Ryzen AM5 ; 7000

[RasmusP]

More like 5600CL28 or 5200CL26 if such things exist. But no worries, don't waste your time on benching those sims, I was just curious.

Looks like hwub heard us and did what I had planned, lol:

 

[BrunoBæ]

My former German IT bible PCGamesHW have made a compare between AMD 5800X3D and 5800X.
Sorry about the photo quality.

 

[Spinelli]

Those timings don't mean much. First of all, they're loose. Second of all, you can get more performance gains from tuning subtimings like secondary and tertiary timings.

I literally got 10-20 % performance gains in some games from fully tuning my RAM. Primary, secondary, and tertiary timings.

 

[RasmusP]

Those timings don't mean much. First of all, they're loose. Second of all, you can get more performance gains from tuning subtimings like secondary and tertiary timings.

I literally got 10-20 % performance gains in some games from fully tuning my RAM. Primary, secondary, and tertiary timings.

Did you check their full sheet with the timings they received from buildzoid?

I don't have much clue about all these timings, but the gains are already there so it can only get better.

 

[Neilski]

Did you check their full sheet with the timings they received from buildzoid?

Some of the games had simply astonishing changes in frame rate from changes to the secondary timings. (Slightly depressing that I know so little about them, but I can console myself by noting that some games weren't remotely sensitive to the memory timings )

 

[pattikins]

The AMD Ryzen 9 7950X3D cpu seems to have taken the performance crown in ACC, but not by as much as I thought it would:

 

[demetri]

Not bad, 5800X3D was 50% faster than 5800X there and 7950X3D is also 50% faster than 7950X, both using some run-of-the-mill 5200 MHz CL38 RAM.
You couldn't really be asking for more

 

[Novopaine]

I'm holding out for the 7800x3d... can't wait. I just hope availability is good.

 

[RasmusP]

https://www.purepc.pl/amd-ryzen-9-7950x3d-vs-intel-core-i9-13900k-test-wydajnosci-najszybszych-procesorow-kto-wygra-pojedynek-tytanow?page=0,20

PurePC.pl are at it again too on day 1

7950X3D = 113.6 fps
5800X3D = 100.6 fps
13900k = 87.5 fps
7950X = 74.9 fps
5800X = 67.0 fps


Not bad! Not as insane as before but I guess the big gains in ACC are due to the massive cache and not really from the CPU performance. So the 3D variants scale from each other like the non-3D CPUs do.

 

[Rebel_488]

Weird they don’t use a cpu limited sim like rf2 to really test things properly.

 

[RasmusP]

Weird they don’t use a cpu limited sim like rf2 to really test things properly.

Nothing apart from ACC is big enough to get tested. Before acc, it was project cars that got tested.
At least ACC is cpu limited during their tests (and it usually is for me, with 10600k & 3080 at 3440x1440, 90 fps limit (100 Hz gsync Monitor)).
And it still is with my 7600X. But only if I make all cars visible in a 40 grid race.

 

[Neilski]

Weird they don’t use a cpu limited sim like rf2 to really test things properly.

Well, the Hardware Unboxed fps variations weren't enormous, but the purepc.pl data that @RasmusP posted had a massive range, even just comparing the newest AMD chips to the 12th gen Intel (like more than 40% speed gain from 12900K to 9750X3D), so that's crazy deep into CPU-bound in my book.
I'm a bit confused though at why the two tests gave such different results, esp. given that the purepc.pl used a 3090, while they had a 4090 for the HWUB tests.
[edit: that'd be a 7950X3D; the 9750 might come along in a few years time ]

 

[RasmusP]

I'm a bit confused though at why the two tests gave such different results, esp. given that the purepc.pl used a 3090, while they had a 4090 for the HWUB tests.

It's down the massively different settings.
PurePc.pl goes for the heaviest settings possible:
Night, thunderstorm, max AI, all visible, max settings.

Hwub goes for more realistic settings:
Day light, clear sky, 20 AI or so, 20 visible, medium settings.

Apparently it doesn't simply scale the fps up/down but instead works differently on amd/Intel/cache size.

 

[demetri]

PurePC.pl also uses potato RAM (both their DDR4 and DDR5 kits), so a CPU with a bigger cache gets a bigger improvement

 

[RasmusP]

PurePC.pl also uses potato RAM (both their DDR4 and DDR5 kits), so a CPU with a bigger cache gets a bigger improvement

Yeah true.. They could really use 3600 CL16 and 5600 CL36.
But they keep a lot of older results in their charts, which I really like, and therefore they keep everything the same as it was back then.

It's not the most complete source but it is a really valuable one imo.
Especially when the 9th and 10th gen CPUs were around, they were one of the very few to test ACC with the i5's pushed to 5 GHz or higher.

 

[Spinelli]

The huge cache on AMD 3D chips means RAM frequency & latency/timings makes much less difference with them.

If you want to see "maxed out" benchmark comparisons where the CPUs and RAMs are properly tuned, wait for Frame Chaser's video.

 

[Spinelli]

7950x VS 13900K VS 7950x3D "maxed out" benchmarks:

If you just care about the benchmarks, they start @ 7:45

He has the 8 non-3D cores on the 7950x3D disabled so he doesn't have to deal with Windows scheduling between 3D and non-3D cores, higher temps (and possibly lower clocks), and all the auto-stuff involved with the AMD's partly-3D, partly-non-3D chips. He also has the e-cores on the 13900K disabled.

He said he'll be doing another video of the 7950x3D with all it's cores (8 3D, lower clocks + 8 non-3D, higher clocks) enabled.

 

[Neilski]

If you just care about the benchmarks, they start @ 7:45

May give it a go but I find his videos reeeeeeallly hard work. From the titles he chooses, to his facial expressions, to his voice, it's 110% clickbait
Wouldn't say no to an executive summary!

 

[RasmusP]

May give it a go but I find his videos reeeeeeallly hard work. From the titles he chooses, to his facial expressions, to his voice, it's 110% clickbait

And he needs a De-Esser. I could barely stand him talking but the content was pretty good!

Wouldn't say no to an executive summary!

The CPU is great. Some games prefer Intel, some games prefer AMD. Where the non-3D lost clearly to the 13900k, the 3D keeps up.
Making it very very even on average at the end!

But the results are a lot better on average with the second CCD disabled. So basically a 7800X3D with bigger standard cache.

And it allows for some tweaking, so you can get quite some higher clock speeds out of it. Not as locked as the 5800X3D is!
Which boosts the performance when the non-3D was better than the 3D.

Overall the 1% lows were a bit behind the very heavily tweaked 13900k. But I'd argue that it's only in games that have high fps anyway, so who cares.
For games like simracing in VR, where the 1% lows are below 100 fps, the 3D seems to be mostly better than the Intel.

 

[Neilski]

he needs a De-Esser

OMG, that's a real thing
Thanks for the summary!