• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Ryzen Owners Zen Garden

I really am impressed with the speed boost over AM4.. and you can bet your patootie that I will be grabbing Zen 6 :)
 
I really am impressed with the speed boost over AM4.. and you can bet your patootie that I will be grabbing Zen 6 :)
If it really comes with 12 cores per CCD, and can just as awesomely be fine-tuned as Zen 4 and 5, then I'll probably grab one, too (my wallet just started crying as I typed this). :D

I'd love to see 24 cores beating under my LP cooler. I'm perverted like that. :D

---

More testing was done.

The test was watching my performance and CPU clock speed while playing Kingdom Come Deliverance 2 while running BOINC/WCG in the background.

Setup 1: 45 W TDP (61 W PPT), Expo disabled, RAM at 4800 MHz, SoC at 1.0 V (auto).
Setup 2: 45 W TDP (61 W PPT), Expo enabled, RAM at 6000 MHz, SoC at 1.3 V (auto).

Setup 1 had the CPU clock speed pegged at 4.3 GHz.
Setup 2 had the CPU clock fluctuate between 3.8 and 3.9 GHz.

Game performance was exactly the same in both situations.
 
Last edited:
This past week I've been trying to tune the memories (4x64gb CUDIMM 6400MHz CL52 from kingston) on my 9950x+x670e pro art a bit further.
I had already managed to achieve 4800MHz 42-42-42 stable from previous testings, with VDDSOC=1.2V, and VDDIO=VDD=VDDQ=1.3V

No matter what, I could not get 5000 or 5200MHz to work properly whatsoever. I even tinkered with the terminations, mostly using random values that I saw from other setups, to no avail. Since I'm using the iGPU, it was clear when thing booted up and were not stable due to all the graphics artifacts going on, and tests such as y-cruncher and stressapptest failed really quick.

Only thing I managed to do was to manually set the FCLK to 2000MHz, and lower the VDDIO/VDD/VDDQ voltages to 1.25V instead of 1.3V.

At the current 4800MHz 42-42-42 settings I can manage ~50GB/s on intel mlc (65% of the theoretical 76.8GB/s bw that 4800MHz should achieve).
For comparison, my 5950x with 3200MHz manages 37~42.5GB/s on MLC (83% of the theoretical 51.2GB/s bw that 3200MHz should achieve).
Those numbers actually seem to be on par with the results from chipsandcheese:
1746634763931.png


Instead of trying to push the memory frequencies further, I'll try to see if I can tune timings a bit more and hope that this improves the memory bandwidth.
So far trying out 4800MHz 28-39-39 made the system not even post, and I stopped at that since I got too lazy to clear the CMOS :p

I've also seen talks that a FCLK out of sync may reduce perf, I'll see if changing this improves memory bandwidth in any way.
 
I really am impressed with the speed boost over AM4.. and you can bet your patootie that I will be grabbing Zen 6 :)
I can confirm this moving from a 5900X to 9900X3D is amazing in the boost in performance.

This past week I've been trying to tune the memories (4x64gb CUDIMM 6400MHz CL52 from kingston) on my 9950x+x670e pro art a bit further.
I had already managed to achieve 4800MHz 42-42-42 stable from previous testings, with VDDSOC=1.2V, and VDDIO=VDD=VDDQ=1.3V

No matter what, I could not get 5000 or 5200MHz to work properly whatsoever. I even tinkered with the terminations, mostly using random values that I saw from other setups, to no avail. Since I'm using the iGPU, it was clear when thing booted up and were not stable due to all the graphics artifacts going on, and tests such as y-cruncher and stressapptest failed really quick.

Only thing I managed to do was to manually set the FCLK to 2000MHz, and lower the VDDIO/VDD/VDDQ voltages to 1.25V instead of 1.3V.

At the current 4800MHz 42-42-42 settings I can manage ~50GB/s on intel mlc (65% of the theoretical 76.8GB/s bw that 4800MHz should achieve).
For comparison, my 5950x with 3200MHz manages 37~42.5GB/s on MLC (83% of the theoretical 51.2GB/s bw that 3200MHz should achieve).
Those numbers actually seem to be on par with the results from chipsandcheese:
View attachment 398560

Instead of trying to push the memory frequencies further, I'll try to see if I can tune timings a bit more and hope that this improves the memory bandwidth.
So far trying out 4800MHz 28-39-39 made the system not even post, and I stopped at that since I got too lazy to clear the CMOS :p

I've also seen talks that a FCLK out of sync may reduce perf, I'll see if changing this improves memory bandwidth in any way.
It is not a slight but CL52 sounds like the bottom of the stack. Team have some 64GB on sale that are 6000 30 for like $200 Cad.
 
It is not a slight but CL52 sounds like the bottom of the stack. Team have some 64GB on sale that are 6000 30 for like $200 Cad.
It was one of the first 64GB sticks that became available in the market, it doesn't even have any EXPO/XMP whatsoever.
Too late to change it now, I'll try to make the best out of those sticks since I paid way too much for them and had to go through a hole saga do get them in my hands haha
 
I've also seen talks that a FCLK out of sync may reduce perf, I'll see if changing this improves memory bandwidth in any way.
depends on what memory frequency you are running, I am on my 9700X with 2000 FCLK with 8000MT's ram CL36 and it better than my 1:1 setting (6200MT's CL28) both tunes have their Nitro setup tuned as well to reduce the latency, both also uses L2 tweak, the 2:1 is doing better overall with better lows than the 1:1 tune..
 
depends on what memory frequency you are running, I am on my 9700X with 2000 FCLK with 8000MT's ram CL36 and it better than my 1:1 setting (6200MT's CL28) both tunes have their Nitro setup tuned as well to reduce the latency, both also uses L2 tweak, the 2:1 is doing better overall with better lows than the 1:1 tune..
I'll likely be running MEMCLK=UCLK since I'm at low frequencies (4800MHz), just need to figure out if leaving the IF at 2000MHz is better than keeping it in sync at 1600MHz, bandwidth-wise.
I don't really mind latency that much, bandwidth is more relevant for me.
 
It was one of the first 64GB sticks that became available in the market, it doesn't even have any EXPO/XMP whatsoever.
Too late to change it now, I'll try to make the best out of those sticks since I paid way too much for them and had to go through a hole saga do get them in my hands haha
Understood. I started at 5200 with 32GB when I went jumped into AM5. I was looking at around the time the 99X3D chips launched for some new RAM. I have seen modern Games are much more demanding. As I said I found 64GB of Team 30 for around $200. You only need the best RAM to satisfy penis envy but having 64GB is much more ideal than people realize,
 
I'll likely be running MEMCLK=UCLK since I'm at low frequencies (4800MHz), just need to figure out if leaving the IF at 2000MHz is better than keeping it in sync at 1600MHz, bandwidth-wise.
I don't really mind latency that much, bandwidth is more relevant for me.
as far as my test went it has/had no impact (going higher FCLK with lower mem clocks and let uclk sync), the more you let is sync (lower FCLK) the performance just tanks, it will tank more performance if its a dual CCD chip.
 
Understood. I started at 5200 with 32GB when I went jumped into AM5. I was looking at around the time the 99X3D chips launched for some new RAM. I have seen modern Games are much more demanding. As I said I found 64GB of Team 30 for around $200. You only need the best RAM to satisfy penis envy but having 64GB is much more ideal than people realize,
My jump to AM5 was mostly due to the possibility of using 4x64GB, since my 4x32GB AM4 setup was holding me back in many different tasks.
Having AVX-512 with Zen 5 is also a really nice bonus.
 
I kinda feel unimpressed.. :nutkick:

I feel ripped..

Max tuned 9950X3D (5.7ghz 3D Vcache cores + 5.9ghz non-X3D + Crosshair Hero X870E + Latest win11 + DDR5 8200MT's/DDR5 6600MT's + 2200 FCLK) and I am still losing FPS vs my max tuned i9.. :roll:

Screenshot 2025-05-19 013711.png
 
I really am impressed with the speed boost over AM4.. and you can bet your patootie that I will be grabbing Zen 6 :)
Wonder that should I get a 9700X myself.
 
Wonder that should I get a 9700X myself.
Sure why not? Get yourself setup for Zen 6.. I paid less overall for this rig then I did for my AM4 rig.. crazy..
 
Sure why not? Get yourself setup for Zen 6.. I paid less overall for this rig then I did for my AM4 rig.. crazy..
Even a 7700X would probably be more then enough (hell, even my current 5800X is fine) so I doubt that I need a Zen6 CPU (at least not instantly in its launch), but 9700X with increased power limit should be a decent upgrade. :)
 
I kinda feel unimpressed.. :nutkick:

I feel ripped..

Max tuned 9950X3D (5.7ghz 3D Vcache cores + 5.9ghz non-X3D + Crosshair Hero X870E + Latest win11 + DDR5 8200MT's/DDR5 6600MT's + 2200 FCLK) and I am still losing FPS vs my max tuned i9.. :roll:

View attachment 400259
Is that benchmark still buggy as hell? The last time I used it, it produced vastly different results on the same system within two consecutive runs. It seems to suffer from some kind of memory overflow issue (unless it's been fixed).
 
The Intel setup is running 2R ram.. maybe try that ram in your X3D system..
 
Is that benchmark still buggy as hell? The last time I used it, it produced vastly different results on the same system within two consecutive runs. It seems to suffer from some kind of memory overflow issue (unless it's been fixed).
its not just out of 2 runs, I tried disabling the 2nd CCD, parking cores manually, force scheduling on the X3D CCD and everything, when I do 1:1 (UCLK=MCLK) the results are far worse..
The Intel setup is running 2R ram.. maybe try that ram in your X3D system..
2x24GB sticks aren't DR or 2 ranked..

2x24GB isn't even stable on the Crosshair Hero.. (Claps to Asus junk) the 2x24GB sticks are only stable on my B650E Aorus Tachyon bench jig with the same chip..

I personally thought, I'd gain anything (even with just 1 bit) on the move, well I ended up wrong, even the lows are horrendous now, on Cyberpunk on the same GPU OC, the i9 averages around 84 FPS on the bench with RT on (PT off) this one is just around 79-80 FPS and on actual gameplay FPS is jiggling around..lmao..

Temps aren't even an issue, I only peak around 43c while gaming with this setup (i9 was just around 36-38c with the same loop and 6Ghz P-Cores/4.6Ghz E-Cores with HT), there were no telemetry or any power monitoring stuff around, as I only use the Steam FPS overlay.

EDIT:

I think I am done for today..
Screenshot 2025-05-19 222242.png
 
Last edited:
Better than AsRock junk haha.

Ahh well.

Yeah, I just saw the Taichi X870E do 8000 on a G.Skill 2x24GB kit..and just when things couldn't get worse, they announced the Taichi OCF X870E.. (facepalm)
 
Yeah, I just saw the Taichi X870E do 8000 on a G.Skill 2x24GB kit..and just when things couldn't get worse, they announced the Taichi OCF X870E.. (facepalm)
Ahh, I see what you are saying. I was looking to the not so distant past.
 
I've had my 5800X3D for about a year now, upgraded from a 5600 on an MSI PRO B550M-VC WIFI motherboard. Also have 32GB DDR4-3600 CL16. I know that the 5800X3D is much more limited than the 7800X3D or 9800X3D in terms of PBO or overclocking. Is there anything else I can do to improve performance, or is just leaving it at stock performance good? To be clear, I'm very happy with the performance I have, so I'm not looking to improve a specific game or get 3D Mark records or anything like that. Just looking for little tips and tricks that would improve things.
 
I've had my 5800X3D for about a year now, upgraded from a 5600 on an MSI PRO B550M-VC WIFI motherboard. Also have 32GB DDR4-3600 CL16. I know that the 5800X3D is much more limited than the 7800X3D or 9800X3D in terms of PBO or overclocking. Is there anything else I can do to improve performance, or is just leaving it at stock performance good? To be clear, I'm very happy with the performance I have, so I'm not looking to improve a specific game or get 3D Mark records or anything like that. Just looking for little tips and tricks that would improve things.
Welcome to Dell life. You can tweak your memory a bit, B-Die is best.. welcome to retirement life, as most forms of tuning are dead with that CPU. You can set your curve to get it to boost a bit better and run a touch cooler.
 
I've had my 5800X3D for about a year now, upgraded from a 5600 on an MSI PRO B550M-VC WIFI motherboard. Also have 32GB DDR4-3600 CL16. I know that the 5800X3D is much more limited than the 7800X3D or 9800X3D in terms of PBO or overclocking. Is there anything else I can do to improve performance, or is just leaving it at stock performance good? To be clear, I'm very happy with the performance I have, so I'm not looking to improve a specific game or get 3D Mark records or anything like that. Just looking for little tips and tricks that would improve things.
2025-05-19 18_55_59-PBO2 Tuner.png
And for the details, you may use search in this thread. Apart from the RAM tuning, that's all that 5800X3D have and need.
 
Back
Top