• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

9800X3D Memory Overclocking

Ran AIDA64 for the first time to check my latencies. 70ns seems alright.

1746997306555.png


I just bumped my MEM VDD and MDM VDDQ from auto (1.35V) to 1.40V. Been having some stability issues and BSODs lately. Not sure if it's memory-related or something else, but memory seemed like the right place to start.


Still not stable, though. Getting errors after 10 minutes of OCCT. RAM temps got up to about 67C, which may be a bit hot. Might look into better heat spreaders for the sticks and/or improving airflow around them.
 
Last edited:
Decided to take the stock spreaders off my RAM. Heat gun and a plastic pry tool made it pretty easy. Based on the impressions left on the thermal tape, several of the modules were not contacting the spreader completely.

I also had a small piece of metal fall out of my PC last time I cleaned it. I could never figure out where it came from, but in the process of removing the memory, I was able to. It seems to come from the end of the memory slot. I'm not really sure what its purpose is. Luckily, I kept it and was able to push it back in. I think it's prevented from falling out again while the memory is installed, but I'll have to be careful removing memory.

20250511_155348.jpg20250511_155324.jpg

Per OCCT, temps are slightly improved without the heat spreaders. Still at 66C 15 minutes into the test, and that seems way too high from what I'm reading. On the other hand, I'm not entirely sure how accurate the SPD Hub Temperature reported in OCCT is.

This time around I started hitting errors 16 minutes into the test. I think I might need some active cooling. Memory is toasty in this little case.
 
I ordered some Bykski copper heatsinks and a 60mm Noctua fan. Hopefully I can get these temps down.
 
Epos7 said:
Decided to take the stock spreaders off my RAM. Heat gun and a plastic pry tool made it pretty easy. Based on the impressions left on the thermal tape, several of the modules were not contacting the spreader completely.

I also had a small piece of metal fall out of my PC last time I cleaned it. I could never figure out where it came from, but in the process of removing the memory, I was able to. It seems to come from the end of the memory slot. I'm not really sure what its purpose is. Luckily, I kept it and was able to push it back in. I think it's prevented from falling out again while the memory is installed, but I'll have to be careful removing memory.

View attachment 729413View attachment 729415

Per OCCT, temps are slightly improved without the heat spreaders. Still at 66C 15 minutes into the test, and that seems way too high from what I'm reading. On the other hand, I'm not entirely sure how accurate the SPD Hub Temperature reported in OCCT is.

This time around I started hitting errors 16 minutes into the test. I think I might need some active cooling. Memory is toasty in this little case.
Click to expand...
1.4V on DDR5 is low/moderate.

You've either got some specific timings set which are affecting the heat/stability. Or......what is your SOC voltage? Maybe it just needs to come up? 1.4 on its own isn't the reason for your high RAM temps.

Its also possible you just got a batch of RAM which isn't as good for tweaking.
 
Last edited:
Epos7 said:
I just bumped my MEM VDD and MDM VDDQ from auto (1.35V) to 1.40V. Been having some stability issues and BSODs lately. Not sure if it's memory-related or something else, but memory seemed like the right place to start.


Still not stable, though. Getting errors after 10 minutes of OCCT. RAM temps got up to about 67C, which may be a bit hot. Might look into better heat spreaders for the sticks and/or improving airflow around them.
Click to expand...

My 24Gb M-Die starts to error out above 65C if I run tREFI at 50000.
 
I have heard that TREFI can be temp sensitive.

Yeah looking at Epos7's Zentimings screenshot:

96GB of RAM + some of the sub-timings are ultra low. Even for a 32GB set, some of those numbers would be a stretch on some A-Die (A-die is generally better for very low timings). And the main timings overall, are a bit relaxed.

Might be able to balance things out better, by lower some of the main timings and relaxing the sub-timings a little bit.

But also....its 96GB of RAM. You shouldn't expect to run parallel to a 32GB A-Die or a good set of 48GB M-Die.

*There is still room to increase the SOC volage. It can go up to 1.3. My 7800X3D would happily accept any RAM tweaks at 1.25. But, other setups can need more. And some can do it with less.
 
  • Like
Reactions: Epos7
like this
I used the Buildzoid vids to gather the settings. He recommended using the manufacturer specs for the base timings, then setting the rest to what he specified.

It's 96GB in an ITX case, so I'm not expecting to set any records. I do think I need to improve heat dissipation.

Will also try bumping SOC. Guessing that'll only raise temps further, though. Based on findings from Igor's lab and TPU, these temps are way too high.
 
Epos7 said:
I used the Buildzoid vids to gather the settings. He recommended using the manufacturer specs for the base timings, then setting the rest to what he specified.
Click to expand...
Well its going to be different for each set of RAM. And you are also running 96GB.

Buildzoid insists that the main timings aren't worth much. And he's sort of correct in that, sub-timings can have as much or more effect. However, lowering the main timings does affect your Aida latency score. Which is why I said that you may be able to relax some sub-timings. But, lower main timings. To potentially balance out end latency VS temps.

It's 96GB in an ITX case, so I'm not expecting to set any records. I do think I need to improve heat dissipation.
Click to expand...
Good luck!
RAM sticks are so close together for ITX, there is barely any gap between the sticks.

I'm now running DDR5 8400 in an ITX build @ 1.5V with an Intel 265K.

I haven't done a 15 minute RAM test. But, its stable for gaming. And I have done like 7 or 8 minutes RAM test in OCCT.
IIRC, my 7-ish minute RAM test temp was about 63C. and my gaming temp was about 59C. The RAM sinks are thick and get real hot. Visually, they seem well attached.
And overall, the air gets kicked around in the case real good. As I have a side mounted 240mm AIO and the fans hang only a couple inches from the RAM. Currently set to exhaust. I tried it with the fans flipped to "pull" through the RAD and have the fan blowing toward the RAM. But, it didn't seem to make a real difference on the RAM. But it did increase my GPU temps 4 degrees.

However, I did just order some fan shroud extensions, to hang the fan even closer to the RAM. Just to see what's up.

Will also try bumping SOC. Guessing that'll only raise temps further, though. Based on findings from Igor's lab and TPU, these temps are way too high.
Click to expand...
I don't think that SOC voltage has much effect on the RAM temps. But I could be wrong. SOC voltage should mostly be about stabilizing the memory controller and fabric. And that can increase the CPU temps a bit.
 
  • Like
Reactions: Epos7
like this
chameleoneel said:
And overall, the air gets kicked around in the case real good. As I have a side mounted 240mm AIO and the fans hang only a couple inches from the RAM. Currently set to exhaust. I tried it with the fans flipped to "pull" through the RAD and have the fan blowing toward the RAM. But, it didn't seem to make a real difference on the RAM. But it did increase my GPU temps 4 degrees.

However, I did just order some fan shroud extensions, to hang the fan even closer to the RAM. Just to see what's up.


I don't think that SOC voltage has much effect on the RAM temps. But I could be wrong. SOC voltage should mostly be about stabilizing the memory controller and fabric. And that can increase the CPU temps a bit.
Click to expand...

Interesting...

I have a 280mm side rad with the fans set to intake. That should be blowing a decent amount of air over the motherboard and RAM slots, but I do have several cables and liquid tubes passing over the sticks. I'm thinking about installing the 60mm fan on the top of my case sucking in cool air. That could point down the length of the RAM slots, and I'm thinking about drawing up and 3D printing a mount for the fan that could also double as an airflow tunnel. Fan and heatsinks should be here within a couple days, so I'm curious to see how much temps improve with the same settings and just adding the copper heatsinks. And again when incorporating the 60mm fan. At that point I'll have a good baseline and can start tweaking my settings further.

First thing to try after the heatsinks and fan will be SOC. Thanks!
 
Epos7 said:
Interesting...

I have a 280mm side rad with the fans set to intake. That should be blowing a decent amount of air over the motherboard and RAM slots, but I do have several cables and liquid tubes passing over the sticks. I'm thinking about installing the 60mm fan on the top of my case sucking in cool air. That could point down the length of the RAM slots, and I'm thinking about drawing up and 3D printing a mount for the fan that could also double as an airflow tunnel. Fan and heatsinks should be here within a couple days, so I'm curious to see how much temps improve with the same settings and just adding the copper heatsinks. And again when incorporating the 60mm fan. At that point I'll have a good baseline and can start tweaking my settings further.

First thing to try after the heatsinks and fan will be SOC. Thanks!
Click to expand...
None of my cables pass over the sticks. I did change the way I insert the RAD, so that the liquid cables no longer cover the RAM Sticks. That seemed to make more of a difference than changing the fans to pull/intake. As the sticks heat up more slowly, when I watch them in OCCT. Compared to when I had the liquid cables over the top of them/laying on them.

**Its also worth keeping in mind that you have a 96GB set with dual sided, 48GB sticks. That's going to run warmer, compared to single sided 16 and 24GB sticks from 32GB and 48GB sets.
 
Last edited:
  • Like
Reactions: Epos7
like this
Epos7 said:
Decided to take the stock spreaders off my RAM. Heat gun and a plastic pry tool made it pretty easy. Based on the impressions left on the thermal tape, several of the modules were not contacting the spreader completely.

I also had a small piece of metal fall out of my PC last time I cleaned it. I could never figure out where it came from, but in the process of removing the memory, I was able to. It seems to come from the end of the memory slot. I'm not really sure what its purpose is. Luckily, I kept it and was able to push it back in. I think it's prevented from falling out again while the memory is installed, but I'll have to be careful removing memory.

View attachment 729413View attachment 729415

Per OCCT, temps are slightly improved without the heat spreaders. Still at 66C 15 minutes into the test, and that seems way too high from what I'm reading. On the other hand, I'm not entirely sure how accurate the SPD Hub Temperature reported in OCCT is.

This time around I started hitting errors 16 minutes into the test. I think I might need some active cooling. Memory is toasty in this little case.
Click to expand...
989136_IMG_20240303_113308703.jpg
I Velcro taped my ram cooler fan to the back of the vector plate and it works like a champ. I decreased the RAM temps by at least 10c degrees or more. I made a thread about it a while ago. Just an idea for you.
 
xDiVolatilX said:
I Velcro taped my ram cooler fan to the back of the vector plate and it works like a champ. I decreased the RAM temps by at least 10c degrees or more. I made a thread about it a while ago. Just an idea for you.
Click to expand...

That looks like a good solution. My case is too compact to have anything above the RAM, however. I'm gonna have to mount something along the top edge of the motherboard, above the sticks blowing down.

I think I can squeeze one of these slim 60mm fans up there. Will find out when it arrives tomorrow.

1747170056316.png
 
Epos7 said:
That looks like a good solution. My case is too compact to have anything above the RAM, however. I'm gonna have to mount something along the top edge of the motherboard, above the sticks blowing down.

I think I can squeeze one of these slim 60mm fans up there. Will find out when it arrives tomorrow.

View attachment 729753
Click to expand...
But traditionally case air flow is front to back bottom to top. You're going to have the RAM air blowing down the case? Perhaps your case air flow is not typical?
 
xDiVolatilX said:
But traditionally case air flow is front to back bottom to top. You're going to have the RAM air blowing down the case? Perhaps your case air flow is not typical?
Click to expand...

I have intake fans on the side and bottom with an exhaust fan on the back. PSU exhausts out the top. I think in these small cases, there isn't really much directionality, you just want to get air where it's needed.
 
Got the Bykski copper heatsinks installed today. With no other changes, I'm seeing memory temps at 55C after 10 minutes in OCCT. That's a 12 degree improvement over the stock "heatsinks" and about 10 degrees over the bare sticks. OCCT is an edge case that's only stressing the memory. In real world cases, I'd be loading up my CPU or GPU which would cause my fan profile to speed up and more air would be moving over the sticks.

Started seeing errors at the 16 minute mark when the sticks were up to 60C. Same point in the test, lower temp than bare sticks (65C).

Next up is adding the 60mm fan. So far, I haven't been able to figure out how to use memory temps as a curve source in Aquasuite.
 
Last edited:
Epos7 said:
Got the Bykski copper heatsinks installed today. With no other changes, I'm seeing memory temps at 55C after 10 minutes in OCCT. That's a 12 degree improvement over the stock "heatsinks" and about 10 degrees over the bare sticks. OCCT is an edge case that's only stressing the memory. In real world cases, I'd be loading up my CPU or GPU which would cause my fan profile to speed up and more air would be moving over the sticks.

Started seeing errors at the 16 minute mark when the sticks were up to 60C. Same point in the test, lower temp than bare sticks (65C).

Next up is adding the 60mm fan. So far, I haven't been able to figure out how to use memory temps as a curve source in Aquasuite.
Click to expand...
Also probably worth leaving OCCT open to monitor RAM temps. And then actually use your computer, and see what your real world temps are.
 
  • Like
Reactions: Epos7
like this
Added the fan and saw another improvement in temps. Was still getting errors in OCCT test so I bumped SOC up to 1.3 as suggested. Made it through an hour of OCCT w/o errors. Temps didn't go over 54C either.



1747295481184.png
 
Last edited:
Ran another OCCT test and hit errors again. Might need to back off some timings.
 
xDiVolatilX said:
Have you considered grabbing the 6000 cl26 kit and just getting it over with? Lol
Click to expand...
The 96GB 6000 CL28 kit is pretty spendy. I can only imagine what the CL26 costs lol.

I have the 6400 CL32 kit. Was hoping to get 6000 CL30 from it, but I might have to back off the timings a bit.
 
Epos7 said:
The 96GB 6000 CL28 kit is pretty spendy. I can only imagine what the CL26 costs lol.

I have the 6400 CL32 kit. Was hoping to get 6000 CL30 from it, but I might have to back off the timings a bit.
Click to expand...
How many GB is your kit?
I have the 64GB GSkill 6400 C32 also.
I think the 6000 C26 is faster right?
I would consider switching to it if it came in 48GB because I don't wanna go back to 32.
Gonna look around at prices although not sure if it is indeed faster than 6400 C32? Lol

I see it for 350 on Amazon holy shat
Does expo work on I tell perfectly fine? I think it doesn't have any xmp profiles ..
 
1747456982888.png


I backed off a few of the timings a bit. Copied these from JSHamlet234 settings above:

tRAS 28 > 48
tRC 68 > 84
tRFC 500 > 540
tRFC2 400 > 290
tRFCsb 300 > 250

Did an hour of OCCT w/o errors, but so did the slightly faster timings. Paid for a license so I can run longer tests.
 
Started P95 Large FFTs test w/ AVX on. So far I'm an hour in without issues. PC is running like butter. CPU has settled in at about 68C.
 
Epos7 said:
Why AVX off?
Click to expand...
Heavier load. Don't want it to downclock in game if the game uses AVX mainly. Oh no I mean AVX offset off in the BIOS I set the AVX offset to 0 so it doesn't downclock itself not AVX off my bad lol. Of course, during the stress test leave AVX on lol i usually just do the blend. Seems to stress everything all around very heavy.
 
  • Like
Reactions: Epos7
like this
Crashed after 5-6 hours of Prime95 Large FFTs. Still seeing some stability issues in general use.
 
Epos7 said:
Crashed after 5-6 hours of Prime95 Large FFTs. Still seeing some stability issues in general use.
Click to expand...
We have a very similar setup. I thought I'd post my RAM timings so you could get an idea of what's worked for me thus far. I'm happy with it.

I played with vsoc a bit but found no added stability over 1.2v. I haven't tried messing with PowerDown. This is the GSkill FlareX kit.

1748141535675.png
 
  • Like
Reactions: Epos7
like this
Astral Abyss said:
We have a very similar setup. I thought I'd post my RAM timings so you could get an idea of what's worked for me thus far. I'm happy with it.
Click to expand...

Thanks! I think I'm going to RMA my kit a second time. I should at least be able to run EXPO. Hopefully they won't notice I took the heatspreaders off then reinstalled.

I bought an identical kit from Amazon while I was waiting for my first RMA, and that kit worked flawlessly for several weeks before I returned it. The original kit and the one I got back from RMA have had a lot of instability every day.
 
Last edited:
I'm kind of an idiot. Didn't realize the Ripjaws series is for Intel and Flare series is for AMD.

My kit is a Ripjaws kit, which might explain why it hasn't been working so well with my AMD setup. At least for the EXPO timings.
 
Epos7 said:
I'm kind of an idiot. Didn't realize the Ripjaws series is for Intel and Flare series is for AMD.

My kit is a Ripjaws kit, which might explain why it hasn't been working so well with my AMD setup. At least for the EXPO timings.
Click to expand...
AFAIK, its the same memory. Simply comes from the factory with different timings/sub-timings. Manually tweaking should net the same results. Aside from the usual chip lottery.

I used some A-Die DDR5 7200 which only had an XMP profile, with my 7800X3D. I manually tweaked down to DDR5 6200 and manually changed most of the timings/sub-timings.
 
chameleoneel said:
AFAIK, its the same memory. Simply comes from the factory with different timings/sub-timings. Manually tweaking should net the same results. Aside from the usual chip lottery.
Click to expand...

I'm pretty sure that was my thought when I bought it, too. Think I've just gotten some bad sticks again. I'm nice and stable at 5600Mhz with auto timings. Started another RMA with G.Skill.

In the meantime I ordered the Flare 6000 CL30 kit. I'll install and keep that one and sell the kit I get back from G.Skill. Bit of a hassle, but I can't be w/o my PC for three weeks waiting on the RMA.
 
Does anyone know if the new 6400 cl26 RAM that is for expo would work on a raptor lake CPU? Curious lol or for those that want it for AMD It could be an option also.
 
Are these good timings for my CL30 ram? Anything I can do to further improve performance?
 

Attachments

  • ZenTimings_Screenshot.png
    ZenTimings_Screenshot.png
    69.5 KB · Views: 0
You must log in or register to reply here.
Back
Top