• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

*** The Official Alder Lake owners thread ***

I understand, as to why you chose different GPU drivers to the game ready ones, IIRC, for your 3090 that stability is of utmost importance to that of absolute performance.
The default for these boards is Gear 2 and the quality of the memory controller varies quite a bit for each CPU. Whether it makes much in the way of real world differences that would have to be checked but stability has to be first achieved. I'm guessing since your test at below XMP speed you are noting issues.

My 4000Mhz RAM will just plod along at Gear 1......

RAM-Test-at-2000-Mhz.png


with 2000Mhz memory controller and a overclock of the CPU.
But that is with 16GB in total and not 32GB sticks, more stress from what I understand on that memory controller.
It would be interesting to see if you put the VCSSA voltage back to auto and then rebooted and went back in the BIOS what the voltage is set to by the board. Bet it isn't 1.3v.
 
Ah I did check by setting auto SA voltage yesterday when faffing about with the settings after rollling back to F6 and I recall seeing the SA voltage showing at 1.47v at one point but then another time it was showing 0.9v so it's a bit all over the place in auto it would seem. Perhaps this is why some are experiencing these things in different configurations :p

My current test is still running and currently at 3600% with no errors. I will let it get to the full 6400% before I call it a safe result for stability at 3400MHz, then will up the DRAM voltage a bit more, maybe 1.43v and put it back to 3600MHz and retest. I recall last night at 3600MHz I was seeing a single error in karhu at around 70% coverage, so quite early on so was able to replicate it via multiple runs so knew it wasn't an anomaly whereas the single error at 5300% could well have been an anomaly as that was at 1.4v 3400MHz whereas the test currently running is at 1.42v.

I will have to do some game benchmarks with Gear 2 vs Gear 1 to figure out if there is any real world impact. If the difference is small then it will be better to edge on the side of stability and use Gear 2 I guess at stock 3600MHz for the RAM.

I am currently using the PC whilst the memory test is running too, I figure this is an ideal way to properly test for stability and errors if I am just generally using the PC, browsing, some media etc. The CPU cores are all at 100% and the temps are 55 degrees so I guess a memory test run like this only puts stress on the memory controller in the CPU rather than other areas that would otherwise generate more heat.
 
mrk said:
I am currently using the PC whilst the memory test is running too, I figure this is an ideal way to properly test for stability and errors if I am just generally using the PC, browsing, some media etc. The CPU cores are all at 100% and the temps are 55 degrees so I guess a memory test run like this only puts stress on the memory controller in the CPU rather than other areas that would otherwise generate more heat.
Click to expand...

This is a common misconception but you want a stress to have full access to the resource to maximize the usage/stress.
 
Hmm I see - Secondary question in that case, in karhu do you leave it on default or also tick the stress FPU box too?
 
Do you have a link to the profile perhaps? Not heard of Testmem 5 but looks like it's a community thing?

Edit*
I realise that Gear 1 allows the CPU MCto run at the same speed as the RAM, so with my 3600 in mind, would that be 1800MHz at the MC. Whereas in gear 2 it gets cut in half to allow for running higher memory bandwidth. For professional worldload stuff the latter would be more beneficial;, for gamers and genera use the latency is more important hence Gear 1.

The ultimate thing I guess is what was mentioned above that the CPU IMC could handle Gear 1 and be stable, or it might not, it varies from CPU to CPU like OC lottery between the same CPU model.
 
Last edited:
@mrk, yup that is about it in terms of speeds and memory controller. Here is mine at Gear 1....

memory.png


Memory controller at 2000Mhz and then the DRAM speed also at 2000Mhz, doubled to 4000Mhz Memory. If at Gear 2 it would be then 1000Mhz for that frequency on the memory controller.

Unless things have changed Intel, IIRC, rate anything above 3200Mhz as being overlocked in terms of memory and the default is Gear 2. Motherboard manufacturers have different interpretations on that, altho I do not know what my MSI board set as default, but it was only around Rocket Lake time that we got the Gear 1 and 2 choices.
I remember watching "buildzoid" over at YT explain in one recent(ish) video that his memory speeds achieved were limited due to the poor memory controller on the AL CPU he was using at the time. I suppose there are elements of pot luck, along with settings and testing etc that go into this.
And then "does it actually matter" in terms of real life performance loss or gain for the trouble taken to get there, if possible.

Changing over the Gear 2 is easy enough for you to test both stability and performance.
 
Cheers for those, I guess I can only try Gear 2 3600 and see what things are like. I will stay on BIOS F6 though I think as everything works fine and then explore future options at the next update.

My Gear 1 vs 2 testing will be based on exporting a batch of photos in Lightroom, then running a game benchmark. If the difference is small enough to not matter, then I'll stick with Gear 2 I guess for outright stability.

The fun continues :D
 
I suspect you are correct and have been mulling it over whether the slight bit of faff of testing, rebooting, changing setting, rebooting, testing again etc is worth the time - The karhu run has surpassed 6700% with no errors so I have stopped it. According to their FAQ, up to 6400% coverage is a 99.4% error detection rate so I am happy that the current config at 3400MHz @ 1.43v and SA at 1.3v is completely stable in Gear 1.

I think what I will do now is bump it up to 3600MHz and then raise the DRAM voltage one step up and retest just out of curiosity to rule out DRAM voltage being the reason I got the error at 5300% coverage last night.

Edit1*
Back at 3600MHz and RAM at 1.43v - Still using Karhu for now but 1 error at 681% coverage which is around 85% error detection rate.

Bumping up to 1.44v and re-running. SA still at 1.3v.

Edit2*
3600MHz @ 1.44v and 1 error at 1045% coverage.

Edit3*
3600MHz @ 1.46v, 6400% coverage and zero errors. I have found the stable point for my build now. Finally.
 
Last edited:
Looks like I have found the full stable settings for 3600MHz Gear 1, see edits above!
 
I suspect that is indeed the case strange how for a short while everything seemed fine though, maybe it's the memory training thing that has settled and started to highlight the need for more voltage or something?

Either way, this is good stuff, I even played an hour of Cyberpunk with the latest patch and has no CTD.
 
Vimes said:
Good result there all round, whilst still maintaining Gear 1. Now keep on F6 BIOS and you will be fine :D
Click to expand...

F6 is here to stay unless GB put some great new features in a future update :p

Just did my first 4K 60fps video production in Davinci Resolve and while it uses Cuda during the final render, the CPU utilisation during editing and seeking etc was superb. Took seconds to produce the end video.

EGXtU7v.jpg
 
@mrk. You must be well pleased overall with your AL platform, especially coupled with your GPU. I'll remind you of sticking with F6 when you next post about F8 causing stability issues :D
 
You must log in or register to reply here.
Back
Top Bottom