News AMD Ryzen 9 7950X3D Tested in Blender, Geekbench 5

[Admin]

AMD's Ryzen 9 7950X3D sneaks into performance databases of popular benchmark suites.

AMD Ryzen 9 7950X3D Tested in Blender, Geekbench 5 : Read more

 

[Deleted member 2731765]

"Of course, we cannot draw accurate conclusions based on test results from one or two pre-production systems "

Yes, while the X3D CPU lags behind the Ryzen 9 7950X on both counts, it doesn't matter because neither test can take full advantage of its 3D V-cache. Also, the average 7950X score is based on many results/scores, whereas the 7950X3D's scores are from only one run.

Even the memory/RAM speed specs are unknown for the X3D part, unless I'm mistaken.

Geekbench and Blender don't tell the complete story as they can't take advantage of the massive 144 MB of L2 + L3 cache on the processor. Then again, one can rest easy knowing the processors are out and about, and it shouldn't be long before someone tests them with AAA games.

In GAMING workloads the X3D SKU should be definitely faster, though this remains to be seen.

BTW, the new X3D processor has also been spotted in "Sisoftware" sandra benchmark.

EDIT:

Also, the deficit in multi-core performance is to be expected considering the Ryzen 9 7950X3D doesn't has a lot of voltage or the thermal headroom to work with as the standard non-3D CPUs.

Not to mention that 7950X3D chips have a different CCD configuration, with one CCD running at higher clocks and the other with conservative limits set in place due to the 3D V-Cache stacking structure.

The benchmark log shows that the chip maxed out around a peak 5.5 GHz boost clock which is slightly slower than the 5.7 GHz boosts that the 7950X non-3D can hit and explains the lower multi-core numbers.

 

[richardvday]

This News is not really news at all.
It's already been established that the benchmark's talked about here do not benefit from the increased cache.

This is not the intended primary workflow for these CPUs anyway. This is meant for primarily gaming but I still need to do some multi threaded work reasonably well. It's a compromise.
Personally I would build a separate system for that but maybe some people don't want to or can not justify two systems.

 

[-Fran-]

You guys should start adding the 5800X and 5800X3D to extrapolate a bit more.

While I think these are interesting, they don't have enough information to draw any conclusions, really.

Regards.

 

[pointa2b]

I don't think it was ever argued the 7950X3D would be superior to the 7950X in all applications. At worst against the X, its basically just a slightly underclocked version of it. What would be really interesting to see is how it compares with games/applications that benefit heavily from a larger cache. Thats what it all boils down to. 😎

 

[Quirkz]

So now the glaring weakness of thee 5800X3D is resolved: You no longer have to give up have your productivity performance, you only sacrifice a little. These results are amazing, and it's weird they making it sound like it's bad with words like 'unimpressed'.

Now the question is did they solve part 2: Is the 7950X3D as big an improvement in gaming as the 5800X3D was.

 

[moongrim]

Its all about the games games games, just few select applications take advantage of extra cache, synthetic benchmarks are useless in measuring performance in these cases

 

[hotaru251]

whats that? a non cache heavy benchmark makes a cpu with lower clocks perform worse?

[insert surprised pikachu face here]

 

[Ravestein NL]

So now the glaring weakness of thee 5800X3D is resolved: You no longer have to give up have your productivity performance, you only sacrifice a little. These results are amazing, and it's weird they making it sound like it's bad with words like 'unimpressed'.

Now the question is did they solve part 2: Is the 7950X3D as big an improvement in gaming as the 5800X3D was.

Indeed, this is excellent, about the same speed in productivity and 30% less power needed (did anyone notice that). If this CPU will perform the same in games like it's predecessor than we couldn't be happier!

 

[BeedooX]

Also, the deficit in multi-core performance is to be expected considering the Ryzen 9 7950X3D doesn't has a lot of voltage or the thermal headroom to work with as the standard non-3D CPUs.

I suspect that's just as much to do with the fact that AMD realised they didn't need to put oodles of voltage through their CPUs after all. You run your 7950X in ECO mode and will probably be equally happy.

 

[btmedic04]

So now the glaring weakness of thee 5800X3D is resolved: You no longer have to give up have your productivity performance, you only sacrifice a little. These results are amazing, and it's weird they making it sound like it's bad with words like 'unimpressed'.

Now the question is did they solve part 2: Is the 7950X3D as big an improvement in gaming as the 5800X3D was.

seeing that your new here, allow me to give you a heads up. you can expect any written by Anton or Aaron regarding an amd part to have a negative connotation. Their own personal bias is always apparent

 

[Maebius]

We really need to be patient for the new X3D chips' benchmarks. (but can't 😛)
Many things are different than the last one, so we can't even safely extrapolate by just comparing 5800x to 5800x3D
-Better IPC for the new generation
-Higher Clock Speed
-DDR5
-The 79XX x3Ds might have core "assignment" issues at first, like the intel E-cores had
-The new ones can do PBO
-Intel chips have a bit more cache than the previous generation

Overall, I'm really looking forward to the 7800x3D (for gaming).
I'm not sure what I hope for really... if the 7950x3D has really good gaming benchmark results, I might not be able to resist 😊 holding off for the 7800, which I kind of suspect might turn out to be equally good (for gaming).

 

[hannibal]

5800x3D lose in blender to 5800x so this is just like it should be! If you are doing productivity. Don´t lose money on 3dcache versions!
Because this only has one chip with 3dcache it is avtually allmost as fast as normal 7950 so this is good news to any people with production heavy PC usage that also play games sometimes!

 

[Friesiansam]

Gaming performance is what will matter to the target buyers, not Blender and Geekbench...

 

[JayGau]

I am so disappointed by this Tom's Hardware article. It makes me worried about your competencies guys. Just take a look at the 5800X3D. The chip is indeed slightly slower than it's regular counterpart because of the lower max clock frequency but destroys almost every other CPUs in gaming, which is the main purpose of the 3D models. They are gaming CPUs, not for productivity. And everything I wrote here was known since the launch of the 5800X3D 3 years ago. The 7000 models will most likely do the same and I can't wait to see the results in game benchmarks.

 

[blppt]

As others have pointed out, the GB results for the lower clocked 5800X3D were terrible in comparison with similarly priced and cored CPUs. It means nothing to the audience interested in the X3D models.

 

[bit_user]

This is hardly surprising. These same benchmarks also favored the 5800X over the 5800X3D.

The (risky) bet AMD is making is that apps like games will either have patches which apply their own thread-affinities to the chiplet with extra cache, or that a Windows 11 patch can monitor cache usage stats by threads and steer them appropriately. I'm skeptical it'll work terribly well, but I guess we'll soon find out.

 

[bit_user]

Many things are different than the last one, so we can't even safely extrapolate by just comparing 5800x to 5800x3D

I disagree. Something that wasn't cache-limited on Zen 3 still probably won't be cache-limited on Zen 4. While Zen 4 has more IPC, DDR5 does a much better job of feeding it.

The new ones can do PBO

Isn't PBO limited to the die without the extra cache?

 

[bill001g]

Until we get gaming tests none of this really matters but I guess this data does confirm what people have guessed so that might be a good sign for the other benchmarks to come.

Problem is a lot of the so called gaming benchmarks are as invalid as blender. The ones that use aggregated numbers can be greatly skewed by games such as tomb raider being included which is greatly impacted by the 3d cache where a different combination of games will give different result.
This is where people have to try to find game benchmarks for the games they actually play rather than be lazy and base their choices on some review numbers.

Since most people spending this kind of money for parts are going to run higher than 1080 it will be interesting to see how much difference this makes at higher resolutions. We have seen the current cpus tend to have to wait on the video card at higher resolutions so higher clock and core counts make less difference.
So if the 3d cache makes a bigger difference at 4k does that mean a 7800x3d or 7900x3d be a better option than the 7950x3d if we take the cost of the cpu into consideration.

 

[Maebius]

I disagree. Something that wasn't cache-limited on Zen 3 still probably won't be cache-limited on Zen 4. While Zen 4 has more IPC, DDR5 does a much better job of feeding it.

Indeed but these things might not offset each other in a linear manner to make a good guess. And since DDR5 might feed the cpu better, the extra cache's help might not be as pronounced as it is with the 5800x3d

Isn't PBO limited to the die without the extra cache?

Is it? It's available for the 7800x3d so why wouldn't it be available for both 79xx dies?

We'll see. /shrug

 

[aberkae]

For the price of one 7900x 3d today you can get the 7900x, cl30 32 gigs of gskill ram kits at ddr5 6ghz and an Asus strix F b650e motherboard at microcenter currently. For reference and FYI.
I understand that people want the Ryzen 4 3d cpus to be great at gaming , some through uncocoius bias. We should attemp to keep things in perspective. The hype train will lead to AMD to making off like a bandit Nvidia style in terms of pricing going forward. Also Ryzen 4 doesn't stop being good at gaming for those professionals who casually game.



AMD Ryzen 9 7900X, ASUS B650E-F ROG Strix Gaming WiFi, G.Skill Flare X5 Series 32GB DDR5-6000 Kit, Computer Build Combo
Original Price$857.96
Save $257.97
$599.99
Source local microcenter

Now let's add the b650e and 32 gigs of ddr5 ram to the 7900x 3d which likely will not have any deals thanks to the super hype train surrounding it and we are looking at an additional $500 premium just to " casual game" for professionals. Thoughts?

 

[Maebius]

Thoughts?

I think not everyone lives in the US and even there, not all cities have a Microcenter.
Obviously, if there are no similar bundles, the X3Ds will be of poor value, unless they're super good.
But there are no similar bundles in most places around the world so, who knows? (the above example costs 900ish+ euros incl VAT in Europe)

 

[bit_user]

For the price of one 7900x 3d today ...

Well, a lot of people don't really care if it's not the best value. If it can deliver most of the benefits of the 5800X3D without most of the downsides, then it'll sell well enough. That's what AMD seems to be counting on. I think it's a somewhat risky bet, and I expect a lot of gamers will still prefer the 7800X3D.

I'm finding myself wondering why they didn't enable cache stacking on the I/O die. I guess that won't really help with EPYC, but it seems like somewhere you could put extra L3 (or L4) without the clock speed tradeoffs.

 

[-Fran-]

Well, a lot of people don't really care if it's not the best value. If it can deliver most of the benefits of the 5800X3D without most of the downsides, then it'll sell well enough. That's what AMD seems to be counting on. I think it's a somewhat risky bet, and I expect a lot of gamers will still prefer the 7800X3D.

I'm finding myself wondering why they didn't enable cache stacking on the I/O die. I guess that won't really help with EPYC, but it seems like somewhere you could put extra L3 (or L4) without the clock speed tradeoffs.

Interconnects and speed (+ latency). If you put the extra cache there, the amount of latency and interconnects you'd need will blow the packaging cost, I'm sure. Plus, it wouldn't be L3 cache anymore. Probably more like ERAM or HBM.

Regards.

 

[bit_user]

Interconnects and speed (+ latency).

They already have high-bandwidth interconnects between the CCD and IOD, because every time a CCD gets a cache miss, it needs to snoop & potentially fetch the cacheline from other CCD. According to this, Zen 4 beefed up the bandwidth to about 1.5 TB/s:

Alleged AMD Ryzen 9 7900X "Zen 4" CPU Offers Over 50% Cache Bandwidth Versus Zen 3

An alleged cache benchmark of AMD's Ryzen 9 7900X "Zen 4" CPU has been published by leaker, HXL, showing a 50% bandwidth gain over Zen 3.

wccftech.com

Plus, it wouldn't be L3 cache anymore. Probably more like ERAM or HBM.

No, you probably wouldn't move all your L3 there, but I did offer that it could be L4.

Be careful about throwing around terms like HBM, because that means something very specific and it's not at all what I was talking about.