9800x 3d vs 12900k - Battle of the Century

[JustBenching]

So, since I'm "upgrading" to the 9800x 3d, im getting some footage together and ill end up doing a side by side video between 12900k and 9800x 3d. Depending on how much time I have, ill try to compare both stock and both OCed with different OSes. Sadly I had to restart the whole process a couple of times cause I figured out win 11 24h2 doesn't play nice with Intel, at all. Im seeing 20-25% performance difference between 24h2 and win 10.

If you got any game suggestions (no multiplayer games lads, they are really not easy to do) im all ears. I have already tested CP2077, TLOU, Hogwarts, Spiderman, Veilguard, Baldurs gate 3, Starfield , Kena, HZD. The settings im using are ULTRA + RT max on all games, resolution is 4k DLSS ultra performance or 1080p DLSS Ultra performance if the former is GPU bound. Games that are already GPU bound even at 1080p with Ultra Performance I just drop, there is no value in this comparison since youll need a 7090 to see any difference between the chips.

My 9800x 3d will arrive by Friday so hopefully ill get it done during the weekend. Some early teasers about what a tuned 12900k can do - same scene same setting as HUB (4k balanced - High preset)

 

[phanbuey]

Interested in your take on the smoothness factor. Hogwarts for me has more frame excursions with 9800x3d 6400mhz vs the 13700KF w/ 7600 c36 48GB ramkit -- nothing bad but the 13700 was smoother in this title. Remnant 2 is the other way around - the 13700KF 24H2 rode the struggle bus on that one and the 9800X3D feels awesome. Excited to see your results

 

[JustBenching]

Interested in your take on the smoothness factor. Hogwarts for me has more frame excursions with 9800x3d 6400mhz vs the 13700KF w/ 7600 c36 48GB ramkit -- nothing bad but the 13700 was smoother in this title. Remnant 2 is the other way around - the 13700KF 24H2 rode the struggle bus on that one and the 9800X3D feels awesome. Excited to see your results

Have you tried latest versions of hogwarts? It's butter smooth on 12900k and i assume it has to be the same on 9800x 3d. They improved performance a lot.

 

[phanbuey]

Have you tried latest versions of hogwarts? It's butter smooth on 12900k and i assume it has to be the same on 9800x 3d. They improved performance a lot.

This is the latest version -- it was butter smooth on the 13700KF like a week ago -- and stuttery af on the 64GB 6000 cl30 kit on the 9800x3d - swapping ram kits made it better but it's still not as buttery as it once was. It's definitely fine to play but if you look closely you can see a bit of jerk if as you're running around. On the 64GB ramkit it was actually unplayable - massive stuttering.

Which is odd - all the other games are crushing rn - PBO +200 im getting some monster fps in other titles. But Im like 50% done with HL and wanted to close it out.

 

[JustBenching]

This is the latest version -- it was butter smooth on the 13700KF like a week ago -- and stuttery af on the 64GB 6000 cl30 kit on the 9800x3d - swapping ram kits made it better but it's still not as buttery as it once was.

Which is odd - all the other games are crushing rn - PBO +200 im getting some monster fps in other titles. But Im like 50% done with HL and wanted to close it out.

Ok, ill check this one thoroughly. But reviews have the 9800x 3d sitting at the top blasting on everything else in this game.

 

[phanbuey]

I think it's my ram configuration - i think the old 64GB maybe too much with the dual rank? -- 48GB is way better, maybe a fast 32GB kit is what's needed for the optimal combo.

 

[JustBenching]

I think it's my ram configuration - i think the old 64GB maybe too much with the dual rank? -- 48GB is way better, maybe a fast 32GB kit is what's needed for the optimal combo.

Im gonna stick to my 2x16 Adies and try to hit 8k+ - if not possible I might consider going for the new 6000c28 kits. We shall see.

Are you overall happy - neutral - disappointed with 13700k ---> 9800x 3d? Do you notice the difference in MT workloads? That's my biggest fear, that's it's going to be very noticeable.

 

[oxrufiioxo]

Just commenting for the notifications.

Im gonna stick to my 2x16 Adies and try to hit 8k+ - if not possible I might consider going for the new 6000c28 kits. We shall see.

Are you overall happy - neutral - disappointed with 13700k ---> 9800x 3d? Do you notice the difference in MT workloads? That's my biggest fear, that's it's going to be very noticeable.

A borked windows 23H2/MT perfomance is the reason I swapped a 7800X3D for a 7950X3D...... Hoping the 9950X3D requires zero user intervention...

 

[JustBenching]

Just commenting for the notifications.



A borked windows 23H2/MT perfomance is the reason I swapped a 7800X3D for a 7950X3D...... Hoping the 9950X3D requires zero user intervention...

Repeating it here as well. After a lot of testing and retesting and clean installing windows.

Win 10

Ecores on = 90 fps
Ecores off = 96 fps

Win 11 24h2

Ecores on = 80 fps
Ecores off = 75 fps

Intel is borked and I was really wondering if I should stick with windows 10 for the 9800x 3d as well.

 

[phanbuey]

Im gonna stick to my 2x16 Adies and try to hit 8k+ - if not possible I might consider going for the new 6000c28 kits. We shall see.

Are you overall happy - neutral - disappointed with 13700k ---> 9800x 3d? Do you notice the difference in MT workloads? That's my biggest fear, that's it's going to be very noticeable.

Overall very happy -- the system is actually very snappy and most games have monster gains in FPS and run smoothly. The MT workloads are not noticeable at all to me - but I primarily focus on compile and data backend stuff which isn't like rendering. Things definitely open and load faster though, so that's nice. The VMs also run really smoothly, I was worried about virtualbox/python performance since our VMs runs kind of badly on our 7900x system, but 9xxx3D seems to have no lag or issues with it.

TBH the Zen 5 cores kind of perform like Alder Lake P cores, but instead of E cores they just put a giant swimming pool of cache.

 

[Frick]

Late game Dwarf Fortress. Late game Stellaris/EU4/CK3.

 

[JustBenching]

Late game Dwarf Fortress. Late game Stellaris/EU4/CK3.

I have no idea how to play these, if you cna share some save files so I can just load and test I can give it a go no problem.

Overall very happy -- the system is actually very snappy and most games have monster gains in FPS and run smoothly. The MT workloads are not noticeable at all to me - but I primarily focus on compile and data backend stuff which isn't like rendering. Things definitely open and load faster though, so that's nice. The VMs also run really smoothly, I was worried about virtualbox/python performance since our VMs runs kind of badly on our 7900x system, but 9xxx3D seems to have no lag or issues with it.

TBH the Zen 5 cores kind of perform like Alder Lake P cores, but instead of E cores they just put a giant swimming pool of cache.

That's good news actually.

 

[mb194dc]

Why run DLSS ultra performance in single player games? Just makes the visuals massively worse, makes no sense, is glorified 720p, lol, ? Unless you're struggling for frames I guess.

 

[phanbuey]

Why run DLSS ultra performance in single player games? Just makes the visuals massively worse, makes no sense, is glorified 720p, lol, ? Unless you're struggling for frames I guess.

To simulate what games that put more load on the CPU and faster graphics cards are going to look like.

 

[mb194dc]

To simulate what games that put more load on the CPU and faster graphics cards are going to look like.

Makes no sense, unless op really spent 3k on a system to run games with visuals from the PS2 era 20 years ago. I would be interested to see 9800x3D v 7950x with SMT off at realistic settings, given they're similar price now.

Op seems to have Dual QHD (5120x1440) screen, let's see it shine at native res.

 

[JustBenching]

Makes no sense, unless op really spent 3k on a system to run games with visuals from the PS2 era 20 years ago. I would be interested to see 9800x3D v 7950x with SMT off at realistic settings, given they're similar price now.

Op seems to have Dual QHD (5120x1440) screen, let's see it shine at native res.

Nah, sorry, it's an old signature. I don't have that monitor anymore.

I'm testing the cpu so I'm dropping the resolution.

 

[phanbuey]

Makes no sense, unless op really spent 3k on a system to run games with visuals from the PS2 era 20 years ago. I would be interested to see 9800x3D v 7950x with SMT off at realistic settings, given they're similar price now.

Op seems to have Dual QHD (5120x1440) screen, let's see it shine at native res.

makes perfect sense -- look at 5800x3d reviews from 2 years ago with a 3090 and now compare them to games running today with 4080/4090. Also the 9800X3D will annihilate a 7950X with SMT off in gaming in anything that isn't completely GPU bound.

Ryzen 7 9800X3D, Really Faster For Real-World 4K Gaming?
2022:

2024:

 

[Dr. Dro]

Im gonna stick to my 2x16 Adies and try to hit 8k+ - if not possible I might consider going for the new 6000c28 kits. We shall see.

Are you overall happy - neutral - disappointed with 13700k ---> 9800x 3d? Do you notice the difference in MT workloads? That's my biggest fear, that's it's going to be very noticeable.

6000 C28 with 1:1 internal ratios remain optimal path for 9800X3D IMO

 

[JustBenching]

6000 C28 with 1:1 internal ratios remain optimal path for 9800X3D IMO

Is it - even if you go above 8k? Not much tests about stuff like that so ill have to experiment a bit. Also there is the option of pushing to 6200-6400 with 1:1 imc (assuming i don't get a garbage bin).

AFAIK the FCLK doesn't depend on IMC / memclocks anymore right, so I can just run it as fast as my chip can handle?

 

[Dr. Dro]

Is it - even if you go above 8k? Not much tests about stuff like that so ill have to experiment a bit. Also there is the option of pushing to 6200-6400 with 1:1 imc (assuming i don't get a garbage bin).

AFAIK the FCLK doesn't depend on IMC / memclocks anymore right, so I can just run it as fast as my chip can handle?

It's the same as Zen 4. Going above 6200~ will necessarily have you reduce the FCLK/UCLK ratio to stay under the MHz limit, and this halves internal bandwidth causing a bottleneck. Ideally you should target 6200 to 6400~ with the lowest latencies achievable

https://www.reddit.com/r/overclocking/comments/18z4rm9
Much, much, much harder to go above with massively diminishing results. IMO - 6200 MT/s ~C28-C30 3100/3100/2167 is what you should target

 

[R-T-B]

Late game Dwarf Fortress. Late game Stellaris/EU4/CK3.

Late game high part count modded ksp as well, though, you'd need to know how to play it which is a big blocker.

 

[rv8000]

Is it - even if you go above 8k? Not much tests about stuff like that so ill have to experiment a bit. Also there is the option of pushing to 6200-6400 with 1:1 imc (assuming i don't get a garbage bin).

AFAIK the FCLK doesn't depend on IMC / memclocks anymore right, so I can just run it as fast as my chip can handle?

There is a latency penalty for odd fclk uclk ratios, it can be overcome by pushing fclk far enough.

6000:3000:2000
6200:3100:2066
6400:3200:2133
6600:3300:2200

Higher uclk requires more vsoc which makes fclk more unstable which is why 6600 1:1 is a unicorn status cpu.

Running 6400:3200:2200 may be faster than 2133 flck overall, but you run the risk of having unstable fclk and or requiring vddg tuning.

 

[JustBenching]

There is a latency penalty for odd fclk uclk ratios, it can be overcome by pushing fclk far enough.

6000:3000:2000
6200:3100:2066
6400:3200:2133
6600:3300:2200

Higher uclk requires more vsoc which makes fclk more unstable which is why 6600 1:1 is a unicorn status cpu.

Running 6400:3200:2200 may be faster than 2133 flck overall, but you run the risk of having unstable fclk and or requiring vddg tuning.

Oh, didn't know that about FCLK. I thought more faster = more better

I guess I need to do the tests and see what works.

 

[Frick]

I have no idea how to play these, if you cna share some save files so I can just load and test I can give it a go no problem.

Dwarf Fortress is the one likely to be best suited, the game is free and there are a bunch of saves avaliable, and there's a built in FPS counter... The game is more dependant on memory though I think, but it would still be interesting. Will try to find something this evening or tomorrow after work.

Also, plz post something here as well along with the video (just some numbers written down is enough).

Late game high part count modded ksp as well, though, you'd need to know how to play it which is a big blocker.

I forget how the game works, is it possible to save during something FPS-draining?

 

[rv8000]

Oh, didn't know that about FCLK. I thought more faster = more better

I guess I need to do the tests and see what works.

An average-above average chip should be able to do something like the following easily…

VSOC range will be roughly 1.2-1.3 depending on chip quality. Safe trfc for a-die is also significantly lower (120-130ns). Memory voltages will vary based on your kit.