 or 329. That should speed things up very much!

 

[tbob22]

I have done a lot of rev E tweaking on single rank that is. Post your zentimings and I can try to help you if you want?

Sure, I did have to loosen some timings but I was able to tighten others. I can't seem to get much lower than 600 on tRFC, it won't even post most of the time at 554. 15 on tRDCWR and tRP was causing major stability problems, as did tRAS at 32. My board makes it a pain because I have to manually reset the bios every time it fails to post and lose all my settings, it will recover maybe 1 out of 10 times.

 

[Cowboystrekk]

Sure, I did have to loosen some timings but I was able to tighten others. I can't seem to get much lower than 600 on tRFC, it won't even post most of the time at 554. 15 on tRDCWR and tRP was causing major stability problems, as did tRAS at 32. My board makes it a pain because I have to manually reset the bios every time it fails to post and lose all my settings, it will recover maybe 1 out of 10 times.

View attachment 201887

Okay. Safe suggestions: Try tRAS 34 (tCL+tRCDRD) and tRFC 580 or 590, tRRDS 5, tRRDL 7, tFAW 20, tWR 12, tRTP 8 or 6, tWTRS 4, tWTRL 12.

Fast: tRFC 570 or 560, tRRDS 4, tRRDL 6, tFAW 16, tWR 10, tRTP 5, tWTRS 4, tWTRL 8 or 10.

What ram voltage are you running? If 1.45V or above I would consider tRCDRD 20, then you can often lower tRP by 1 or 2.

 

[tbob22]

Okay. Safe suggestions: Try tRAS 34 (tCL+tRCDRD) and tRFC 580 or 590, tRRDS 5, tRRDL 7, tFAW 20, tWR 12, tRTP 8 or 6, tWTRS 4, tWTRL 12.

Fast: tRFC 570 or 560, tRRDS 4, tRRDL 6, tFAW 16, tWR 10, tRTP 5, tWTRS 4, tWTRL 8 or 10.

What ram voltage are you running? If 1.45V or above I would consider tRCDRD 20, then you can often lower tRP by 1 or 2.

Thanks, that will give me some more values to try. Currently at 1.4v on the ram.

I did try tRFC of 590 and was getting errors after 10-20min of testing, similar issues with tFAW at 24, so I just bumped it up to 36.
I ended up enabling GDM which changed tCL to 16, just wasn't stable at 1t. Lowering those may have opened up some headroom for the other timings though, so I may be able to revisit those.

With those changes it does seem to be a much more stable, been testing for about 12h so far.

 

[Space Lynx]

Jeebus was at my side

View attachment 201292

this might be a world record for AMD ram oc's... cause its never been about speed but FCLK matching max...

@TheLostSwede @Andy Shiekh thoughts? do you think this is a world record here?