Associate
@MrPils
Thanks for the info, I'll let you know how I get on. Correct VSOC seen in HWinfo and Ryzen master.
Thanks for the info, I'll let you know how I get on. Correct VSOC seen in HWinfo and Ryzen master.
8 PACK MEMORY RANGE GROWING: SAY HELLO TO 8 PACK RIPPED EDITION & 32GB KITS!!!
- Thread starter Gibbo
- Start date
Did you exchange for a different model of PSU? I've seen first hand in the past a perfectly functional power supply be incompatible with a perfectly functional motherboard from brand A when brands B and C worked fine with it. I would still try a brand and model swap of the PSU if you can.
If you do have a bios flashback button on your board then the bios update is also a feasible possibility, though as I said it could be that the flash procedure makes the reset more likely to happen. As long as you have that hardware button though it should always be recoverable even if you get restarts mid flash.
Finally eliminate one piece of hardware at a time via testing in windows - I mentioned testmem5 for testing memory but there is also furmark for testing your GPU and various Linpack based tests you can use for your cpu (though these have some memory dependancy so test memory first).
No worries, glad to hear vSOC is reported correctly elsewhere. As long as vSOC is correct the other voltages can't be higher than it so you've nothing to worry about in terms of extra voltage being added you didn't know about
Associate
No worries, glad to hear vSOC is reported correctly elsewhere. As long as vSOC is correct the other voltages can't be higher than it so you've nothing to worry about in terms of extra voltage being added you didn't know about
Worked through your instructions and learnt a bit more from it and saw big gains during the inter change tests.
She is now doing her Memtest and will be for a couple of hours. If it passed I'll move on knowning she is reasonably stable to real world uses.
I'll post the results later tonight.
Thanks again for all your help.
Good job man, its all about the gains
What I've tried to outline in my replies throughout the thread is that there are two routes to the end goal, you either pick your voltage you are comfortable with running before you begin clocking or start clocking at maximum safe to see where you end up and then have a go at tuning voltage down at the end. Both are valid approaches, it's all about optimising your memory subsystem for your individual goals. Having a kind of crib sheet to run through as you do it eliminates a lot of the frustration involved compared to taking the blind adjustments route. I learnt everything from watching/reading others tuning memory and blind trial and error experience myself - the most frustrating part honestly is that none of this is documented in guide form by any manufacturers. If I have managed to figure enough out from my own research and experience to be able to help others then there is no reason Corsair, G.Skill, Kingston, Teamgroup or any other module manufacturer could not publish this information. They choose not to and instead leave users to blindly adjust timings or run poorly optimised XMP/Auto configurations and you have to question why? When a user with no memory tuning experience is encouraged by a manufacturer to spend hundreds of pounds on top of the line memory they should not be chucked up river without a paddle. If that information was made available from the source of the modules themselves then there would not be so much unintentional misinformation spread by the media and well intentioned users. I guess it benefits sales more to have tuning memory be some kind of dark art, after all they would rather you spend more money buying higher rated (more expensive) kits under the guise that they are "faster" when in reality a lot of the time that's simply not the case.
Sorry, a bit of a mini rant fell out at the end there
Associate
Happy with grumpyness, I'm on the same boat sailing in the same direction.
For my tune, i used your prompts but picked a voltage I was happy with. In this case 1.45v.
This has meant, the primaries have not moved much from my picture and also trfc but I feel it's a good level tune.
She has been running for 1.5 hours in Memtest without error. I'll let it go to 2 and then start using it.
For my tune, i used your prompts but picked a voltage I was happy with. In this case 1.45v.
This has meant, the primaries have not moved much from my picture and also trfc but I feel it's a good level tune.
She has been running for 1.5 hours in Memtest without error. I'll let it go to 2 and then start using it.
Associate
@MrPils
Here are my results:
Aida:
Read: 59359
Write: 59027
Copy: 54955
Lat: 63.2
Thanks again for you help with this and sharing your knowledge
Here are my results:
Aida:
Read: 59359
Write: 59027
Copy: 54955
Lat: 63.2
Thanks again for you help with this and sharing your knowledge
@MrPils
Here are my results:
Aida:
Read: 59359
Write: 59027
Copy: 54955
Lat: 63.2
Thanks again for you help with this and sharing your knowledge
Slight correction for a typo in my guide post you followed - tWR can always be equal to tCWL, not tCL...the W got away
. You can drop your tWR down from 15 to 14 for free if you want. That change will be pretty much unnoticeable though in the scheme of things and certainly not worth retesting for.Really good results there and welcome to the sub 100 second club in Memtest Easy...not many make it that far! Well done dude
Associate
Slight correction for a typo in my guide post you followed - tWR can always be equal to tCWL, not tCL...the W got away
Really good results there and welcome to the sub 100 second club in Memtest Easy...not many make it that far! Well done dude
Wouldn't of got there if you had not helped. I'll mirror what others have said, the info you have should made into a sticky - you know the process
Those are really nice dude, I was waiting and waiting for the 4400 C16's to come in stock and ended up losing patience and jumping on the 4000 C15's instead. Then a week later the 4400's came in stock...typical! The 4000 C15's are roughly the same as your 4400 C16 kit at 3800mhz, both our kits do TRCDRD 14 whereas my Patriot 4400 binned kit will only do 15. Your kit there should technically be a slightly better bin, I don't think we will be able to see extra performance from that until we can get past 3800mhz without WHEA errors though. The timings are already at minimum for both kits at the current limit of 3800mhz, but you are running a lower voltage there than I used. To be fair I didn't try reducing, I just went straight in at what I had the Patriots at (1.51v) as a quick test in a known setup before I test them in the new 5600x build (which will hopefully be tomorrow).
I like the cheeky little bclk bump btw - I see what you did there
. Was that from spread spectrum being enabled or did you manually set it? That chip has a strong fabric for Ryzen 3000.Christ I didn't see that was a 5950x...must be tired I read it as 3950x lol.
I'm on the latest Impact bios and my 5600x will get into windows at 4266mem/2133IF...just spewing WHEA errors tho. Anything using over the 1900 setting generates constant errors, it just generates more at higher speeds so its not truly stable over 1900mhz regardless. I've seen this with all Ryzen 5000 chips over 1900 so far - AMD need to work more on the AGESA to get the cpus stable up there (if its even possible at all).
Associate
I'm currently running my sticks at 1.44v, I know its been discussed but is there any real danger of running more into them? I mean, I don't want to break them.
I'm currently running my sticks at 1.44v, I know its been discussed but is there any real danger of running more into them? I mean, I don't want to break them.
1.5v is fine, just don't stress test them at that voltage without a fan over them. You can remove the fan once you're done stress testing, gaming won't get them anywhere near that hot.
Dual Rank vs Single Rank comparison update:
I hit problems completing the Dual Rank vs Single Rank comparison today. The Crosshair 8 Impact keeps throwing a fit with the single rank sticks requiring a cmos clear and a full retune back to tweaked settings. I've no idea as of now what's causing it, I spent around 4 hours messing with it today. I got the majority of the benchmarking done before it threw its first wobbler up until which point the results I was getting were totally in line with my Gigabyte board. I even tried using my Patriot sticks in case the G.Skill kit had gone faulty, but it (mis)behaved just the same with those installed too. I'm beginning to wonder if the board has a training issue when using 1T command rate, I'll do some more investigating into that in the next few days - probably Saturday. Whatever the issue was it would suddenly start erroring in benchmarks when it had already passed full stress testing. Restarts, voltage changes, reducing mhz or slackening off timings didn't help, the only way to get back stable was a cmos clear and manually retune the timings back to exactly what they were when it went unstable from full auto values. A cmos clear and loading any saved profile went straight back to erroring and blue screening again. Truly bizarre behaviour and something I haven't come across before, especially considering putting the dual rank kit back in and loading up its profile was instantly stable.
Despite the issues late on in the session I can give a quick idea of where things end up though:
MaxxMem2 and Geekbench 3 gave a slight lead to the dual rank sticks (around 1%)
Geekbench 4, x265 4k and Y-Cruncher 500m tests were a draw
Y-Cruncher 1b gave a slight lead to single rank sticks (around 1%)
Y-Cruncher 2.5b, Membench Easy, Membench Default, Timespy and Timespy Extreme gave a large lead to the single rank sticks. (2% to 5%)
I did not run AIDA64 as I got side-tracked trying to figure out what was going on with the board, however I would expect the single rank sticks to have lower latency whilst the dual ranked sticks had higher bandwidth.
When (if?) I manage to figure out what's going on with the board with the single rank sticks I'll finish the AIDA bench, rerun all the benchmarks to make sure nothing was wrong with the results and then post up the screenshots with a proper conclusion as per the first part of the comparison. There's so much extra junk in the Asus bios I'm going to have a lot of things to mess about with to try and track down the cause of the random instability.
I hit problems completing the Dual Rank vs Single Rank comparison today. The Crosshair 8 Impact keeps throwing a fit with the single rank sticks requiring a cmos clear and a full retune back to tweaked settings. I've no idea as of now what's causing it, I spent around 4 hours messing with it today. I got the majority of the benchmarking done before it threw its first wobbler up until which point the results I was getting were totally in line with my Gigabyte board. I even tried using my Patriot sticks in case the G.Skill kit had gone faulty, but it (mis)behaved just the same with those installed too. I'm beginning to wonder if the board has a training issue when using 1T command rate, I'll do some more investigating into that in the next few days - probably Saturday. Whatever the issue was it would suddenly start erroring in benchmarks when it had already passed full stress testing. Restarts, voltage changes, reducing mhz or slackening off timings didn't help, the only way to get back stable was a cmos clear and manually retune the timings back to exactly what they were when it went unstable from full auto values. A cmos clear and loading any saved profile went straight back to erroring and blue screening again. Truly bizarre behaviour and something I haven't come across before, especially considering putting the dual rank kit back in and loading up its profile was instantly stable.
Despite the issues late on in the session I can give a quick idea of where things end up though:
MaxxMem2 and Geekbench 3 gave a slight lead to the dual rank sticks (around 1%)
Geekbench 4, x265 4k and Y-Cruncher 500m tests were a draw
Y-Cruncher 1b gave a slight lead to single rank sticks (around 1%)
Y-Cruncher 2.5b, Membench Easy, Membench Default, Timespy and Timespy Extreme gave a large lead to the single rank sticks. (2% to 5%)
I did not run AIDA64 as I got side-tracked trying to figure out what was going on with the board, however I would expect the single rank sticks to have lower latency whilst the dual ranked sticks had higher bandwidth.
When (if?) I manage to figure out what's going on with the board with the single rank sticks I'll finish the AIDA bench, rerun all the benchmarks to make sure nothing was wrong with the results and then post up the screenshots with a proper conclusion as per the first part of the comparison. There's so much extra junk in the Asus bios I'm going to have a lot of things to mess about with to try and track down the cause of the random instability.
Woomack did a lot of testing. 4200 is about the max you can expect unless you have a really great board. The Msi z490i Unify is around this for reference.
4400 C17 requires 1.6v if you have the board to clock it. I guess evga ones?
4200 16-17-17-38 1.50V is around the max you will get most of the time with these.
https://www.overclockers.com/forums...on-2x16GB-DDR4-3600-CL16-TDPPD416G3600HC16GBK
Associate
Thanks
Soldato
How does mine look?
I tried IF at 1900, it booted to the login screen then crashed, then I had a hard time getting it back again, every time I entered the bios it would hang completely, took me several attempts to change it back to 1800, save and reboot. I'm worried to try over 1800 again, getting to the cmos reset is a pain in this ITX system.
I tried IF at 1900, it booted to the login screen then crashed, then I had a hard time getting it back again, every time I entered the bios it would hang completely, took me several attempts to change it back to 1800, save and reboot. I'm worried to try over 1800 again, getting to the cmos reset is a pain in this ITX system.