Recommended voltages for 7800X3D

[Klemc]

What difference does it make?

It's broken games, that Intel/AMD could give better results bc of this reason (only)

CP 2077 is the best for battle the two CPU

What about last Hellblade II ?

And record gaming session in video, so that's not only bench numbers

 

[fevgatos]

Sure, what games do you want to test?

Shall we open a different thread or it's fine here?

 

[nguyen]

Shall we open a different thread or it's fine here?

I think it's fine here, but I'm going on a 3 days vacation tomorrow so you can posting some benchmarks first .

Here are list of games that I have + test at 1080p high using built in benchmark
CP2077
Shadow of the Tomb Raider
FF XXIV dawntrail benchmark
Horizon Zero Dawn
Total War Three Kingdoms
Total War Warhammer 3
The Talos Principle 2
Forza Horizon 5
Gears Tactics

I don't have TLoU nor Jedi Survivor :/

 

[Taraquin]

I think it's fine here, but I'm going on a 3 days vacation tomorrow so you can posting some benchmarks first .

Here are list of games that I have + test at 1080p high using built in benchmark
CP2077
Shadow of the Tomb Raider
FF XXIV dawntrail benchmark
Horizon Zero Dawn
Total War Three Kingdoms
Total War Warhammer 3
The Talos Principle 2
Forza Horizon 5
Gears Tactics

I don't have TLoU nor Jedi Survivor :/

I can compare aswell to show differences with ram only have CP and SOTTR on that list. But I get gpu bound eith my 4070. Can we test CP with dlss performance to see if I get cpu-bound? Tombraider we can watch cpu game avg/mins.

 

[fevgatos]

I think it's fine here, but I'm going on a 3 days vacation tomorrow so you can posting some benchmarks first .

Here are list of games that I have + test at 1080p high using built in benchmark
CP2077
Shadow of the Tomb Raider
FF XXIV dawntrail benchmark
Horizon Zero Dawn
Total War Three Kingdoms
Total War Warhammer 3
The Talos Principle 2
Forza Horizon 5
Gears Tactics

I don't have TLoU nor Jedi Survivor :/

I've already run some but not the ingame benchmarks. CP2077 ill send you a save when you are back, taraquin has already tested in the same area.

 

[Taraquin]

I've already run some but not the ingame benchmarks. CP2077 ill send you a save when you are back, taraquin has already tested in the same area.

But he may need dlss aswell. Only your 4090 is powerful enough to be cpu bound in some games using native

 

[fevgatos]

But he may need dlss aswell. Only your 4090 is powerful enough to be cpu bound in some games using native

Oh well, we don't need to run 1080p native. We can test at lowest possible resolution each game allows, maybe DLSS / FSR on top if it's available. I think cyberpunk will be mostly CPU bound in the area we tested, it hammers CPUs.

 

[Taraquin]

Oh well, we don't need to run 1080p native. We can test at lowest possible resolution each game allows, maybe DLSS / FSR on top if it's available. I think cyberpunk will be mostly CPU bound in the area we tested, it hammers CPUs.

If you remeber on my setup I had to lower res and use DLSS performance due to 4070 tanking compeltely when using pathtracing. 4090 is waaaaay stronger with RT.

CPU game fps are the ones that matter and shows how much cpu is capable of outputting.

 

[fevgatos]

If you remeber on my setup I had to lower res and use DLSS performance due to 4070 tanking compeltely when using pathtracing. 4090 is waaaaay stronger with RT.

View attachment 356421
CPU game fps are the ones that matter and shows how much cpu is capable of outputting.

That takes me back a couple of years. Was testing SOTR like crazy, very high + TAA (basically guru 3d's settings). I'm fairly sure i was getting exactly 320-330 fps at stock (obviously talking about CPU game). Dropping to high won't really make a difference, but you can try it out - see if very high drops your fps, I think it won't. I can get to around ~380 fps by going completely crazy with clocks - disabling ecores and boosting the cache to 5ghz but for sure those are not stable settings.

I don't have it currently so I'll download it again (I have steam version, which one are you using?) and see what's what. But yeah, the 7800x 3d is faster here for sure, I mean im around the 5800x 3d ballpark at stock. It's only the heavy core games that the 12900k ties or beats the 7800x 3d (once human, tlou, cyberpunk etc.)

 

[Taraquin]

Wait this is looking pretty funny
Cyberpunk 2077 1080p High
Core boost disabled, all cores 4.2ghz, CO -30
View attachment 356089

Core boost enabled, 65W PPT, CO -30
View attachment 356090

Not only all cores 4.2ghz gives higher avg FPS, but also much higher 1% Low FPS, about 15C cooler and 20W less

Cyberpunk 2077 probably doesn't care about clocks but rather the Cache

Edit: with Core Performance Boost disabled, CO disabled the performance is way way higher. Look like Curve Optimizer might be detrimental to performance
View attachment 356093

I see now you used framegen, try without and use dlss perfprmance to make totally cpu bound.

As for CO on/off:
Avg CO -35

No CO

Score is 18092 CO off, 18550 CO on.

CO off. 5C higher temp, 5% worse lows, 3% lower avg.

CO on:

I`m still gpu bound a few times in certain scenes, lol.

 

[Bloste]

You may want to turn your heatsink 180º: They are asymetrical, and as you have installed it's too close to the GPU. It will cach some heat from it. Turn it 180º, change the fans to blow in the correct direction, and test the temps. It should go down some degrees.

 

[Blaeza]

You lot are making me want to try this, but I don't know what resolution a MSi 1660 super ventus xs oc could bottleneck a 7800X3D lol.

 

[nguyen]

You may want to turn your heatsink 180º: They are asymetrical, and as you have installed it's too close to the GPU. It will cach some heat from it. Turn it 180º, change the fans to blow in the correct direction, and test the temps. It should go down some degrees.

The CPU fan is set as intake from the back actually, the top fan is exhaust that will pull both CPU+GPU heat upward.

 

[Bloste]

The CPU fan is set as intake from the back actually, the top fan is exhaust that will pull both CPU+GPU heat upward.

Not the fan, the hestsink is asymmetrical:

As you can see, it veers to one side. Right now the heatsink is too close to the backplate of the GPU, so some heat from it can jump to the CPU heatsink. If you rotate the heatsink 180º, it will increase the distance between them.

 

[nguyen]

Not the fan, the hestsink is asymmetrical:

View attachment 356437

As you can see, it veers to one side. Right now the heatsink is too close to the backplate of the GPU, so some heat from it can jump to the CPU heatsink. If you rotate the heatsink 180º, it will increase the distance between them.

Nice spotting, I was kinda annoyed that the heatsink sit so close to the backplate of the GPU haha.

 

[Taraquin]

Nguyen: can you rerun CP2077 without framegen? Interesting to see how much 4am imoacts vs my setup

 

[nguyen]

Nguyen: can you rerun CP2077 without framegen? Interesting to see how much 4am imoacts vs my setup

Just got back with a trainwreck of a body LOL, here ya go
720p Low settings with DLSS Ultra Performance, all Steam edition
Core Boost disabled (4200mhz clocks)


Stock (~4700mhz clocks) #1 run

#3 run

I don't quite like the idea that performance is reduced when the Cooler is heat soak causing lower boost. That means people in warmer climate will always get lower performance LOL. This 7800X3D sample doesn't react well with CO at all, -20 CO gives lower performance.

It's also very interesting to find that Frame Gen actually doesn't like the core clocks fluctuation at all

Bonus:
13700K Stock clocks with 7200MT Tuned DDR5

 

[DLAV]

How about per core CO? I would set all cores to zero, start with core 0 at minus 60. Once you know how low it can go, set it to zero and move to core 1 and so on.

 

[Taraquin]

Just got back with a trainwreck of a body LOL, here ya go
720p Low settings with DLSS Ultra Performance, all Steam edition
Core Boost disabled (4200mhz clocks)
View attachment 356735

Stock (~4700mhz clocks) #1 run
View attachment 356738
#3 run
View attachment 356737
I don't quite like the idea that performance is reduced when the Cooler is heat soak causing lower boost. That means people in warmer climate will always get lower performance LOL. This 7800X3D sample doesn't react well with CO at all, -20 CO gives lower performance.

It's also very interesting to find that Frame Gen actually doesn't like the core clocks fluctuation at all

Bonus:
13700K Stock clocks with 7200MT Tuned DDR5
View attachment 356741
View attachment 356742

It's strange that CO gives worse score. On my setup my -35 avg CO is always better. If we compare you see disabling boost lowers your score a lot. Try -10 CO? I bet maybe one or 2 cores are bad and gets throttling at -20 CO.

If we compare scores we get quite similar stock with some ramtuning, but I excel with CO. In shadow I get 60 more fps cpu game. Do you have boost disabled?

 

[nguyen]

It's strange that CO gives worse score. On my setup my -35 avg CO is always better. If we compare you see disabling boost lowers your score a lot. Try -10 CO? I bet maybe one or 2 cores are bad and gets throttling at -20 CO.

If we compare scores we get quite similar stock with some ramtuning, but I excel with CO. In shadow I get 60 more fps cpu game. Do you have boost disabled?

I didn't test SoTR on the 7800X3D (still need to copy it to my friend PC), that was with 13700K on my own PC

The boost clocks on 7800X3D is so temp sensitive that the results are rather erratic, having low ambient temperature and an AIO would stabilize the core clocks. Overall though I think my friend would enjoy the stable frame rates with Frame Gen enabled as he will be playing games at 4K.

 

[Taraquin]

I didn't test SoTR on the 7800X3D (still need to copy it to my friend PC), that was with 13700K on my own PC

The boost clocks on 7800X3D is so temp sensitive that the results are rather erratic, having low ambient temperature and an AIO would stabilize the core clocks. Overall though I think my friend would enjoy the stable frame rates with Frame Gen enabled as he will be playing games at 4K.

Try occt avx2 extreme. Run smudebugtool where you can set CO on the fly. You need to restart occt after chsnging CO. If 1 core needs -10 and the rest can do -30 you can get a long way reducing temps with per core CO As I showed earlier in this tread my CO boost fps by 3-5% while lowering temps by 5C in CP2077. One core needs -21, rest do -34 to -48. My avg CO is -37 I just calculated.

 

[fevgatos]

Can you try once human? I've seen that once you unlock the framerate it plays like absolute crap on the 7800x 3d. It pegs all cores to 100% and cause excessive stuttering. It's a free game

Spoiler

And this is how it plays on 12900, the game has a 240 framecap

Spoiler

 

[Taraquin]

Can you try once human? I've seen that once you unlock the framerate it plays like absolute crap on the 7800x 3d. It pegs all cores to 100% and cause excessive stuttering. It's a free game

Spoiler

And this is how it plays on 12900, the game has a 240 framecap

Spoiler

Nvm I can try later.

 

[fevgatos]

Nvm I can try later.

Just play past the tutorial, although with the 4070 your card will be the limit. For some reason the game is heavier when running 4k + 60% res scale than when you are running 1080p. And I mean heavier on the cpu.

 

[oxrufiioxo]

Just play past the tutorial, although with the 4070 your card will be the limit. For some reason the game is heavier when running 4k + 60% res scale than when you are running 1080p. And I mean heavier on the cpu.

On the 7950X3D it just uses one core at 100% every other core is 5-10% or lower sill hits 225-240fps when I drop to 90% gpu utilization through scaling To make sure I am cpu limited very high settings.

Games is DX11 apparently in 2024 so not surprised.