AMD Ryzen 7 5800X3D Retail CPU Gets First Independent Tests

[TheLostSwede]

An early retail unit of AMD's upcoming Ryzen 7 5800X3D has made its way to a Peruvian site called XanxoGamging, who put it through its paces in a few benchmarks, of which none so far are game related. The tests run on the upcoming CPU suggests that it's about as fast as a Ryzen 7 5700X in most single and multi-core tests. This should largely be down to the slower clock speeds of the Ryzen 7 5800X3D, which holds it back in these benchmarks compared to the older Ryzen 7 5800X.

However, it seems like some benchmarks can take advantage of the extra cache and the Ryzen 7 5800X3D is outperforming the 5800X in Blender, by a small margin. That said, the Cinebench R23 results are not overly impressive, neither are the CPU-Z or Geekbench 5 numbers. None of this is really unexpected though, especially as AMD has specifically mentioned that the 3D V-Cache doesn't bring additional performance to most software. XanxoGaming has promised more benchmarks and game tests tomorrow, but mentions that it feels strange losing performance in normal software due to the lower clocks, but that they hope the performance can be improved over time by an improved UEFI/AGESA.



View at TechPowerUp Main Site | Source

 

[Bwaze]

Could it be that this reviewer does not have proper bios and AMD driver? I doubt AMD would hype such an uninspiring product.

 

[TheLostSwede]

Could it be that this reviewer does not have proper bios and AMD driver? I doubt AMD would hype such an uninspiring product.

The UEFI/AGESA has been available for about a month and new drivers are out, so highly unlikely.
The tests were done with 3200MHz CL14 RAM, so that could account for some of it, but not much.

 

[Dragokar]

Just wait for the gaming benchmarks, I can still remember Clawhammer vs Newcastle in Socket 754 and 1024 KiB vs 512 KiB L2 Cache

Also 1.2.0.6 C is needed, but the full support version is 1.2.0.7 with the “for now” unofficial chipset driver.

 

[Bwaze]

Single core Cinenench R23 (by my opinion most indicative of gaming speeds here) is crap, slower than stock 5800x. Something must be wrong, or this won't beat the 5600x in gaming.

 

[TheLostSwede]

Also 1.2.0.6 C is needed, but the full support version is 1.2.0.7 with the “for now” unofficial chipset driver.

I guess we'll have to wait and see if it makes any real difference.

 

[Mussels]

You need a new AGESA and chipset driver, and it only benefits games.


So of course, people test synthetic benches that don't benefit from the cache.

Single core Cinenench R23 (by my opinion most indicative of gaming speeds here) is crap, slower than stock 5800x. Something must be wrong, or this won't beat the 5600x in gaming.

Oooooooor you could understand that's totally irrelevant, as the cache doesn't help cinebench performance at all.

 

[The Quim Reaper]

The tests were done with 3200MHz CL14 RAM, so that could account for some of it, but not much.

Cinebench scores aren't really affected by RAM speed to any great degree.

 

[Cowboystrekk]

A few apps can utilize large L3 cache, most notable some mining software, but in general gaming is where cache matters most.

 

[Wavetrex]

Not expecting any miracles from this CPU.
At best I would consider it a beta-test for future technologies...

 

[mb194dc]

Single core Cinenench R23 (by my opinion most indicative of gaming speeds here) is crap, slower than stock 5800x. Something must be wrong, or this won't beat the 5600x in gaming.

The extra cache doesn't help in Cinebench because rendering an image is a completely different workload to any game engine?

Pretty obviously, extra cache will only help workloads where the same instructions are continually repeated and enough of them will fit in the cache to give that advantage.

Any benefit will be game and engine specific. What code will fit in the extra 64mb l3 that would have come from RAM before?

 

[Bwaze]

Why would gaming be something uniquely affected by cache size? I'd think it would be similar to higher RAM speeds - relevant in some cases (especially min frames), largely unimportant in most cases?

Will we see games that won't benefit, or be even slower due to slower core clock, and some that will benefit more? And reviewer coul then pick and choose which message he wants to display?

 

[The King]

Why would gaming be something uniquely affected by cache size? I'd think it would be similar to higher RAM speeds - relevant in some cases (especially min frames), largely unimportant in most cases?

Will we see games that won't benefit, or be even slower due to slower core clock, and some that will benefit more? And reviewer coul then pick and choose which message he wants to display?

This video should answer your question.

Interesting note in the F36c bios on that mobo.

Source website

Exclusivo: Filtración de primeros benchmarks del Ryzen 7 5800X3D

Benchmarks del procesador AMD Ryzen 7 5800X3D. Cinebench R23, Geekbench 5 entre otros cosas más adicionales en la siguiente nota.

xanxogaming.com

 

[Bwaze]

Ah, that guy. Remember his explanations of PBO +200 MHz overclocking? How did that work out?

 

[TheLostSwede]

Ah, that guy. Remember his explanations of PBO +200 MHz overclocking? How did that work out?

It did, after some time.

Cinebench scores aren't really affected by RAM speed to any great degree.

Cinebench wasn't the only benchmark they ran though.

 

[Chomiq]

Ah, that guy. Remember his explanations of PBO +200 MHz overclocking? How did that work out?

I will take TechTechPotato and Wendell's opinion on extra cache, which pretty much sums up to:
- benefits gaming,
- no major increase in consumer workloads,
- benefits in enterprise workloads (depending on use case).

Granted, no one will use 5800X3D for enterprise workloads, but that's where Milan-X steps in, which is already used in the industry.

But as usual, wait for actual gaming benchmarks.

 

[Bwaze]

Funny that "3200MHz CL14 RAM" is here noted as a downside, as reviewer could have used much quicker RAM.

That's Samsung B die. Was quite expensive, and can be easily overclocked to maximum Ryzen can even handle in 1:1. How many Ryzen users have quicker RAM? And how much does it even matter in benchmarks that do scale - as oposed to Cinebench, also most games outside fringe cases, min frames or ultra high framerate?

And no, "PBO +200MHz overclocking" never worked as AMD explained in that video. A lot of time and resources was spent to even get Ryzen 3000 processors to achieve marketed boost clocks (even for just a fraction of a second), forget achieving anything higher in a meaningful way.

 

[Tsukiyomi91]

another chip to milk the already dried up AM4 socket.

 

[Mussels]

Funny that "3200MHz CL14 RAM" is here noted as a downside, as reviewer could have used much quicker RAM.

That's Samsung B die. Was quite expensive, and can be easily overclocked to maximum Ryzen can even handle in 1:1. How many Ryzen users have quicker RAM? And how much does it even matter in benchmarks that do scale - as oposed to Cinebench, also most games outside fringe cases, min frames or ultra high framerate?

And no, "PBO +200MHz overclocking" never worked as AMD explained in that video. A lot of time and resources was spent to even get Ryzen 3000 processors to achieve marketed boost clocks (even for just a fraction of a second), forget achieving anything higher in a meaningful way.

3200C14 is damn near king of the hill - look how well 3600C16 does in the AL testing
DDR5 Memory Performance Scaling with Alder Lake Core i9-12900K - Game Performance 720p | TechPowerUp

And true, 3000 series never boosted that high for long. Zen 3 sure did, easily.

It's an overclocking feature for tweakers, it was never advertised by amd as "+200Mhz all the time" or anything

 

[Frick]

another chip to milk the already dried up AM4 socket.

Or more upgrade paths for folks still on older chipsets.

 

[Chrispy_]

Those numbers are below my 5800X, which is unsurprising given the clockspeed deficit.

AMD themselves said that the biggest beneficiary of all that extra cache is gaming. I expect things like rendering and compiles to get negligible benefit as those things are generally not cache-limited in any way on the existing 5800X.

 

[Mussels]

Those numbers are below my 5800X, which is unsurprising given the clockspeed deficit.

AMD themselves said that the biggest beneficiary of all that extra cache is gaming. I expect things like rendering and compiles to get negligible benefit as those things are generally not cache-limited in any way on the existing 5800X.

It's looking like a 5700x, in anything the cache doesn't actively help

 

[Cowboystrekk]

Funny that "3200MHz CL14 RAM" is here noted as a downside, as reviewer could have used much quicker RAM.

That's Samsung B die. Was quite expensive, and can be easily overclocked to maximum Ryzen can even handle in 1:1. How many Ryzen users have quicker RAM? And how much does it even matter in benchmarks that do scale - as oposed to Cinebench, also most games outside fringe cases, min frames or ultra high framerate?

And no, "PBO +200MHz overclocking" never worked as AMD explained in that video. A lot of time and resources was spent to even get Ryzen 3000 processors to achieve marketed boost clocks (even for just a fraction of a second), forget achieving anything higher in a meaningful way.

On ryzen 5000 +200 pbo works perfectly On 3000 not so much.

 

[GerKNG]

Cinebench is not that sensitive to cache or ram. you can go from a weak 3600 CL16 XMP Kit to a fine tuned 4000 CL14 Kit and the score barely changes. same with strong cache OCs or more cache in general.
i guess it'll be as fast as a 5800X at 5Ghz all core in games.

 

[Denver]

Single core Cinenench R23 (by my opinion most indicative of gaming speeds here) is crap, slower than stock 5800x. Something must be wrong, or this won't beat the 5600x in gaming.

AMD EPYC 7773X "Milan-X" Benchmarks Show Very Strong HPC Performance Upgrade Review - Phoronix

www.phoronix.com

The performance gain is not in all tasks. But the Milan-X benchmarks have shown that there are substantial gains, it remains to be seen if any of these apply to our use.