Poll: Ryzen 7950X3D, 7900X3D, 7800X3D

[01001101]

Yeah. I'm guessing if its a specific timing causing the error. Can try WR at a higher value too like 90 see if its the culprit.

Ok. Cool. I'll try 90 WR in a bit. I adjusted the RFC 500, RFC2 400, WR 54 and ran tests with both GDM enabled and disabled. Both had errors. It's really puzzling. I wonder if it's something I've not done right in the BIOS or just the RAM not playing nice with the Mobo?

 

[LtMatt]

Ok. Cool. I'll try 90 WR in a bit. I adjusted the RFC 500, RFC2 400, WR 54 and ran tests with both GDM enabled and disabled. Both had errors. It's really puzzling. I wonder if it's something I've not done right in the BIOS or just the RAM not playing nice with the Mobo?

Yeah I'm stumped, must be the kit you are using I guess.

You should try running Karhu with the memory using the EXPO II profile see if it passes.

 

[01001101]

Yeah I'm stumped, must be the kit you are using I guess.

You should try running Karhu with the memory using the EXPO II profile see if it passes.

I might do that dude. Process of elimination. For the price I paid for the kit I would expect it just to work. I was just looking on Corsairs website and it's currently £40 off rrp!

 

[LtMatt]

I might do that dude. Process of elimination. For the price I paid for the kit I would expect it just to work. I was just looking on Corsairs website and it's currently £40 off rrp!

If it fails with EXPO, maybe its a bad kit.

 

[01001101]

Yeah. I'm guessing if its a specific timing causing the error. Can try WR at a higher value too like 90 see if its the culprit.

90WR gave me the best PB yet. Error count 1 @ 27m:48. I'll go load EXPO defaults now and run RAM Test again to see what we get.

 

[bluesouljah]

90WR gave me the best PB yet. Error count 1 @ 27m:48. I'll go load EXPO defaults now and run RAM Test again to see what we get.

I have the same mobo and RAM and it just works mate. Maybe it's the kit?

 

[01001101]

If it fails with EXPO, maybe its a bad kit.

I have the same mobo and RAM and it just works mate. Maybe it's the kit?

It's still going. Default bios settings with only EXPO II enabled.

I'll leave it going. See what happens?

 

[LtMatt]

It's still going. Default bios settings with only EXPO II enabled.

I'll leave it going. See what happens?

Looks like the MEM is okay then, you just have a turd bin.

Start changing the timings one or two at a time until you figure out which one it is then we can relax that one.

 

[01001101]

Looks like the MEM is okay then, you just have a turd bin.

Start changing the timings one or two at a time until you figure out which one it is then we can relax that one.

Turning out to be a fun little project.
I think I should change 1 at a time.
What's the best testing method?
Test, revert back, change another, test and repeat? Or just change, test, change, test and so on?
Sorry if I'm overcomplicating it. Just want to be methodical in testing.

 

[LtMatt]

Turning out to be a fun little project.
I think I should change 1 at a time.
What's the best testing method?
Test, revert back, change another, test and repeat? Or just change, test, change, test and so on?
Sorry if I'm overcomplicating it. Just want to be methodical in testing.

If it was me I'd just buy a better memory kit Lol, but I like getting every drop of performance from my system.

Best to start off with the 5 primary timings and go from there, changing two at a time after those primaries are confirmed as stable.

 

[Gavster]

Yeah I'm stumped, must be the kit you are using I guess.

You should try running Karhu with the memory using the EXPO II profile see if it passes.

Keep the fire extinguisher handy

 

[jigger]

Keep the fire extinguisher handy

Rtx4090?

 

[Legion]

It's still going. Default bios settings with only EXPO II enabled.

I'll leave it going. See what happens?

If these are Samsung IC's and not Hynix M-die. You are unlikely to get these running at much more beyond EXPO / XMP. They just don't scale like Hynix M (and A) does.
Matt is also giving you suggestions specifically for Hynix sets here as well, not Samsung IC based sticks.
CPU-Z will tell you which IC's are on the mem sticks.

 

[01001101]

If it was me I'd just buy a better memory kit Lol, but I like getting every drop of performance from my system.

Best to start off with the 5 primary timings and go from there, changing two at a time after those primaries are confirmed as stable.

I might have to. I guess I'm just committed now to finding out where the problem is in the hope the error isn't between the chair and the keyboard.

Keep the fire extinguisher handy

It's the warmest room in the house. Saves putting the heating on.

If these are Samsung IC's and not Hynix M-die. You are unlikely to get these running at much more beyond EXPO / XMP. They just don't scale like Hynix M (and A) does.
Matt is also giving you suggestions specifically for Hynix sets here as well, not Samsung IC based sticks.
CPU-Z will tell you which IC's are on the mem sticks.

Good to know. I'm pretty sure these are Hynix. I'll double check when I get home.

 

[LtMatt]

If these are Samsung IC's and not Hynix M-die. You are unlikely to get these running at much more beyond EXPO / XMP. They just don't scale like Hynix M (and A) does.
Matt is also giving you suggestions specifically for Hynix sets here as well, not Samsung IC based sticks.
CPU-Z will tell you which IC's are on the mem sticks.

Yeah we checked with CPU-Z earlier and they were hynix AFAIK.

 

[01001101]

Yeah we checked with CPU-Z earlier and they were hynix AFAIK.

Can confirm.

 

[Addicted]

Surely it's just luck of the draw and your binning isn't so good as others?

 

[01001101]

Surely it's just luck of the draw and your binning isn't so good as others?

I've started testing again. See how it goes.
I don't suppose a bad bin is covered under warranty? I'll open up a ticket with Corsair. See what they have to say.

 

[Legion]

Can confirm.

I'd try 30, 38, 38 and then tighten everything else as per Matt's suggestions. Primary timings don't really make a big impact. Seconds and tertiarys do however and this is where you want to focus your efforts to get anything out of that kit above EXPO settings.
Go look at the "Actually hardcore overclocking" channel on youtube. Buildzoid (the channel creator) has some vids to help you grasp DDR5 tuning without just following settings with little to no knowledge of what it all means.
Learning the nuances of DDR5 will help you tune your kit without just blindly following settings being given to you.

Hynix M really doesnt like those RAS to CAS 36 and RAS Precharge 36 settings. you can take it much further with those 2 settings at 38 or 39. It's entirely bin dependent sticks do vary a fair amount.
38 and 39 will also let you take those sticks to 6200 and 6400 if the rest of your system can handle it.

 

[LtMatt]

I'd try 30, 38, 38 and then tighten everything else as per Matt's suggestions. Primary timings don't really make a big impact. Seconds and tertiarys do however and this is where you want to focus your efforts to get anything out of that kit above EXPO settings.
Go look at the "Actually hardcore overclocking" channel on youtube. Buildzoid (the channel creator) has some vids to help you grasp DDR5 tuning without just following settings with little to no knowledge of what it all means.
Learning the nuances of DDR5 will help you tune your kit without just blindly following settings being given to you.

Good suggestion, could well be RCD tbf. RCD does help lower latency in Aida so always want it as low as you can go, assuming stability of course.