9800X3D, hotter than the first day, pulling more wattage?

[Carlyle2020hs]

If you follow those stories a lot of asrock users mention having used Ryzen Master.
So maybe RM did/overwrote something?

As a sanity check how about:
De-install RM / load bios defaults and set it up anew / kindly share results ?

 

[rv8000]

What resolution do you run the game at? I'm running at 1440p and I just played with the framerate limited to 85 in the Nvidia control panel. The session resulted in 95.8C max on the 9800X3D and 51.3C average (although I forgot to reset max values before play I think) . The 4080S maxed out at 85.3C on the hotspot and 71.6C average.


Damn, I just did a repaste with PTM 7950 on my GPU recently. Sounds like Kryosheet might be in my future lol

Is there any reason to worry about Kryosheet destroying your system? I've seen people discussing the fact that it is conductive so any debris from the sheet could cause an electrical short.

1440p maxed minus AA and motion blur (if there is a setting for mb). Your temps seem fine tbh, outside of map load in slamming CPU package temp for 2-3 seconds (definitely a HD2 problem), I don’t think there’s much of an issue.

 

[A Computer Guy]

If you follow those stories a lot of asrock users mention having used Ryzen Master.

It wouldn't surprised. Ryzen Master defaults have been dumb and in the past found it to be a bit wonky where I had no choice in the past but to reset UEFI/BIOS back to defaults after messing with it. It's an interesting concept but I never found it able to consistently apply and test good Curve Optimizer values so it's kind of useless to me for reconfiguring the system and I don't trust it to work correctly.

I can only speculate that with these CPU blowing up Asrock simply got caught tweaking the motherboards like most of the vendors did back in AM4 days before HWiNFO64 provided a way to detect it. If I recall properly it was reported Asrock said something to the effect of TDC and EDC parameters needed to be adjusted. However if everyone else is using the standard PPT/TDC/EDC values by default for the CPU's why would CPU's only blow up for Asrock significantly more? If I was a Asrock 9000 cpu owner I'd use PBO to manually adjust TDC and EDC values downward a bit from default until it's confirmed Asrock has the problem solved.

 

[Pets4Ever]

I actually just did a build in a fractal north for a friend and I really liked it. Only a 240mm up top tho... and if you have front intake rad your GPU clearance is actually not great.

Meshify 3 is also nice but not sure about that front panel.

I helped my brother with his build in a Fractal North and it is a nice case. There is actually an XL version that allows for a 360mm rad on the top, but I would personally prefer it if there was a version with a metal side panel instead of glass for mesh. I don't like glass because then the second it becomes a little dusty it becomes an eyesore, and I'd feel more pressure to color coordinate parts.

Apparently the Meshify 3 has killer thermals but I agree it wouldn't be my first choice aesthetically. There is at least a no-rgb solid side panel variant if I decided to go that way though.

Got a chance to run some HD2. Average core temps were 66c, average package temp was 73c, peak package temp hit 95 ONLY when dropped from the hellpod loading into the mission, never came close at any other point. Max ppt hit 133w with no PBO and stock cpu.

So cores and package were below 73c for the entirety of a 33m missile launch mission. Only running a single fan on my NH-D15 at max rpm (for memory cooling).

The 133w peak sounds identical to the results I got last night after wiping my PBO settings. Thanks for the sanity check! I run two fans on the NH-D15 with PWM so they boost higher if the temp gets higher. Is there an advantage to just running one?

Just a thought...If you running hotter and pulling more watts then your thermal paste might be settling in and working better. This could be a good opportunity to increase mounting pressure when the paste is hot and tweak power limits to run cooler assuming your thermal transfer has improved since your initial installation.

Does the Noctua NH-D15 include the AM5 offset mounting kit? If so then I would use it. It's proven on my non-X3D's to be useful on my modest NH-D9L and NH-C14S coolers.

Also it seems PBO + Curve optimizer is the way to go to really tune Ryzen CPU's. Since they hit thermal limits before others the trick is to manage heat production with power limits and curves to get the best performance aside from having a good cooler and well placed cooler contact across the chiplets.

I don't think there is any more adjustment left on the heatsink as far as making it tighter. The version I have did not come with an offset kit, it's technically an AM4 compatible cooler, but Noctua advises it is compatible with AM5 too.

If you are hitting high temps (which you are) during a CPU only load like cinebench then its not your case airflow that's the problem.

I guess I was thinking that if the AIO weren't performing as expected in my case that maybe the top grating could be a factor. Part of me feels like taking the case apart and dremeling out sections of this grate...

If you follow those stories a lot of asrock users mention having used Ryzen Master.
So maybe RM did/overwrote something?

As a sanity check how about:
De-install RM / load bios defaults and set it up anew / kindly share results ?

Steve of Gamers Nexus talked to ASROCK about the failures and IMO it wasn't all that enlightening.

1440p maxed minus AA and motion blur (if there is a setting for mb). Your temps seem fine tbh, outside of map load in slamming CPU package temp for 2-3 seconds (definitely a HD2 problem), I don’t think there’s much of an issue.

Thanks for helping me confirm that there is nothing wrong with my parts. I will endeavor to get this system to a point where it doesn't hit 95C ever in games though, my apartment isn't going to be getting cooler this summer and I don't like 90C for an average in CB either.

It wouldn't surprised. Ryzen Master defaults have been dumb and in the past found it to be a bit wonky where I had no choice in the past but to reset UEFI/BIOS back to defaults after messing with it. It's an interesting concept but I never found it able to consistently apply and test good Curve Optimizer values so it's kind of useless to me for reconfiguring the system and I don't trust it to work correctly.

I can only speculate that with these CPU blowing up Asrock simply got caught tweaking the motherboards like most of the vendors did back in AM4 days before HWiNFO64 provided a way to detect it. If I recall properly it was reported Asrock said something to the effect of TDC and EDC parameters needed to be adjusted. However if everyone else is using the standard PPT/TDC/EDC values by default for the CPU's why would CPU's only blow up for Asrock significantly more? If I was a Asrock 9000 cpu owner I'd use PBO to manually adjust TDC and EDC values downward a bit from default until it's confirmed Asrock has the problem solved.

Yea, Asrock has commented but I didn't think anything they said was very enlightening. (see video further up this post) Some people claim they had these failures when running stock, but you never know who is telling the truth online

 

[rv8000]

I helped my brother with his build in a Fractal North and it is a nice case. There is actually an XL version that allows for a 360mm rad on the top, but I would personally prefer it if there was a version with a metal side panel instead of glass for mesh. I don't like glass because then the second it becomes a little dusty it becomes an eyesore, and I'd feel more pressure to color coordinate parts.

Apparently the Meshify 3 has killer thermals but I agree it wouldn't be my first choice aesthetically. There is at least a no-rgb solid side panel variant if I decided to go that way though.


The 133w peak sounds identical to the results I got last night after wiping my PBO settings. Thanks for the sanity check! I run two fans on the NH-D15 with PWM so they boost higher if the temp gets higher. Is there an advantage to just running one?


I don't think there is any more adjustment left on the heatsink as far as making it tighter. The version I have did not come with an offset kit, it's technically an AM4 compatible cooler, but Noctua advises it is compatible with AM5 too.


I guess I was thinking that if the AIO weren't performing as expected in my case that maybe the top grating could be a factor. Part of me feels like taking the case apart and dremeling out sections of this grate...

View attachment 402474


Steve of Gamers Nexus talked to ASROCK about the failures and IMO it wasn't all that enlightening.

Thanks for helping me confirm that there is nothing wrong with my parts. I will endeavor to get this system to a point where it doesn't hit 95C ever in games though, my apartment isn't going to be getting cooler this summer and I don't like 90C for an average in CB either.


Yea, Asrock has commented but I didn't think anything they said was very enlightening. (see video further up this post) Some people claim they had these failures when running stock, but you never know who is telling the truth online

There’s no advantage to running 1 fan, 2 will provide better temps; my ram is tall and I cannot fit the fan on either side of the HSF due to the vrm/io cover and ram on each respective side.

From the sound of discussion it seems investing in a better case/better airflow case is going to help you more than anything. I firmly believe buying an AIO is a waste of money unless you want it for looks, the temp difference is going to be minimal, but thats just an opinion.

Are you going to be playing CB on a daily basis ? CB temps shouldn’t be a concern.

If you’re serious about squeezing more performance from your system, look at tuning your IF and memory timings/freq, much more benefit focusing there.

 

[Pets4Ever]

There’s no advantage to running 1 fan, 2 will provide better temps; my ram is tall and I cannot fit the fan on either side of the HSF due to the vrm/io cover and ram on each respective side.

From the sound of discussion it seems investing in a better case/better airflow case is going to help you more than anything. I firmly believe buying an AIO is a waste of money unless you want it for looks, the temp difference is going to be minimal, but thats just an opinion.

Are you going to be playing CB on a daily basis ? CB temps shouldn’t be a concern.

If you’re serious about squeezing more performance from your system, look at tuning your IF and memory timings/freq, much more benefit focusing there.

Ah gotcha. I limited myself to shorter ram sticks to deal with the clearance issue but there is a set of DDR5-6000 26CL that would fit if I had an AIO instead of a tower cooler. The kit only comes with tall heat spreaders and RGB unfortunately.

Well, I want my computer to be able to handle sustained loads without getting 90°C average temps. If the AIO isn't able to help with that then it goes back I guess.

CB is not all that crazy of a test, it's not like PRIME95 or something. My 12700k got lower 80s for temps when running the multicore and I'm more comfortable with that extra headroom. My apartment is window AC only, so this summer it will get warmer than it is now. Seems like it'd be better for longevity to stay away from the max temp too.

 

[phanbuey]

Ah gotcha. I limited myself to shorter ram sticks to deal with the clearance issue but there is a set of DDR5-6000 26CL that would fit if I had an AIO instead of a tower cooler. The kit only comes with tall heat spreaders and RGB unfortunately.

Well, I want my computer to be able to handle sustained loads without getting 90°C average temps. If the AIO isn't able to help with that then it goes back I guess.

CB is not all that crazy of a test, it's not like PRIME95 or something. My 12700k got lower 80s for temps when running the multicore and I'm more comfortable with that extra headroom. My apartment is window AC only, so this summer it will get warmer than it is now. Seems like it'd be better for longevity to stay away from the max temp too.

just limit your PBO to whatever you're comfortable with - but really longevity isn't really a concern for these -- raptor lake aside. The AMD 7845HX i've had in my laptop basically idles at 80C and lives life at 92C all day, every day, and will do so for years -- that paste is probably chalk by now.

Also, I was running a flat multiplier on this b650, didn't realize it disables the safeties, and ran my 9800X3D at 118 C AVX-512 for a solid 30 seconds before shutting off the stress. It survived lol. So i think there's some headroom to these even at 92-95C.

Also that ram is amazing - that's going to be some great performance.

 

[Pets4Ever]

just limit your PBO to whatever you're comfortable with - but really longevity isn't really a concern for these -- raptor lake aside. The AMD 7845HX i've had in my laptop basically idles at 80C and lives life at 92C all day, every day, and will do so for years -- that paste is probably chalk by now.

Also, I was running a flat multiplier on this b650, didn't realize it disables the safeties, and ran my 9800X3D at 118 C AVX-512 for a solid 30 seconds before shutting off the stress. It survived lol. So i think there's some headroom to these even at 92-95C.

118°C, whew toasty! I know the hardware is fairly heat resistant, but I think I've read that running hotter would speed up the natural degradation of the chip over time. It might not be drastic or super noticeable unless you are measuring, but still.

Well I hate to limit it or I might lose the performance I paid for. Interestingly, this review seems to indicate that the NH-D15 can handles AMD loads up to 243 watts. Even given an open-air test bench this seems crazy. I'll have to read over the setup and methodology sections.

Arctic Liquid Freezer III Pro 360 A-RGB Review

Arctic is expanding its AIO liquid cooler collection with the Liquid Freezer III Pro 360 A-RGB. The cooler boasts a wider RPM range for its radiator fans, A-RGB lighting, and a proven high-performance design. You could say they took an already excellent product and made it better!

www.techpowerup.com

 

[JustBenching]

118°C, whew toasty! I know the hardware is fairly heat resistant, but I think I've read that running hotter would speed up the natural degradation of the chip over time. It might not be drastic or super noticeable unless you are measuring, but still.

Well I hate to limit it or I might lose the performance I paid for. Interestingly, this review seems to indicate that the NH-D15 can handles AMD loads up to 243 watts. Even given an open-air test bench this seems crazy. I'll have to read over the setup and methodology sections.

Arctic Liquid Freezer III Pro 360 A-RGB Review

Arctic is expanding its AIO liquid cooler collection with the Liquid Freezer III Pro 360 A-RGB. The cooler boasts a wider RPM range for its radiator fans, A-RGB lighting, and a proven high-performance design. You could say they took an already excellent product and made it better!

www.techpowerup.com

The review isnt using a single ccd cpu like thr 9800x 3d though. The d15 can definitely handle 240w on a dual ccd ryzen, but the 9800x 3d isnt that. Realistically nothing can cool an 8 core zen when you are pushing 150w. Heck even intel parts will suffer. If i disable my ecores and just push 200w on just the 8pcores ill get 100c.

 

[phanbuey]

This guy is hilarious.

 

[Pets4Ever]

The review isnt using a single ccd cpu like thr 9800x 3d though. The d15 can definitely handle 240w on a dual ccd ryzen, but the 9800x 3d isnt that. Realistically nothing can cool an 8 core zen when you are pushing 150w. Heck even intel parts will suffer. If i disable my ecores and just push 200w on just the 8pcores ill get 100c.

Ohhh that make sense. I didn't think of that.

This guy is hilarious.

Haven't watched yet, just read the title. Have you heard that Thermal Grizzly is selling pre-delidded 9800X3Ds with a warranty? 60% markup though lol

 

[Carlyle2020hs]

Twice the contact area with only 6 instead of 8 cores in a CCD
That tested 7900X is an overclocking beast, that´s rather easy to cool.


As already stated 9800x3d will struggle, IF let loose.

Even at the best of times:

No expenses spared:

still:
90c+ are possible IF you want the last ounce of power quenched out (CB23 24.000-25.000+ points)

I´m only using air cooling since i don´t have AC and thus a real low ambient noise level that makes any AIO pump very audible.

But i don´t advertise air cooling for tinkerers that can´t run the chip on auto

 

[JustBenching]

Ohhh that make sense. I didn't think of that.


Haven't watched yet, just read the title. Have you heard that Thermal Grizzly is selling pre-delidded 9800X3Ds with a warranty? 60% markup though lol

The best i could do on a 5800x with a flattened aio, T30 fans to 100% speed and liquid metal was around 180w. Seeing how the 9800x 3d has more concentrated heat the fact that you can get to 140 is a miracle all things considered.

 

[phanbuey]

Ohhh that make sense. I didn't think of that.


Haven't watched yet, just read the title. Have you heard that Thermal Grizzly is selling pre-delidded 9800X3Ds with a warranty? 60% markup though lol

yeah - for 0 percent FPS improvement... lol. Crazy.

 

[Pets4Ever]

Is a contact frame useful for a 9800X3D? I thought those were necessary with intel mainly because of their rectangular shape and subpar rigidity in the default mounting system.

yeah - for 0 percent FPS improvement... lol. Crazy.

Haha yea the price is even funnier listening to this video

 

[JustBenching]

I disagree with that, the fact that 9800x 3d isnt as dependant on ram makes it scale really well with frequency in gaming performance. So if you don't mind the extra cost or the extra power draw that comes with it a delided 9800x 3d running at 5.6 is actually naughty.

 

[rv8000]

Is a contact frame useful for a 9800X3D? I thought those were necessary with intel mainly because of their rectangular shape and subpar rigidity in the default mounting system.


Haha yea the price is even funnier listening to this video

Afaik the contact frame isn’t useful or necessary. I think someone working for/from ASUS did say they saw better memory related performance when pushing memory clocks, but I don’t think thats ever been verified/proven.

 

[phanbuey]

So I found something interesting on the 9800x3D....

1. Run superpi on a stock install.

2. Enable hardware shadow stacks, and run again. You will see a 10-15% boost in speed. I dropped to 7.1-7 seconds flat from 7.7 at the same clocks. This also enables VBS that will crush your VM performance and lower fps in some games but still super interesting.

According to the AI:

• Shadow Stack offloads return address validation from software to hardware
• Disables or reduces overhead from StackGuard, CFG, or legacy EMET-style shims
• Results in fewer jumps, cleaner call stack
• Enabling CET may trigger new CPU microcode paths
• May optimize speculative return branch prediction
• May unlock hidden instruction optimizations
• SuperPi is memory latency-sensitive
• CET changes stack layout slightly → affects TLB/cache locality
• Could yield better L1/L2 behavior in tight loops
• CET-enabled builds may change kernel scheduler heuristics
• Slight improvement in single-thread priority & responsiveness

I feel like the AMD scheduler tweaks that they were parading around a few months ago are reliant on CET somehow. Windows definitely feels more responsive with hardware shadow stacks on.

 

[Dorek]

Well, I wonder. Dorek up there says he only hits 80°C with the same cooler, very similar setup. Maybe mine is a bad sample? I don't see how the installation could be wrong, it's so straightforward.


Do you know what your ambient temps might be? I can only guess on mine but I'm thinking 23-24°C last night


You have the rad acting as an intake? What is your exhaust setup then? Doesn't that hurt your GPU temps?

Ambients around 22 so normal but after bios update it seems like temps gone up on the cpu at least on the hotspot is about 49 celsius but otherwise seems okay.

 

[Pets4Ever]

I found an article by MSI where they tested AIO positions in a case. I'm somewhat surprised that Top Mount Intake and Front Intake Tubes Down have such different results. I would have expected those two to be roughly equivalent.

 

[JustBenching]

I found an article by MSI where they tested AIO positions in a case. I'm somewhat surprised that Top Mount Intake and Front Intake Tubes Down have such different results. I would have expected those two to be roughly equivalent.

View attachment 402662

Its highly case dependent. If your case has an exhaust rear fan then you can put your AIO as top exhaust. If it doesn't have a rear exhaust fan then put it front intake. That's my 2 cents after testing and retesting different cases with different coolers.

 

[A Computer Guy]

I found an article by MSI where they tested AIO positions in a case. I'm somewhat surprised that Top Mount Intake and Front Intake Tubes Down have such different results. I would have expected those two to be roughly equivalent.

View attachment 402662

One thing to consider is the front intake avoids recirculating the rear warmer exhaust assuming it's not being deflected back to the front by a wall or something.

 

[phanbuey]

I found an article by MSI where they tested AIO positions in a case. I'm somewhat surprised that Top Mount Intake and Front Intake Tubes Down have such different results. I would have expected those two to be roughly equivalent.

View attachment 402662

Their AIOs have the pump in the radator, so that may have something to do with it. Tubes up has air bubble implications and rad up also has air bubble implecations - so you basically want toe bubble up top in side the rad - but fill the channels as much as possible otherwise.

But you're right in a watercooling system (custom WC loop with a rez) it wouldn't make a difference.

EDIT: nvm - the new models have the pump in the block. I think the older ones used to have a pump in the rad.

 

[Pets4Ever]

Their AIOs have the pump in the radator, so that may have something to do with it. They should normally be about the same.

Oh interesting, did not know that.

 

[Jtuck9]

One thing to consider is the front intake avoids recirculating the rear warmer exhaust assuming it's not being deflected back to the front by a wall or something.

This seems to be my boggle with tempered glass side panelling and something like how the 9070XT Nitro+ cards exhaust hot air. Might be a moot point if you have sufficient airflow from the front?!