Share your AIDA 64 cache and memory benchmark here

[FilipM]

I tried just the primaries at CL16, didnt make much of a difference.

 

[Cowboystrekk]

I tried just the primaries at CL16, didnt make much of a difference.

So those tight timings worked at 3933 but no boot with loose ones at 4000?

 

[FilipM]

As I said, it will boot eventually at 3933 and work fine. I go in the bios, if i hit Save & Exit without touching anything its broken again...then it will boot eventually and work just fine. I have not booted at all at 4000Mhz

 

[tabascosauz]

I tried just the primaries at CL16, didnt make much of a difference.

I don't think it's the board. If you can't POST 4000CL22, then it could be a possibility.

I think you're expecting too much of the A0 kit. A0 PCB is trash in more than one way, iirc they suck at lower timings as well? My 4133CL19 Vipers are A0 and while they are mediocre in binning and hit the usual A0 hard wall at 4600 or so, I haven't ever been able to run CL14. Not at any speed, not at any VDIMM, not ever, period. At CL16 and looser it behaves as you'd expect from your average B-die.

I think your expectations are too high. Your A2 timings are too good, you can't just expect A0 on a lower bin without even hitting 1.5V. Hell, I need 1.5V for just 2 single rank sticks, and I've seen even worse binned B-die before. And 4DIMM may require more. Hit it with some more VDIMM see what happens.

The 4000CL19 and 4133CL19 kits are not on the same level as the 4400 Vipers. I found that out the hard way. If you took a quick peek under the heatspreader and you're sure it's A0, return that shit ASAP. You will not get far.

If you wanna make A0 work, pop out the A2 sticks and work out your limits for A0. They will feel quite different I can assure you.

 

[FilipM]

They are both 4000 C19 bins, it just turned out that one of the Kits has a different PCB... I may give only the A0 a shot, although when i did boot at 3933 with 1.49 Vdimm, ran a few passes of TM5 and had no errors

 

[tabascosauz]

They are both 4000 C19 bins, it just turned out that one of the Kits has a different PCB... I may give only the A0 a shot, although when i did boot at 3933 with 1.49 Vdimm, ran a few passes of TM5 and had no errors

If you have a return window, just get rid of it...it's not worth your time running 2 A0s alone, let alone mixing and matching with A2. Like I said, it's not on the same level as 4400 because the 4000s have a much wider variance for quality.

TM5 never caught 140ns artifacting either so it's not always perfect

Give it up to 1.55V for now to rule out it being the board issue.

I'm stuck with the A0s because for clearance I need to run naked sticks and only these are crappy enough that I would ever take the heatspreaders off them and void warranty, dont be like me

 

[FilipM]

Increased ProcODT to 43.6...booted!

Left TM5 at 3933 running now as Im about to get some sleep, but I am almost certain it'll pass

Returning it will be a nightmare, Im outside of EU and having the hassle of returning is greater than selling it here locally if it doesn't work out the way I want to.

 

[glnn_23]

Very impressive. Very good timings! What voltage are you running on ram?

1.535v vdimm in Bios.

 

[FilipM]

There we go

 

[freeagent]

That is pretty sweet man, congrats!

I just cannot get over 1933 with four sticks. Not sure whos to blame here, the CPU, Board, or the operator

Ill try again later

Edit:

I am going to test my Black and Whites on their own now.. that could be the problem lol.. I don't know if they can actually do 2000 MHz lol

 

[FilipM]

I found a few things

ProcODT needs to increase from 2x8 to 4x8. I am now at 40

CAD bus values of 60 20 20 20 or 60 20 40 20, which worked for GDM Off on 2 sticks dont work, all sorts of trouble, couldnt boot etc. I am down to 24 20 24 24

RTT values of 7 3 4 are so far the best. Not risking 7 3 1 as aparently that can kill your ram

VDDG CCD went up from 940mV to 1000mV

VDDG IOD went down from 1060mV to 1050mV

VSOC stays the same


Hope some of this helps your case

 

[Deleted member 205776]

RTT values of 7 3 4 are so far the best. Not risking 7 3 1 as aparently that can kill your ram

Meanwhile my board on XMP profile:

 

[FilipM]

I posted in another thread as well

Struggled with 4x8 at CL15, so moved to CL16....altering a few subtimigs got me just as quick as CL15, which is rather interesting. Pics below for comparison

 

[fec32a4de]

Can't get my hands on Ryzen 5000 series

 

[freeagent]

Yup four sticks can be a real pain lol. Ryzen seems to really like C16 for some reason..

 

[FilipM]

It's the odd tCL, then the fact that it cannot run tCWL at odd numbers...Im guessing

X58 overclocking didn't like even multipliers and even BCLK past 210, I've seen such oddities before

 

[Cowboystrekk]

Can't get my hands on Ryzen 5000 series

Reduce to 3800\1900IF or 3733\1866IF and reduce latency by 10ns. Also try disabling GDM and using dram calc

 

[freeagent]

*Stock*

And PBO +200

Still stock.. sorta

 

[masterdeejay]

Xeon Gold 6138 Es from china. Mixed ram oc to 2666

 

[tabascosauz]

Trying to chase that 60GB/s copy in a daily stable profile for iGPU gaming:

 

[jesdals]

Thats some crazy Infinity speeds. I wonder how much AMD can push next gen with DDR5 and if we will see some asyncronius magic to utilise the potential of ddr5

 

[tabascosauz]

Thats some crazy Infinity speeds. I wonder how much AMD can push next gen with DDR5 and if we will see some asyncronius magic to utilise the potential of ddr5

The other 4650G guy on I think the last page got much better B-die and a better binned CPU. No idea if that 4533 is daily stable or how much vdimm but it sure is impressive, esp the 2272MHz IF, I can do about 2200

That said, mine isn't in an ATX Hero, it's wrapped up in a 5.3L acrylic case about as tall as a can of pringles sooo there's only so far VDIMM I can go on 2 x 80mm Noctuas

The APUs are already halfway there, if they can further minimize the copy and latency penalty a bit then 2:1 at extremely high speeds could very well be viable, if we have the RAM to do it

 

[kibatronic]

I did some testing with a AMD Ryzen 5 4650G Pro before it was installed in my TrueNAS server. I ran Prime95 (Large FFT's) for 7 hours on these settings so it seemed stable. In the server it is only running at pedestrian 3200MHz with unbuffered ECC modules. The memory modules used for the test are Gskill 2x16GB 3200MHz CL14 B-Die, dual rank.

https://imgur.com/qwxBu6L

https://imgur.com/KV5FqGd

 

[shemuru]

 

[Mordante]

Ran Aida64 for the first time ever, but have no idea what these results mean. Any tips?