Discussion AM5 7950X build - first impressions

CPU - Ryzen 9 7950X (FCLK at 2167MHz)
MB - GIGABYTE B650E AORUS MASTER (default PBO)
RAM - 2x16GB Corsair 6200MT/s DDR5 CL36 (running at CL32)
GPU - RX 6900 XT (1125mV 2600MHz core 2100MHz VRAM w/fast timings)
NVMe SSD - 2TB Silicon Power XS70 PCIe 4.0, 2TB ADATA SX8200 Pro PCIe 3.0
PSU - EVGA 1000 P6 Platinum
Case - NZXT H510 Flow
Cooling - be quiet! Dark Rock TF 2 CPU cooler, 2x NZXT F140Q high airflow fans

Couldn't get system to post with BIOS that came with motherboard (memory is on QVL). Had to do the no CPU/RAM BIOS flashback to get the latest BIOS on. After that and some quirkiness with the RAM (memory power down mode must be enabled - seems to be a common theme) system has been perfectly stable. I'm still tweaking and adjusting but very happy with the performance thus far. I purposely went with this air cooler as a separate experiment. So far I've been pleasently surprised. If I bench/stress test the CPU it does get up to 95ºC and throttle, but throughout normal usage (inclidung hours of gaming) no throttling is seen.

I'm still tweaking/adjusting settings. Will come back to this thread with any other quirkiness I discover along the way.

View: https://imgur.com/a/2vGpOMy


View: https://imgur.com/a/85kHCeD


View: https://imgur.com/a/LftR8TX
 
Looks good! Get an exhaust fan in there, though.
Any exhaust fan would just fight for air with the be quiet!. Those high airflow fans up front are pushing a TON of air into the case. When the system is being put through its paces the hot airflow out the back and up top are as if I had a low RPM fan there anyway.

I've done tons of testing on this. It probably wouldn't hurt CPU temps much, if at all, but it would just add extra noise for absolutely no benefit. Fans fighting for the same air create chaotic vortices. Vortices are the opposite of what you want with good air cooling. Yes, 90+% of those custom builds you see with 6+ fans are done more for RGB show than airflow efficiency.
 
Last edited:
  • Like
Reactions: alceryes
I don't know if you've seen this: https://www.anandtech.com/show/17641/lighter-touch-cpu-power-scaling-13900k-7950x

If you lower the PPT, there's a potential loss of maybe 4-5% in CPU bound tasks, but they're using a 360mm AIO cooler. I'd wager on air it should be able to sustain higher clock speeds than the default setting because it's not trying to pull so much power for the same clock speed.
Yup, seen it. Tech YES City also has a good video showing the crazy amount of power that last 5% of performance requires -
View: https://www.youtube.com/watch?v=7JiYAwKIHRY

I'm still tweaking/playing around. ;)
 
  • Like
Reactions: hotaru.hino

jnjnilson6

Distinguished
CPU - Ryzen 9 7950X (FCLK at 2167MHz)
MB - GIGABYTE B650E AORUS MASTER (default PBO)
RAM - 2x16GB Corsair 6200MT/s DDR5 CL36 (running at CL32)
GPU - RX 6900 XT (1125mV 2600MHz core 2100MHz VRAM w/fast timings)
NVMe SSD - 2TB Silicon Power XS70 PCIe 4.0, 2TB ADATA SX8200 Pro PCIe 3.0
PSU - EVGA 1000 P6 Platinum
Case - NZXT H510 Flow
Cooling - be quiet! Dark Rock TF 2 CPU cooler, 2x NZXT F140Q high airflow fans

Couldn't get system to post with BIOS that came with motherboard (memory is on QVL). Had to do the no CPU/RAM BIOS flashback to get the latest BIOS on. After that and some quirkiness with the RAM (memory power down mode must be enabled - seems to be a common theme) system has been perfectly stable. I'm still tweaking and adjusting but very happy with the performance thus far. I purposely went with this air cooler as a separate experiment. So far I've been pleasently surprised. If I bench/stress test the CPU it does get up to 95ºC and throttle, but throughout normal usage (inclidung hours of gaming) no throttling is seen.

I'm still tweaking/adjusting settings. Will come back to this thread with any other quirkiness I discover along the way.

View: https://imgur.com/a/2vGpOMy


View: https://imgur.com/a/85kHCeD


View: https://imgur.com/a/LftR8TX
That's a really nice setup you've got there!

If it were nearly 12 years ago the same system would harbor an i7-990X and Radeon HD 6990.

Hope you've a lot of fun with this monster and that its electric, zapping capabilities are entrancing and that the fans roar on indefinitely and that it lasts many years on the pinnacle of high-end hardware.

I remember the i7-990X to have been Godlike when it came out. I was really in awe of its cores and power, yet I did get an i7-3770K in 2012. And I did manage to overclock it up to 5 GHz stable using Corsair H110. At this speed it performed a little faster than the i7-3930K (6 cores / 12 threads). Which also means that it was stronger than the i7-990X with the overclock as well.

Basically when you reach about 18% CPU usage on the Ryzen 9 7950X that would be like hitting 100% on an i7-990X.
 
  • Like
Reactions: alceryes
Thanks!
The CPU fans (there's two on the Dark Rock TF 2) are always running at 100% RPM and are whisper quiet. I have the front fans ramp up with the CPU temp so, at stock PBO I do hear them, but they're never annoying, just a slight increase whooshing of air. As @hotaru.hino pointed out, PPT will be my next stop with adjustments/tweaks. (y)
 
  • Like
Reactions: jnjnilson6

jnjnilson6

Distinguished
Thanks!
The CPU fans (there's two on the Dark Rock TF 2) are always running at 100% RPM and are whisper quiet. I have the front fans ramp up with the CPU temp so, at stock PBO I do hear them, but they're never annoying, just a slight increase whooshing of air. As @hotaru.hino pointed out, PPT will be my next stop with adjustments/tweaks. (y)
The other day I ran a Windows Defender Full Scan. My Core i7-12700H got to about 100% for 20 minutes. Temperatures started rising though and I suppose the system automatically lowered the CPU usage to about 10-20%. I checked with NitroSense (software for Acer Nitro laptops which displays the fan speed and CPU and GPU temperatures) and the CPU temp was stuck at 92 C. I suppose the tension on the CPU was thus lowered from 100% down to 10-20% because the machine had to maintain that 92 C and not go higher.
 
The other day I ran a Windows Defender Full Scan. My Core i7-12700H got to about 100% for 20 minutes. Temperatures started rising though and I suppose the system automatically lowered the CPU usage to about 10-20%. I checked with NitroSense (software for Acer Nitro laptops which displays the fan speed and CPU and GPU temperatures) and the CPU temp was stuck at 92 C. I suppose the tension on the CPU was thus lowered from 100% down to 10-20% because the machine had to maintain that 92 C and not go higher.
Laptops are notorious for having very thin thermal envelopes, but this still doesn't sound right. I could see it throttle down to 40-60% after getting really hot but 10-20%?? That's too much.
 
  • Like
Reactions: jnjnilson6
There you go! PPT set to a manual 125W in BIOS and only lost 3-4% in TIme Spy CPU score.

View: https://imgur.com/a/6z7Sy1V


It just feels wrong to type in a number this large when dealing with system voltage! :D
It does say mV in the lower left though. Sure enough, 125000 is the correct setting to limit the PPT to 125W.

View: https://imgur.com/a/TwUk5vc


Super heavy CPU benchmarks/loads will suffer more (Cinebench R23 loses around 10%), but the literal 100W power reduction is phenomenal!!!
NOTE - These settings are specific to my particular system. Do NOT just grab settings you see here and use them in your system or components may go pop!!
 
Last edited:

jnjnilson6

Distinguished
Laptops are notorious for having very thin thermal envelopes, but this still doesn't sound right. I could see it throttle down to 40-60% after getting really hot but 10-20%?? That's too much.
Perhaps you're right! It might have to do with the software of Windows Defender too. I was in bed though, so I suppose the cooling couldn't kick-in as big a punch as it would've been able to on a desk (where cooling space would have been better).
The temperature (92 C) was continually the same for hours while Defender worked, so I suppose that had the percentages of CPU usage been higher the temperature would have risen as well. I am almost certain that keeping in mind the fact the temp did not variate but remained at 92 C the usage I was getting was the maximum possible in order that aforesaid temperature could be maintained without escalation.

Otherwise, as I am not using the machine for gaming, I am very happy with it; everything seems to work smoothly and at incredible speeds and heretofore all the components and capabilities are of high quality and there isn't really anything I could point my finger at and mention as a con. Really a great machine; the highest class laptop I've ever owned.
 

jnjnilson6

Distinguished
Laptops are notorious for having very thin thermal envelopes, but this still doesn't sound right. I could see it throttle down to 40-60% after getting really hot but 10-20%?? That's too much.
I was thinking that since an i7-12700H at 19% would perform the same way an i7-3770K would maxed out at 100%, it is not unreasonable going down to 10-20% usage on the i7-12700H to maintain agreeable temperatures. It's just that the CPU is too powerful for even the best cooling of a laptop. For example, as you've said, laptop processors do throttle down to maintain good temperatures, yet if a Quad Core would go down to about 60% on a laptop, an Octa Core may go down to 40% and the Core i7-12700H -> 10-20% (but in the end the i7-12700H's performance at the aforementioned percentages would be equal to or higher than that of the Quad and Octa core CPUs providing higher usage). I think the best explanation is that the processor is too powerful and produces too much heat to remain at high usage throughout long periods of time. And again, keeping in mind that an i7-12700H at 19% is like a 100% i7-3770K, I am more than satisfied.

The average CPUs don't have usage decreased so drastically because they are slower and produce less heat. As I've said, a more powerful processor like the i7-12700H may go down to 10-20% due to the fact it produces more heat, yet that would not actually make it slower than the average CPU which would produce the same amount of heat at 50% which the i7-12700H does at 20%.
 
Last edited:
I was thinking that since an i7-12700H at 19% would perform the same way an i7-3770K would maxed out at 100%, it is not unreasonable going down to 10-20% usage on the i7-12700H to maintain agreeable temperatures. It's just that the CPU is too powerful for even the best cooling of a laptop. For example, as you've said, laptop processors do throttle down to maintain good temperatures, yet if a Quad Core would go down to about 60% on a laptop, an Octa Core may go down to 40% and the Core i7-12700H -> 10-20% (but in the end the i7-12700H's performance at the aforementioned percentages would be equal to or higher than that of the Quad and Octa core CPUs providing higher usage). I think the best explanation is that the processor is too powerful and produces too much heat to remain at high usage throughout long periods of time. And again, keeping in mind that an i7-12700H at 19% is like a 100% i7-3770K, I am more than satisfied.

The average CPUs don't have usage decreased so drastically because they are slower and produce less heat. As I've said, a more powerful processor like the i7-12700H may go down to 10-20% due to the fact it produces more heat, yet that would not actually make it slower than the average CPU which would produce the same amount of heat at 50% which the i7-12700H does at 20%.
No. 10-20% is not normal for a laptop's CPU to throttle down to, unless maybe you were chocking performance in a battery saver mode(?) Sitting the laptop on top of blankets definitely makes the cooling solution worse.
 
  • Like
Reactions: jnjnilson6

jnjnilson6

Distinguished
No. 10-20% is not normal for a laptop's CPU to throttle down to, unless maybe you were chocking performance in a battery saver mode(?) Sitting the laptop on top of blankets definitely makes the cooling solution worse.
Yeah, you're right. There are a few options.

  1. Being in bed.
  2. The software may have simply stopped extracting such high load from the CPU on its own.
-- I was on Balanced mode with the screen on the entire time.

As sarcastically impudent as it may seem, I am generally using the computer for the same things I used my Core i3-8130U. That would fall in the gap of Office use and entertainment (not concerning video games; I've stopped those for a long time, yet do not feel I should tell other people whether to play or read books or do something else, because they're entitled to their own opinion). The only more serious thing I've done was start up a Virtual Machine and install Windows 11 on it which happened almost instantaneously keeping in mind the SSD and CPU power. I used to dabble deep within the hardware sphere before a little less than a decade now; currently the ghosts of different affiliations have slanted my perceptions toward different, multicolor things. Call it artist's passion, but my computer usage has fallen down an immense crest and landed spunkily into the void of hardly any tension upon the components and hardly any tension at all.

It is exhilaratingly daring to run innumerable Windows and programs continually in light and exultant moods as the 12700H cuts through the cycles with incalculable speed. In the end, that's what I got it for. :)
 
There you go! PPT set to a manual 125W in BIOS and only lost 3-4% in TIme Spy CPU score.

View: https://imgur.com/a/6z7Sy1V


It just feels wrong to type in a number this large when dealing with system voltage! :D
It does say mV in the lower left though. Sure enough, 125000 is the correct setting to limit the PPT to 125W.

View: https://imgur.com/a/TwUk5vc


Super heavy CPU benchmarks/loads will suffer more (Cinebench R23 loses around 10%), but the literal 100W power reduction is phenomenal!!!
NOTE - These settings are specific to my particular system. Do NOT just grab settings you see here and use them in your system or components may go pop!!
Interesting side effect to setting a static PPT...?

Purely anecdotal, but I'm running through the same parts of a game (Witcher 3 next-gen) that I initially ran through when I didn't have a PPT limit set. Previously, with no PPT limit, I saw several cores on the first CCD peak above 70% max usage and most of the cores on the second CCD never peak above 20%. Now, with a 125W PPT limit set, I'm seeing more even usage across all the cores of both CCDs. The lowest core peak usage on any of the 16 cores is 39%!! Setting a lower PPT seems to change the workload behavior across all the cores in addition to setting a power ceiling. As far as performance goes, I couldn't tell any difference between the two. Both are buttery smooth.

If anyone else has a 7900X or 7950X and doesn't mind testing it would be interesting to see if this is normal behavior. I actually like that the workload (burst job threads) at least appear to be more spread out across the whole CPU.
 

Brian D Smith

Prominent
Mar 13, 2022
116
68
660
I don't know if you've seen this: https://www.anandtech.com/show/17641/lighter-touch-cpu-power-scaling-13900k-7950x

If you lower the PPT, there's a potential loss of maybe 4-5% in CPU bound tasks, but they're using a 360mm AIO cooler. I'd wager on air it should be able to sustain higher clock speeds than the default setting because it's not trying to pull so much power for the same clock speed.

I saw that article. I wonder...at what point in lower power could one go with fans only to cool the CPU? 65W? 105W? That would be good to know for someone like myself who might not want to do liquid cooling.
 
I saw that article. I wonder...at what point in lower power could one go with fans only to cool the CPU? 65W? 105W? That would be good to know for someone like myself who might not want to do liquid cooling.
It depends on what you're willing to put up with. If you're willing to work with a Nocuta NH-D15, which is one of the best air coolers, it can handle an i9-13900K. But you have to be weary of all the clearance issues it presents. But if you want something tamer, like an Arctic Freezer 34 eSports Duo, then you're looking at nothing more powerful than a Core i7 or a Ryzen 7.

It also depends on what value helps you sleep at night. If you're after something like 55-60C, then you're going to have to knock the CPU tier down a notch and maybe even more. I run a Freezer 34 with my Ryzen 5600X and it floats around 60-65C if I leave it alone with a Handbrake job on a reasonably quiet fan profile.
 
With my Dark Rock TF 2, and a 125W PPT set, I see my 7950X peak in the mid-80s C. Without any PPT set it would march right up to 95C and start throttling - but only when benchmarking/stress testing. Note that at 125W, I only lose 3-4% peak CPU gaming performance. This DOESN'T mean that I am automatically getting 3-4% lower FPS. I'm only talking peak gaming performance. This CPU chews through any game I've thrown at it with ease, so far. Since I play at 4K I'm gonna be GPU limited anyway.

Remember, these chips are designed to work at 95C all the time (24x7x365). Depending on your workload, a very good air cooler would suffice just fine. If you want it to NEVER hit 95C then you'll need to power limit the CPU or get a good 280+mm AiO.
 
Remember, these chips are designed to work at 95C all the time (24x7x365). Depending on your workload, a very good air cooler would suffice just fine. If you want it to NEVER hit 95C then you'll need to power limit the CPU or get a good 280+mm AiO.
To add to this bit, Zen 4 CPUs are designed to hit the thermal limit first, rather than the power limit. Previous Zen generations were designed to hit the power limit first.

I'm convinced AMD did this only to make their product look slightly better than Intel's, since apparently Intel's CPUs since Alder Lake operate on the same principle as Zen 4.
 
UPDATE -
I am experiencing firsthand how sensitive AMD chips are to memory timings and subtimings. My last AMD chip was a Bulldozer-era CPU. AMD has come a LONG way since then. With a near optimal setting for memory timings and subtimings, I'm seeing a 6-7% increase in my Time Spy CPU score. AIDA64 shows memory bandwidth has also increased dramatically and latency is now below 58ns.

To put this into perspective, adjusting all the memory timings and subtimings to near optimal settings has improved CPU and memory performance twice as much as increasing the FCLK from 2000 to 2167 AND the MCLK from 3000 to 3100 (6000 to 6200MT/s). It's crazy how much good timings increase performance.

Of course, CPU tasks that don't rely on the memory much won't show as much of an improvement but going from a Time Spy CPU score of 17637 to 18763 just by adjusting memory timings...I'm kinda at a loss for words! :D

Edit - the difference in performance was with timings and subtimings changes other than CL. CL was set to 32 previously and is still at 32.
 
Last edited: