Poll: Ryzen 7950X3D, 7900X3D, 7800X3D

[Grim5]

These are the caches for each CPU. The 7800X3D has zero advantages here. The 7950X3D can run 3D cache programs like a 7800X3D can but also has another CCD that can provide 8 cores to support non 3D cache processes. It’s far more versatile being able to support 3D cache and 16 cores at the same time.

That's an optimistic view. You're assuming windows is going to do a great job when it comes to assigning threads to tasks and games efficiently. Where as the 7800x3d might have less cache it doesn't have to worry about core to core latency of windows does a bad job, being that it only has a single ccd. While I don't have the answers, I think it curious that the 7800x3d was delayed after the two other models, it makes me wonder if the 7800x3d is the fastest for games but AMD wants to sell higher core count chips to make more money

 

[jinxpad]

what do you think the odds are we'll see v-cache on multiple ccd's in the AM5 life cycle?

 

[Iron Man]

what do you think the odds are we'll see v-cache on multiple ccd's in the AM5 life cycle?

As long as the majority of games only use 8 cores then there will be no reason for AMD to make both ccds have V-cache. AMD have said they tested having both ccds with V-cache but it added extra cost and didn’t benefit gaming performance.

 

[Bencher]

what do you think the odds are we'll see v-cache on multiple ccd's in the AM5 life cycle?

It's pointless. They better spend the time to work with microsoft on the scheduler, that might give them more benefit than a 2nd vcache ccd. Probably work with developers as well to make games - besides taking advantage of the cache, maybe also taking advantage of the extra clocks at the 2nd ccd for threads that need it.

Im very interested to see MSFS performance. This thing would fly on that game

 

[jinxpad]

interesting indeed, will be interesting to see how the gaming performance turns out then. I'm leaving performance on the table it seems for gaming with my current setup.

 

[Deleted member 258511]

That's an optimistic view. You're assuming windows is going to do a great job when it comes to assigning threads to tasks and games efficiently. Where as the 7800x3d might have less cache it doesn't have to worry about core to core latency of windows does a bad job, being that it only has a single ccd. While I don't have the answers, I think it curious that the 7800x3d was delayed after the two other models, it makes me wonder if the 7800x3d is the fastest for games but AMD wants to sell higher core count chips to make more money

I think it’s obvious at this point that the 7800x3d is the one to go for. What worries me is the lack of any sort of “bragging and boasting”. It’s gone very quiet since the reveal which was a bit of a **** show let’s be honest.

I think the 7950x is going to be much closer in performance than expected.

 

[Iron Man]

I think it’s obvious at this point that the 7800x3d is the one to go for. What worries me is the lack of any sort of “bragging and boasting”. It’s gone very quiet since the reveal which was a bit of a **** show let’s be honest.

I think the 7950x is going to be much closer in performance than expected.

I think the 7950x3D performance in gaming will be mostly identical and especially if there is no frequency boost for the ccd v-cache cores over the 7800x3D. You will be able to have bragging rights over being an early adopter over 7800x3D users though which is something I’m looking forward to

 

[delta0]

I think the 7950x3D performance in gaming will be mostly identical and especially if there is no frequency boost for the ccd v-cache cores over the 7800x3D. You will be able to have bragging rights over being an early adopter over 7800x3D users though which is something I’m looking forward to

They will be close for games that are optimised for the 3D v-cache. For games that aren’t then there will be a significant difference. Where there will be an edge even when running on the 3D v-cache is being able to run processes on the other CCD at the same so you have processes running on the CCD they are optimised for. It’s also entirely possible there is a frequency difference as you mention. We’ll soon know.

 

[Quixote]

My experience of mixed use systems, ones where you can do two things at once i.e. video rendering and gaming, is that sometimes you can get a combination of programs that do weird things which don't get shown in benchmarks. I had an experience of this last year when I built a system with 12700 for a friend.

When doing some stability test on it though I would take the opportunity to compare it to my 3900X with a lengthy handbrake transcode. Everything was great with the 12700 when I set it going and it gave excellent performance, better than my 3900X. However, all that changed when I opened chrome to browse OCUK. Handbrake was immediately moved to the E cores, causing the performance to crater. Chrome was using a tiny fraction of the P cores which were all now mostly at idle. As soon as I closed Chrome Handbrake shifted back to the P cores. It made a system with a similar CPU completely unusable for what I need to do with my system. One the other hand the person that I built it for could not be happier with it as they don't run the same combination of programs that I do.

My main concern with the configuration of the dies in these CPUs is that we will get interactions between programmes both wanting the cores on the die with the extra cache and the scheduler not being smart enough to know if a the program actually needs the extra cache. It will probably lead to some very wired performance issues that won't show though in just pure single program benchmarks.

I was more meaning using the system for different things at different times, rather than at the same time.
So a system you'd do production work on, then kick back & do some gaming at another time.
You're right though & there will surely be edge cases with odd performance, especially at the start.
I've long noticed there can be some architectural differences with CPUs under heavy load. How much they seem to leave for general system responsivness in those circumsances.

 

[Deleted member 258511]

I think the 7950x3D performance in gaming will be mostly identical and especially if there is no frequency boost for the ccd v-cache cores over the 7800x3D. You will be able to have bragging rights over being an early adopter over 7800x3D users though which is something I’m looking forward to

Me too haha.

 

[GoogalyMoogaly]

I'm sure I don't really need the 7950X3D and would be fine with the 7800X3D but I'm impatient and basically waiting on these CPUs to do my build. So I'll probably end up getting the 7950X3D for that reason and I'm aware that's almost certainly why AMD are staggering the releases like this.
It annoys me that I'm going to "fall" for it, but it is what it is.

Unless the reviews aren't great and then I might go back to my Intel plan.

 

[junpit]

I'll be staying with my 5800X3D. I don't like how AMD are only selling the expensive chip first and are holding the 7800X3D back.

 

[jinxpad]

could you effectively use something like Process Lasso to assign the v-cached cores to be the only ones used when gaming on the 7900/7950 X3D chips?

 

[Joxeon]

I'll be staying with my 5800X3D. I don't like how AMD are only selling the expensive chip first and are holding the 7800X3D back.

Probably to segment the reviews, if they released all 3 at the same time then reviewers would probably say for anyone gaming just to get the 7800X3D. Now however the reviews for these wont include the 8 core so they'll get judged on performance rather than value for money.

 

[delta0]

Probably to segment the reviews, if they released all 3 at the same time then reviewers would probably say for anyone gaming just to get the 7800X3D. Now however the reviews for these wont include the 8 core so they'll get judged on performance rather than value for money.

There could be many reasons. They may not have enough chips to supply both at once. They may be binning chips that run at higher frequency and building up a stock of chips that can’t clock as high (like they are using 6 core chips for the 7900 which clearly are chips with 2 failed cores).

 

[junpit]

I just think it’s simply profiteering. They know many people want to game on the 3D and won’t wait to own one so will pay top money.

 

[cliffy]

I just think it’s simply profiteering. They know many people want to game on the 3D and won’t wait to own one so will pay top money.

Shock horror, a company trying to make a profit to satisfy their share holders, whatever next

 

[GoogalyMoogaly]

Shock horror, a company trying to make a profit to satisfy their share holders, whatever next

To be fair people seem rather annoyed when Nvidia (or Intel) do that.

I'm sure there are use-cases where the 7950X3D makes sense, but I'm sure there will be lots of people gaming (like myself) that won't be getting the full benefit due to mostly gaming (or currently sitting there looking at my games library unable to decide what to play). That does mean I'll be paying more than needed, but hopefully I'll be able to use the chip for a good long while (and hope that gaming starts to use more than 16 threads).

 

[BigBANGtheory]

Shock horror, a company trying to make a profit to satisfy their share holders, whatever next

dividends presumably, followed by an AGM and a private meeting with select shareholders so they can fix the market with Nvidia without getting caught

 

[Poneros]

My experience of mixed use systems, ones where you can do two things at once i.e. video rendering and gaming, is that sometimes you can get a combination of programs that do weird things which don't get shown in benchmarks. I had an experience of this last year when I built a system with 12700 for a friend.

When doing some stability test on it though I would take the opportunity to compare it to my 3900X with a lengthy handbrake transcode. Everything was great with the 12700 when I set it going and it gave excellent performance, better than my 3900X. However, all that changed when I opened chrome to browse OCUK. Handbrake was immediately moved to the E cores, causing the performance to crater. Chrome was using a tiny fraction of the P cores which were all now mostly at idle. As soon as I closed Chrome Handbrake shifted back to the P cores. It made a system with a similar CPU completely unusable for what I need to do with my system. One the other hand the person that I built it for could not be happier with it as they don't run the same combination of programs that I do.

My main concern with the configuration of the dies in these CPUs is that we will get interactions between programmes both wanting the cores on the die with the extra cache and the scheduler not being smart enough to know if a the program actually needs the extra cache. It will probably lead to some very wired performance issues that won't show though in just pure single program benchmarks.

Interesting, and that's actually part of my mixed use too (run handbrake in background). Is it not possible to manually assign (and lock) cores for each program?