Poll: Ryzen 7950X3D, 7900X3D, 7800X3D

[alec]

If you turn off one of the CCD's on the 7950X

look at 7700X being so high, basically answers it
And that with 200MHz clock disadvantage vs "single-ccd 7950X"

But Doom is an outlier. Game that wants frequency and not many threads, but still threaded enough to go to ther CCD and start causing issues

 

[LtMatt]

but neither are cool.

7950X3D runs 60c or lower for me in games, normally low-mid 50s. 360MM AIO as intake (O11 Dynamic chassis with side panel off) with liquid metal on the CPU. 7950X was a little higher mind you but not by much using the same setup.

Warzone 2 Al Mazrah | 7950X3D 7900 XTX Nitro | 1440P Competitive Settings | Smart Access Memory Off

The 7950X3D and 7900 XTX Nitro average 324 FPS at 1080P, with 1% lows of 200 with a render count of 6 with Smart Access Memory disabled. 7950X3D PBO Auto OC ...

www.youtube.com

 

[GoogalyMoogaly]

7950X3D runs 60c or lower for me in games, normally low-mid 50s. 360MM AIO as intake (O11 Dynamic chassis with side panel off) with liquid metal on the CPU. 7950X was a little higher mind you but not by much using the same setup.

Warzone 2 Al Mazrah | 7950X3D 7900 XTX Nitro | 1440P Competitive Settings | Smart Access Memory Off

The 7950X3D and 7900 XTX Nitro average 324 FPS at 1080P, with 1% lows of 200 with a render count of 6 with Smart Access Memory disabled. 7950X3D PBO Auto OC ...

www.youtube.com

I thought they were supposed to run at 89ºC?

60ºC is a lot better than I was expecting, but I'm not sure I'd describe it as cool.
I also have friends that tell me their 13700k run around that sort of temp in games and I believe the 13900k runs cooler than the 13700k, so probably not a lot in it still.

 

[JGF111]

Yes true if only it was actually available at MSRP and not ebay scalper prices it is only $250 more and probably better binned 3D cache CCD as it is out of stock everywhere.

Since the 7800X3D is best match for your needs and you have run out of patience, do you plan to disable the 2nd CCD to make it exactly the same as a 7800X3D?

No definitely not. About the most fiddling I will do would be to manually assign affinity for any software that shows an egregious mismatch between core allocation.

To be fair the 13900k is probably a reasonable consideration too. Competes in all area (games & productivity) despite having 16 e-cores and is cheaper (than the 7950X3D). Probably runs a little hotter, but neither are cool.

You also forgot option 4:
4. How big do you need your e-peen?

The 13900K was definitely under heavy consideration for sure. But ultimately I was swayed by two things:

1. the 7950's efficiency (and yes I understand the 13900K can be more aggressively power managed). I am spending my power budget on the 4090 so the cpu being extra efficient is important to me.
2. the allure of being able to drop in a brand new cpu in 4 or 5 or 6 years time. I have to admit, seeing people drop a 5800X3D into a 4+ year old system has been pretty impressive... even if I never do this myself (I tend to be a long hauler these days), the ability or option to do so is definitely compelling.

 

[LtMatt]

I thought they were supposed to run at 89ºC?

60ºC is a lot better than I was expecting, but I'm not sure I'd describe it as cool.
I also have friends that tell me their 13700k run around that sort of temp in games and I believe the 13900k runs cooler than the 13700k, so probably not a lot in it still.

89c is the max temperature before throttling.

Not sure if you were aware of the recent Nvidia driver bug that caused 5-10% cpu utilisation of a program called Nvidia container? Well when that bug was a thing my 7950X3D would run the same game at 80c+.

It's fixed now luckily so temps back to normal and obviously not an issue for AMD GPU users.

 

[crazyswede]

look at 7700X being so high, basically answers it
And that with 200MHz clock disadvantage vs "single-ccd 7950X"

But Doom is an outlier. Game that wants frequency and not many threads, but still threaded enough to go to ther CCD and start causing issues

You are right, the 7700X answers it. Would have been interesting to have seen "single-ccd 7950X" in that table. I need to have a look at the whole review.

I think I am like a lot of others having a hard time deciding between... 7800X3D, 7950X3D or 7950X. Arghhh

 

[GoogalyMoogaly]

No definitely not. About the most fiddling I will do would be to manually assign affinity for any software that shows an egregious mismatch between core allocation.



The 13900K was definitely under heavy consideration for sure. But ultimately I was swayed by two things:

1. the 7950's efficiency (and yes I understand the 13900K can be more aggressively power managed). I am spending my power budget on the 4090 so the cpu being extra efficient is important to me.
2. the allure of being able to drop in a brand new cpu in 4 or 5 or 6 years time. I have to admit, seeing people drop a 5800X3D into a 4+ year old system has been pretty impressive... even if I never do this myself (I tend to be a long hauler these days), the ability or option to do so is definitely compelling.

The ability to drop in a new CPU also seems nice to me, but like you I don't upgrade very often. I'll be coming from a 3930K which I got near release. So I'm not sure I'll be buying a new CPU in 4 or 5 years time. I'm not sure if my motherboard will support the new CPUs that far down the line (just because it's AM5 doesn't guarantee it will support every AM5 cpu right?). There's also the aspect that if you're spending £750 on a CPU are you really so budget contious that the cost of a motherboard is what would stop you upgrading (although motherboard prices are no joke these days either, GPU and CPU prices get criticised but motherboards seem to have flown under the radar a bit).

The thing with the 5800X3D as I see it is that it was so good because it came at the end of the AM4 lifecycle and there was nothing like it before. It was a lot better than everything that came before it. That's not going to be the case on AM5, the X3D chips have been released with the first generation of chips, so if the last generation of chips gets one it might not be such a big leap. Imagine if we'd had a 1800X3D, a 2700X3D and a 3800X3D. Would the 5800X3D have been as impressive?

 

[KompuKare]

The issue is also for any application which might like 3D V-Cache too and they are expecting MS to solve this for them. Not even Intel trusted MS to do a decent job so included hardware scheduling. Did MS do a good job with the Bulldozer type design - nope.

This is another self created AMD problem. Intel has that problem because they can't get away from being stuck on what it essentially a 10NM process node,so needs all these janky core configurations so they can compete in some way. AMD end up making a janky core configuration to save a few dollars. The 3D V-Cache this generation is smaller and cheaper than the first generation job - the Ryzen 9 3D V-Cache prototypes had cache on both chiplets.

But the thing as usual AMD gets a pass on this,but Intel didn't. Wait until the Ryzen 9 8950X3D has cache on both chiplets and everyone will forget about this.

I still think Intel have far more problems than their process node.

Yes, possibly because they had projected (their foundry people told their CPU designer people that they'd have x density by year y) they went too big and wasted transistors but basically the P cores are far too big to be efficient.

If server buyers weren't so conservative ("nobody ever got fired for buying for buying Intel"), then Intel would be in a lot more trouble.

I think that once Intel's designers realised that the density which Intel foundry promised the designers wasn't going to get there in time (at all?), they panicked and at that stage they couldn't undo what they had done with the post Skylake mobile cores so they decided to revive Atom cores and do a hybrid design. Both because making a 16-core P cores monster would be cost Intel too much with low yields and a huge die and because even if they ate the huge die margin loss, it would run too hot and with lower clocks it would still lose most benchmarks anyhow..

Intel's big.LITTLE copy is all about saving Intel wafer space and not really the same thing as the original ARM version.

I too think that AMD have been too cheap with a single 3D stacked CCD. A dual 3D stacked part would have lost a bit of ST performance due to no higher ST boost, but would have aged a lot better.

So both are a compromise. Having had to suffer Intel Hybrid nonsense at work with the at times painfully slow Alder Lake laptop I think AMD's compromise is the better one though.

 

[GoogalyMoogaly]

You are right, the 7700X answers it. Would have been interesting to have seen "single-ccd 7950X" in that table. I need to have a look at the whole review.

I think I am like a lot of others having a hard time deciding between... 7800X3D, 7950X3D or 7950X. Arghhh

It's a pity that none of us are considering the 7900X3D.
Feels like that was a bit of a misstep. Has had a similar impact as the 7900XT in that it's sort of left people asking "why?". Again, maybe it's a price issue on that part, maybe if it was much closer to the 7800X3D it might get a bit more traction. Seems like it'll be worse in game but better in productivity (which we already have thanks to chips like the 7950X).

 

[LtMatt]

It's a pity that none of us are considering the 7900X3D.
Feels like that was a bit of a misstep. Has had a similar impact as the 7900XT in that it's sort of left people asking "why?". Again, maybe it's a price issue on that part, maybe if it was much closer to the 7800X3D it might get a bit more traction. Seems like it'll be worse in game but better in productivity (which we already have thanks to chips like the 7950X).

It doesn't help that there are very few reviews of it, mostly because it wasn't sampled at launch. Toms3D have a review up though, one of the few online for the 7900X3D. https://www.tomshardware.com/reviews/amd-ryzen-9-7900x3d-cpu-review

Spoiler

 

[GoogalyMoogaly]

It doesn't help that there are very few reviews of it, mostly because it wasn't sampled at launch. Toms3D have a review up though, one of the few online for the 7900X3D. https://www.tomshardware.com/reviews/amd-ryzen-9-7900x3d-cpu-review

Spoiler

I think GamersNexus did one too, although they came to the conclusion "No" despite it seemingly not performing that poorly.

 

[Bencher]

I think I've found a good example to correct this long running myth once and for all.

I get your point, however that's not true as I've said before.

Modern Warfare 2 (MW2) using the default out of the box configuration, requests use of all 16 cores of the 7950X3D as it uses a default CPU render worker thread count of 16 for 16 core CPUs.


Now, based on your assumption, the workload would be stuck on the first CCD (cache) and the second CCD (frequency) will be asleep.

Here's a video of MW2 using the default out of the box configuration for 16 core CPUs. Note the workload spread over all 16 cores.

Now, I've edited the MW2 config to make it use 6 render threads instead of 16.


Here's a video of MW2 requesting 6 CPU worker threads. Note the workloads are now pinned to CCD0 because the game is not requesting to use more than 8 cores.

The workload is not stuck on one CCD, it is spread out and both are active during the workload if the game wants to use more than 8 cores. So to say that 7950X3D when gaming is stuck to one CCD only is false. However, most games do not request more than 8 cores for workloads. Most games work best with 8 or less CPU worker threads. The games that actually benefit from more are few and far between. Even the example I've now used, MW2, performs better if you override the default settings and change the worker count to 6 or 7. Of course most games won't let you change this behaviour, it's hard coded but the point remains.


I could be wrong about Doom then. I may have been thinking of a different game.

But you are still not answering my point. You are running a game that needs 8 cores. Simultaneously, you are running 1-2-3-5 apps that use 1 core each. Aren't all of those apps, includiing the game, going to run on one ccd while the other is parked?

 

[Bencher]

That's a contrive situation though, you're literally breaking it to say it s broken.

Just leave it alone and it will work properly.

I've no idea what you are talking about? What are you breaking exactly?

 

[LtMatt]

But you are still not answering my point. You are running a game that needs 8 cores. Simultaneously, you are running 1-2-3-5 apps that use 1 core each. Aren't all of those apps, includiing the game, going to run on one ccd while the other is parked?

No, that Was MW2, that wants 16 cores and it can use them. It’s not stuck to one CCD at all. Read my post after that post you quoted where I try to interpret your point.I think I get your point see my follow up posts.

 

[Anth-h90]

But you are still not answering my point. You are running a game that needs 8 cores. Simultaneously, you are running 1-2-3-5 apps that use 1 core each. Aren't all of those apps, includiing the game, going to run on one ccd while the other is parked?

I believe this was answered in the review pack sent to journo's. If the thread utilisation is high enough the cores on the parked CCD would be enabled as needed. So in your example above the 2nd CCD would be enabled to run the other tasks.

 

[Bencher]

I believe this was answered in the review pack sent to journo's. If the thread utilisation is high enough the cores on the parked CCD would be enabled as needed. So in your example above the 2nd CCD would be enabled to run the other tasks.

Yes but pcworld's test proved that that's not whats happening. They run a game (SOTR) alongside cinebench with 4 threads and the 2nd ccd was still parked!

 

[LtMatt]

Yes but pcworld's test proved that that's not whats happening. They run a game (SOTR) alongside cinebench with 4 threads and the 2nd ccd was still parked!

As I said to you earlier, that just doesn’t make any sense to do that. Let Windows manage it and both run fine as I’ve shown in multiple videos over several games and apps.

So what’s the issue here? Just the fact that you can break something by manually assigning affinity in a completely unrealistic and nonsensical manner? If that’s your point then I agree with @humbug.

 

[Hotwired]

Yes but pcworld's test proved that that's not whats happening. They run a game (SOTR) alongside cinebench with 4 threads and the 2nd ccd was still parked!

Not proof of?

If thread utilisation is high enough

Starting CB with 4 threads alongside SOTR isn't counting as high enough then.

The moment the scheduler opens up the second CCD it introduces issues which harm most games and it's trying to run for the best operation of games.

 

[Bencher]

As I said to you earlier, that just doesn’t make any sense to do that. Let Windows manage it and both run fine as I’ve shown in multiple videos over several games and apps.

So what’s the issue here? Just the fact that you can break something by manually assigning affinity in a completely unrealistic and nonsensical manner? If that’s your point then I agree with @humbug.

Βut he didn't assign affinity from windows. He chose the "run on 4 threads" option from within cinebench.

Man im asking something simple and im not getting an answer..say im playing a game that maxes 8 cores, and im running 4 instances of cinebench running the single thread benchmark. Will the 2nd CCD activate or not? It's a simple yes or no question...

 

[J.D]

Βut he didn't assign affinity from windows. He chose the "run on 4 threads" option from within cinebench.

Man im asking something simple and im not getting an answer..say im playing a game that maxes 8 cores, and im running 4 instances of cinebench running the single thread benchmark. Will the 2nd CCD activate or not? It's a simple yes or no question...