Review Zen4 3D review thread

[Hitman928]

Thread to focus on Zen4 3D cache CPUs.

New gaming king (as most expected), though the 2 CCD 7950x3d does seem to have issues with some games, more than I would expect of it getting stuck on the "wrong" CCD. I imagine it will get cleared up with subsequent updates but we'll see. Simulated 7800X3D showed no such issues and overall has the gaming lead (real product might be slightly slower though depending on in game clocks).

Computerbase also has the 7950x3d as the gaming champ. They (and TPU) also show that efficiency while gaming is extremely good.

Ryzen 9 7950X3D im Test gegen 5800X3D und 13900K(S)

Zen 4 3D ist da. Im Test erweist sich AMDs Ryzen 9 7950X3D gegen 7950X, 5800X3D und Core i9-13900K(S) als schnell und effizient.

www.computerbase.de

Just to toot my own horn a little, it landed spot on with my prediction of fastest gaming CPU but not significantly so over a 13900k on average, but with much higher efficiency.

Additional reviews, will add more later.

AMD Ryzen 9 7950X3D processor review

We review the new Ryzen 9 7950X3D processor from AMD. The much-anticipated processor series comes with an added cache that will help predominantly with gaming. This processor series was designed for ...

www.guru3d.com

AMD Ryzen 9 7950X3D Review - Best of Both Worlds

The Ryzen 9 7950X3D is the spearhead of the AMD Zen 4 X3D lineup. In our performance review we test AMD's new 16-core flagship with dozens of applications and 14 games at up to 4K, to see whether AMD can take the performance crown from Intel, how power consumption is affected, and what can be...

www.techpowerup.com

The AMD Ryzen 9 7950X3D Review: AMD's Fastest Gaming Processor

www.anandtech.com

Gamers Nexus

AMD Ryzen 9 7950X3D Linux Performance Review - Phoronix

www.phoronix.com

AMD Ryzen 9 7950X3D Review: AMD Retakes Gaming Crown with 3D V-Cache

AMD takes a bite out of Raptor Lake.

www.tomshardware.com

AMD Ryzen 9 7950X3D review: Team Red retakes the lead with its best CPU ever

The AMD Ryzen 9 7950X3D is the best gaming CPU there is, and it's not close

www.techradar.com

 

[Geddagod]

Wonder if he takes requests though?
The power usage in apps chart got me thinking:

AMD Ryzen 9 7950X3D Review - Best of Both Worlds

The Ryzen 9 7950X3D is the spearhead of the AMD Zen 4 X3D lineup. In our performance review we test AMD's new 16-core flagship with dozens of applications and 14 games at up to 4K, to see whether AMD can take the performance crown from Intel, how power consumption is affected, and what can be...

www.techpowerup.com

View attachment 77327
So I took those "numbers" (visually, any mistakes are mine), and then benchmark scores to get this figured who "won" each bench and by how much and then took that to do a (linear) power AMD and Intel would need and got something like this:

Looking at it, that may be too much (so for the first one, if AMD were 100% then Intel use about 180% the power, or if Intel were 100% then AMD would use about 55%), so I guess the blue lines on their own should be easier to digest:

This time sorted by Intel's power usage: the first seven Intel wins, the rest AMD wins sometimes by huge margins.

Unless Intel can finally come up with something better soon (the famous Intel mantra for the last few years), then their server marketshare could really nosedive. There is almost nothing where their hybrid architecture does well at.

Ye, and I'm sure the worse efficiency of RPL boils down to the E-cores. Nothing to do with Zen 4 using 5nm while RPL is on Intel 7. Surely it's all on the hybrid architecture....

 

[biostud]

This is why the 7950x came first. For the sales.

I completely get it - depending on who I was working with for a build, I'd do it in a heartbeat. If they are making a $1,600 line item in their PC budget for a 4090 I'd be saying we are going the 7950x as a matter of course.

Sadly, I only do a build or two like that per year... if that.

If anything, I am surprised they bothered with the 12 core part unless it's only their to reinforce the value of the higher priced SKU and give the OEMs a solid offering. This wouldn't surprise me, good - better - best is classic.

Value is the 7800x, epeen and just overall mastery is the 7950x.

Yeah the 7900X3D must be for OEM market mostly. If they for some reason had a large surplus of CCDs with non functional cores, a 7600X3D could be priced relatively high and still be a great budget gaming CPU.

 

[Hitman928]

In this case the average is compared to the stock 7950x3d, so yes we can just look at the average fps chart and make direct correlations. Its literally impossible for a 12900k to be any % more relatively faster since the 0ccd fps result is a greater number than the 12900k result. 130fps would always be a higher percentage in relative terms, yet the graphs show it being lower somehow. The charts are using different datasets or there is an error like 12900k results swapped with the 0ccd on the fps chart.

@Sunaiac is correct. A simple example is if you tested two games with the following results:

CPU A: 50 fps in game 1 and 660 fps in game 2.

CPU B: 60 fps in game 1 and 600 fps in game 2.

If you use average fps, then CPU A is 355 fps average and CPU B is 330 fps average. If you calculate relative performance from this, then CPU A would be 7.6% faster.

However, if you first take the relative performance of each game first, then take the average of the relative performance results, you would get CPU B as being 5.5% faster. Different results but both are using the same data.

In the first case you are giving more weight to the tests where both run high fps, just because they run higher numbers. In the second example you are making each test equal weight so a 10% win means the same thing no matter how high or low the absolute fps numbers are.

I used averages here but geometric means are probably more appropriate in the second case. Not sure exactly how TPU calculates the relative performance chart though.

 

[DAPUNISHER]

Wonder if he takes requests though?

I know he does a good job keeping his finger on the pulse of the enthusiast community. And he has been known to add stuff if enough of us request something.

On the power use: Does anyone know if it's the Asus board or what, that is causing the high idle power usage with the Zen 4 CPUs in the Tech Testers review?

 

[Makaveli]

I know he does a good job keeping his finger on the pulse of the enthusiast community. And he has been known to add stuff if enough of us request something.

On the power use: Does anyone know if it's the Asus board or what, that is causing the high idle power usage with the Zen 4 CPUs in the Tech Testers review?

View attachment 77343

That is a good question considering the gigabyte board was doing like 23w on idle.

 

[Hitman928]

I know he does a good job keeping his finger on the pulse of the enthusiast community. And he has been known to add stuff if enough of us request something.

On the power use: Does anyone know if it's the Asus board or what, that is causing the high idle power usage with the Zen 4 CPUs in the Tech Testers review?

View attachment 77343

Could be the motherboard not allowing the CPU to actually idle, or a bad Windows setting. Could also just be the application not properly reporting idle power as has happened in the past.

 

[q52]

anyone know if these work out of the box with the new 48GB memory modules?

https://www.corsair.com/us/en/Categories/Products/Memory/VENGEANCE-RGB-DDR5-%E2%80%94-Black/p/CMH96GX5M2B5200C38

AMD's website still says that 7950X3D only supports 128GB of memory; not sure if thats an actual hard-limit on the CPU or if they only wrote that because at the time only 32GB memory modules were available?

 

[Spicedaddy]

TBH, the moment I saw 13900K/KS pull north of 200W in Spider-Man was the last straw for me, def not building a Intel system anytime in the foreseeable future.

I guess that's at 1080p? I get around 100W CPU package power on 13900k at 4K RT.

1080p benchmarks are fun to see theoretical performance, but most people building a new system aren't targetting 1080p.

 

[biostud]

I guess that's at 1080p? I get around 100W CPU package power on 13900k at 4K RT.

1080p benchmarks are fun to see theoretical performance, but most people building a new system aren't targetting 1080p.

And most people aren't buying either a 13900k or a 7950X3D

 

[PhonakV30]

Those who own Flight simulator , Don't underestimate them ! , They're all crazy/insane people and Always tend to smoke their money!!

 

[Joe NYC]

anyone know if these work out of the box with the new 48GB memory modules?

https://www.corsair.com/us/en/Categories/Products/Memory/VENGEANCE-RGB-DDR5-%E2%80%94-Black/p/CMH96GX5M2B5200C38

AMD's website still says that 7950X3D only supports 128GB of memory; not sure if thats an actual hard-limit on the CPU or if they only wrote that because at the time only 32GB memory modules were available?

Seems like a solution looking for a problem. To fragment the market more and unnecessarily.

And BTW, you can get a a 2x 32 GB EXPO DDR5 6000 with decent timings for ~ $300

Are you a human?

www.newegg.com

Faster, Better, Cheaper.

Seems like it breaks one of the fundamental laws of the universe:

 

[q52]

Seems like a solution looking for a problem. To fragment the market more and unnecessarily.

And BTW, you can get a a 2x 32 GB EXPO DDR5 6000 with decent timings for ~ $300

Are you a human?

www.newegg.com

Faster, Better, Cheaper.

not sure what you mean, I dont care about price, dont actually even care about timings and speed either, I am trying to get the most CPU cores and memory into a small desktop package like mITX, and 2x48GB gives a lot more memory than 2x32GB (which I already have)

 

[Joe NYC]

not sure what you mean, I dont care about price, dont actually even care about timings and speed either, I am trying to get the most CPU cores and memory into a small desktop package like mITX, and 2x48GB gives a lot more memory than 2x32GB (which I already have)

I see, I misread it, I thought it was 2x 24GB.

Yeah, 2x48 would give you more, but I don't think it is supported right now on AM5 motherboards, AFAIK. May need some BIOS upgrades, assuming memory controller can handle it.

 

[KompuKare]

Seems like it breaks one of the fundamental laws of the universe:

Never mind that, a memory module whose capacity isn't base 2 is very strange.

Is it 32GB + 16GB in some asymmetric design? Is it some kind of MLC design where each cell has 3 states?
A quick search shows that Micron and Hynix have both introduced these new 24Gb chips so they are going to become more common.

TBH, the moment I saw 13900K/KS pull north of 200W in Spider-Man was the last straw for me, def not building a Intel system anytime in the foreseeable future.

Those figures I transcribed from TPU to see perf/watt in their apps are bad, and while CPU power usage isn't that important in gaming (since GPUs use so much power), the ComputerBase gaming power usage table was actually pretty brutal:

Ryzen 9 7950X3D im Test gegen 5800X3D und 13900K(S)

Zen 4 3D ist da. Im Test erweist sich AMDs Ryzen 9 7950X3D gegen 7950X, 5800X3D und Core i9-13900K(S) als schnell und effizient.

www.computerbase.de

	Ryzen 9 7950X3D Package-Power (AVG/Max)	Ryzen 9 7950X Package-Power (AVG/Max)	Ryzen 7 5800X3D Package-Power (AVG/Max)	Core i9-13900K Package-Power (AVG/Max)	Core i9-13900KS Package-Power (AVG/Max)
Age of Empires IV	53/71 Watt	68/72 Watt	49/50 Watt	70/73 Watt	72/75 Watt
CoD Warzone 2.0	78/83 Watt	112/115 Watt	82/85 Watt	153/158 Watt	158/166 Watt
Cyberpunk 2077	75/83 Watt	108/115 Watt	77/82 Watt	165/174 Watt	171/181 Watt
Dead Space	81/85Watt	117/125 Watt	92/97 Watt	163/180 Watt	177/191 Watt
Diablo II: Resurrected	63/66 Watt	91/95 Watt	65/70 Watt	106/112 Watt	107/113 Watt
Dota 2	72/76 Watt	101/106 Watt	75/79 Watt	129/140 Watt	133/145 Watt
Forspoken	85/89 Watt	123/130 Watt	84/87 Watt	181/186 Watt	191/196 Watt
Fortnite	73/84 Watt	116/120 Watt	75/81 Watt	152/157 Watt	161/164 Watt
Ghostwire: Tokyo	69/73 Watt	111/121 Watt	76/81 Watt	148/156 Watt	155/163 Watt
Resident Evil Village	74/76 Watt	95/101 Watt	77/80 Watt	120/127 Watt	123/131 Watt
Spider-Man Remastered	82/86 Watt	127/130 Watt	82/85 Watt	182/187 Watt	194/200 Watt
The Callisto Protocol	66/74 Watt	91/94 Watt	69/73 Watt	114/128 Watt	119/131 Watt
The Witcher 3	71/75 Watt	110/126 Watt	79/81 Watt	143/149 Watt	156/172 Watt
Uncharted	69/71 Watt	106/110 Watt	84/88 Watt	145/150 Watt	147/152 Watt
Durchschnitt	72/78 Watt	105/111 Watt	76/80 Watt	141/148 Watt	147/156 Watt

charting only 7950X3D and the 13900KS we get something like this:

For Spider-Man not only is the ratio about x2.37, those figures aren't even scaled for performance. Since the 7950X3D is 6% to 7% faster while using about 42% of the power of the KS.
That's actually worse than some of the application benchmarks.

Unsure whether Intel are repeating P4 or trying to copy Bulldozer, but brute force sure is the wrong approach.

 

[lightmanek]

I know he does a good job keeping his finger on the pulse of the enthusiast community. And he has been known to add stuff if enough of us request something.

On the power use: Does anyone know if it's the Asus board or what, that is causing the high idle power usage with the Zen 4 CPUs in the Tech Testers review?

View attachment 77343

From what one of our forum members discovered, lower power C states are not working on ASUS and possibly other branded boards with recent BIOS. Gigabyte so far is really delivering the best out of the box XMP memory performance and also lowest power at similar feature level compared to other manufacturers. MSI is catching up looking at few reviews published recently.
ASUS seems to be in a mode - max. performance and OC results, efficiency be damned!
As always with new platforms, revision 2 of chipsets and boards will be almost perfect (X770Eee or whathever name it might be).

 

[Kaluan]

Interesting info catch here, AMD + AMD combos seem to win big(ger), at least for these V-Cache parts:


vs


Do note that ComputerBase tests with native supported RAM (so 5200MT for Raphael/Raphael-X, 5600 for 13th gen) and... as I'll explain in the next reply:

I guess that's at 1080p? I get around 100W CPU package power on 13900k at 4K RT.

1080p benchmarks are fun to see theoretical performance, but most people building a new system aren't targetting 1080p.

It's actually 1280x720 + FSR Performance or 37% res scaling (depending on what's available). lmao

 

[alexruiz]

Interesting info catch here, AMD + AMD combos seem to win big(ger), at least for these V-Cache parts:

View attachment 77377
vs
View attachment 77380

Do note that ComputerBase tests with native supported RAM (so 5200MT for Raphael/Raphael-X, 5600 for 13th gen) and... as I'll explain in the next reply:


It's actually 1280x720 + FSR Performance or 37% res scaling (depending on what's available). lmao

Ryzen runs better with Radeon, and Radeon runs better on Ryzen

 

[scannall]

While these reviews are pretty much academic for me, since I just upgraded a year ago I have to admit the inner geek is straining at the bars of the cage. But the old man in me has the power of wisdom. ;-)

For gamers I am thinking the 8 core will be the one to get. We'll be sure after the reviews go live.

 

[poke01]

Never mind that, a memory module whose capacity isn't base 2 is very strange.

Is it 32GB + 16GB in some asymmetric design? Is it some kind of MLC design where each cell has 3 states?
A quick search shows that Micron and Hynix have both introduced these new 24Gb chips so they are going to become more common.



Those figures I transcribed from TPU to see perf/watt in their apps are bad, and while CPU power usage isn't that important in gaming (since GPUs use so much power), the ComputerBase gaming power usage table was actually pretty brutal:

Ryzen 9 7950X3D im Test gegen 5800X3D und 13900K(S)

Zen 4 3D ist da. Im Test erweist sich AMDs Ryzen 9 7950X3D gegen 7950X, 5800X3D und Core i9-13900K(S) als schnell und effizient.

www.computerbase.de

	Ryzen 9 7950X3D Package-Power (AVG/Max)	Ryzen 9 7950X Package-Power (AVG/Max)	Ryzen 7 5800X3D Package-Power (AVG/Max)	Core i9-13900K Package-Power (AVG/Max)	Core i9-13900KS Package-Power (AVG/Max)
Age of Empires IV	53/71 Watt	68/72 Watt	49/50 Watt	70/73 Watt	72/75 Watt
CoD Warzone 2.0	78/83 Watt	112/115 Watt	82/85 Watt	153/158 Watt	158/166 Watt
Cyberpunk 2077	75/83 Watt	108/115 Watt	77/82 Watt	165/174 Watt	171/181 Watt
Dead Space	81/85Watt	117/125 Watt	92/97 Watt	163/180 Watt	177/191 Watt
Diablo II: Resurrected	63/66 Watt	91/95 Watt	65/70 Watt	106/112 Watt	107/113 Watt
Dota 2	72/76 Watt	101/106 Watt	75/79 Watt	129/140 Watt	133/145 Watt
Forspoken	85/89 Watt	123/130 Watt	84/87 Watt	181/186 Watt	191/196 Watt
Fortnite	73/84 Watt	116/120 Watt	75/81 Watt	152/157 Watt	161/164 Watt
Ghostwire: Tokyo	69/73 Watt	111/121 Watt	76/81 Watt	148/156 Watt	155/163 Watt
Resident Evil Village	74/76 Watt	95/101 Watt	77/80 Watt	120/127 Watt	123/131 Watt
Spider-Man Remastered	82/86 Watt	127/130 Watt	82/85 Watt	182/187 Watt	194/200 Watt
The Callisto Protocol	66/74 Watt	91/94 Watt	69/73 Watt	114/128 Watt	119/131 Watt
The Witcher 3	71/75 Watt	110/126 Watt	79/81 Watt	143/149 Watt	156/172 Watt
Uncharted	69/71 Watt	106/110 Watt	84/88 Watt	145/150 Watt	147/152 Watt
Durchschnitt	72/78 Watt	105/111 Watt	76/80 Watt	141/148 Watt	147/156 Watt

charting only 7950X3D and the 13900KS we get something like this:

For Spider-Man not only is the ratio about x2.37, those figures aren't even scaled for performance. Since the 7950X3D is 6% to 7% faster while using about 42% of the power of the KS.
That's actually worse than some of the application benchmarks.

Unsure whether Intel are repeating P4 or trying to copy Bulldozer, but brute force sure is the wrong approach.

What Intel needs is break and a clean sheet design. Focus on efficiency first and performance will follow.

 

[Timmah!]

The 7950X3D power limit is how the 7950X should have been released. Bringing the power consumption to 170W for a couple percent gain in MT was an error on AMD's side IMHO.

To be fair, if that was the case, the only incentive to buy the vanilla chip, at least as far i am concerned, would be somewhat lower price. This way, as it was released, there is at least decision to be made, as vanilla 7950x has those higher clocks thanks to its higher power limits, and thats something to consider for some people. Like for me for example, obviously, as everyone else i would love to have better FPS in games and alleviate the CPU bottleneck, but then again, if i run 3dsmax, autocad, photoshop, opera, winamp, etc... at the same time, its kinda nice to know that wont be randomly shuffled by Windows to a v-cache core, that can do 5GHz tops. As it is, i for one am leaning to keep my 7950x, though its really not an easy decision.

Additionally, i checked the TPU numbers in games, and obviously everyone is concentrating on the 720p numbers, which are indeed great, beating even Raptor Lakes, significantly superior to vanilla chips. I get it, its tested like that to avoid being bottlenecked by GPU, it makes perfect sense, if you want to see what benefits the v-cache brings. But, ultimately, its really not very real-world scenario. You wont be playing games with such CPU and 4090 at 720p or 1080p, it will be obviously 4K. And if you look at the numbers, relative difference between vanilla and 3D chip shrinks from cca 15 percent at 720p to meager 4~5 percent at 4K. Its kinda assumed that the low-res test is important, because if the CPU is more powerful there, it will be pushing more FPS even at higher resolutions, thus lessening the bottleneck....while that is true, its clearly not that hot, as the differential diminishes to single-digit numbers in there - on average per TPU tested games.

So ultimately its about how much gaming you want to do or what games do you specifically want to play, i would say for mixed use (both games and productivity) its a cointoss between 7950x a 7950x3D, unless those games are MSFS, Factorio, Stellaris etc...For purely gaming machine 7800x3D will be then way to go. But thats kinda obvious i guess.

 

[utahraptor]

Those who own Flight simulator , Don't underestimate them ! , They're all crazy/insane people and Always tend to smoke their money!!

Do the other simulators like train/farm/trucking ect have similiar demands are is it unique to the flight one?

 

[ZGR]

Do the other simulators like train/farm/trucking ect have similiar demands are is it unique to the flight one?

For MSFS2020 and especially in VR you definitely will want the X3D to maintain 90-120fps. I play at mostly low settings since mods allow to spawn way more AI air traffic to simulate real life. Busy airports are insanely CPU demanding if you do not cap the AI count in settings.

Disabling VR greatly reduces CPU load. My 3080 at 1.9 GHz is CPU limited at ~1700x1700 while in VR.

And it is kinda an MMO. So many actual players flying around it can definitely have a performance impact. It is the only game I am playing funnily enough.

 

[eek2121]

One of my workloads is Factorio. Guess I should get the 7950X3D. Dunno what I would do with my 7950X though. 🤣

 

[DAPUNISHER]

One of my workloads is Factorio. Guess I should get the 7950X3D. Dunno what I would do with my 7950X though. 🤣

Sell it. I just saw one sell here in FS/FT yesterday. Just call any loss a rental fee.

 

[igor_kavinski]

Dunno what I would do with my 7950X though. 🤣

I am open to a generous donation!