AMD Ryzen 7 5800X3D Geekbenched, About 9% Faster Than 5800X

[btarunr]

Someone with access to an AMD Ryzen 7 5800X3D processor sample posted some of the first Geekbench 5 performance numbers for the chip, where it ends up 9% faster than the Ryzen 7 5800X, on average. AMD claimed that the 5800X3D is "the world's fastest gaming processor," with the 3D Vertical Cache (3D V-cache) technology offering gaming performance uplifts over the 5800X akin to a new generation, despite being based on the same "Zen 3" microarchitecture, and lower clock speeds. The Ryzen 7 5800X3D is shown posting scores of 1633 points 1T and 11250 points nT in one run; and 1637/11198 points in the other; when paired with 32 GB of dual-channel DDR4-3200 memory.

These are 9% faster than a typical 5800X score on this benchmark. AMD's own gaming performance claims see the 5800X3D score a performance uplift above 20% over the 5800X, closing the gap with the Intel Core i9-12900K. The 3D V-cache technology debuted earlier this week with the EPYC "Milan-X" processors, where the additional cache provides huge performance gains for applications with large data-sets. AMD isn't boasting too much about the multi-threaded productivity performance of the 5800X3D because this is ultimately an 8-core/16-thread processor that's bound to lose to the Ryzen 9 5900X/5950X, and the i9-12900K, on account of its lower core-count.



View at TechPowerUp Main Site | Source

 

[Taraquin]

I'm impressed it helped that much in GB. Gaming should be even better!

 

[SL2]

Oh no, this is almost flamebait for those who doesn't know (or doesn't care) how Gbench works (or doesn't work).

Incoming bitter alarmists any second now, who feel backstabbed lol.

Nobody ever said X3D would raise performance equally across any workload.

I'm impressed it helped that much in GB. Gaming should be even better!

Yeah, a lower score than that wouldn't have surprised me.

 

[Taraquin]

Oh no, this is almost flamebait for those who doesn't know (or doesn't care) how Gbench works (or doesn't work).

Incoming bitter alarmists any second now, who feel backstabbed lol.

Nobody ever said X3D would raise performance equally across any workload.


Yeah, a lower score than that wouldn't have surprised me.

Considering it runs lower clock and GB don't care that much about cache I'm impressed

 

[fevgatos]

I'm impressed it helped that much in GB. Gaming should be even better!

Gb is more cache and memory sensitive than any game though

 

[Taraquin]

Gb is more cache and memory sensitive than any game though

Hmm, didn't know it was very cache sensitive, but considering 5800X3D runs a few hundred MHz slower and vs 5800X it's not bad.

 

[DeathtoGnomes]

Gauge with anything but geekbench!



Wanna bet its an engineering sample? We dont know any facts, nor what build its in, so shouldnt be taken as fact, wait for the real reviews.

 

[fevgatos]

Hmm, didn't know it was very cache sensitive, but considering 5800X3D runs a few hundred MHz slower and vs 5800X it's not bad.

I think cache and memory performance is always related, although in this case the 5800x3d tries to bypass the memory latency issues of Zen 3 with a big ass cache.

I hope it will fix the issues that zen 3 had in gaming, basically the performance was...let's say, very inconsistent. For example, since I play a lot of warzone, there are areas in the game that the 5950x obliterated everything getting 400+ fps (while my 10900k got around 300-320), and then there are areas where the 5950x couldn't even much an 8700k at 150 fps, while my 10900 was getting 200-220.

 

[Garrus]

Gb is more cache and memory sensitive than any game though

Not true at all. With Geekbench 5 they removed the memory sensitive tests from GB 4 etc. You are not correct. Anyways I don't care about GB, I care about gaming.

 

[Taraquin]

I think cache and memory performance is always related, although in this case the 5800x3d tries to bypass the memory latency issues of Zen 3 with a big ass cache.

I hope it will fix the issues that zen 3 had in gaming, basically the performance was...let's say, very inconsistent. For example, since I play a lot of warzone, there are areas in the game that the 5950x obliterated everything getting 400+ fps (while my 10900k got around 300-320), and then there are areas where the 5950x couldn't even much an 8700k at 150 fps, while my 10900 was getting 200-220.

This is more of an issue with the 2 CCDs partly due to higher latency\inter CCD latency which in some games result in lower 1% lows. 5600X and 5800X had less of an issue with this and was more consistent on lows.

You can see it here in Warzone on heavily tuned CPUs:

19:45

12:32

5800X is more consistent on avg vs lows vs 10700K than 5950X is vs 10900K and 12900K.

In general though there are very few games where Zen3 struggles with 1% lows although 2 CCDs tends to have worse lows vs Intel and 1 CCDs. In some games Intel fare far worse, check Troy at 17:40.

 

[Bwaze]

"These are 9% faster than a typical 5800X score on this benchmark. AMD's own gaming performance claims see the 5800X3D score a performance uplift above 20% over the 5800X, closing the gap with the Intel Core i9-12900K."

Isn't this uplift in Geekbench mainly in multicore? Why would we see an ever greater uplift in gaming, which mainly cares for single / low core speed?

Single core result actually seems to be even lower than standard 5800X:

5800X3D: 1637 single-core, 11250 multi-threaded points.

5800X: 1671 points single-core, 10333 points in the multi-core tests.

 

[Taraquin]

"These are 9% faster than a typical 5800X score on this benchmark. AMD's own gaming performance claims see the 5800X3D score a performance uplift above 20% over the 5800X, closing the gap with the Intel Core i9-12900K."

Isn't this uplift in Geekbench mainly in multicore? Why would we see an ever greater uplift in gaming, which mainly cares for single / low core speed?

Due to higher cache. Cache is king in most games. Generally higher cache matters more for fps than how many cores you have

 

[fevgatos]

This is more of an issue with the 2 CCDs partly due to higher latency\inter CCD latency which in some games result in lower 1% lows. 5600X and 5800X had less of an issue with this and was more consistent on lows.

You can see it here in Warzone on heavily tuned CPUs:

19:45

12:32

5800X is more consistent on avg vs lows vs 10700K than 5950X is vs 10900K and 12900K.

In general though there are very few games where Zen3 struggles with 1% lows although 2 CCDs tends to have worse lows vs Intel and 1 CCDs. In some games Intel fare far worse, check Troy at 17:40.

That was a really detailed video, wow. Although tbf I don't agree with the results, especially on warzone, but I guess it might the 3600c14 ram, I used to run 4400c16 super tight on my 10900k, so that might be the difference maker.

Isn't this uplift in Geekbench mainly in multicore? Why would we see an ever greater uplift in gaming, which mainly cares for single / low core speed?

Games care about memory / cache latency first, and then single thread performance. For example tuning your ram gives you more fps than overclocking your CPU.

 

[Richards]

So much cache but little uplift.. shows l2 cache is more important thats why intel will increase it for raptor lake

 

[Taraquin]

That was a really detailed video, wow. Although tbf I don't agree with the results, especially on warzone, but I guess it might the 3600c14 ram, I used to run 4400c16 super tight on my 10900k, so that might be the difference maker.


Games care about memory / cache latency first, and then single thread performance. For example tuning your ram gives you more fps than overclocking your CPU.

Results will vary depending on tuning, what route they use on the map etc. With your settings you got a different result. Unless we could use a built in bench it`s hard to compare, but i2hard is a good review-channel since the test both stock and properly tuned (I know there is some headroom on both Intel and AMD, but the tuning they apply is generally very good in my opinion). Recommend to watch that channel, lots of good test of different hardware if you had run 3800cl14\15 tuned on the 5950X it would have been closer I bet, faster ram with tighter timings help on lows a lot, but maybe 10900K would still be best on lows

Some games care a lot about latency, some games prefer bandwith. Watch the other gametests on i2hard. Cyberpunk and Troy don`t care about latency, but loves BW. DDR5 which has bad latency destroys DDR4, but check out Star craft 2, it loves latency and DDR4 is much better than DDR5

 

[Bwaze]

"Cache is king in most games. Generally higher cache matters more for fps than how many cores you have"

"Games care about memory / cache latency first, and then single thread performance."

Any proof of this? I've seen a lot of Ryzen overclocking, optimisation videos, and while you can benefit from lower memory latency and higher memory speed, it's usually very cherry picked results, like 0.1% lows in a couple of carefully selected games. Any benchmarking with wider array of games has shown very little performance uplift from cheap RAM and stock Infinity fabric clock to ultra expensive ones and with huge OC in FCLK.

 

[Space Lynx]

@btarunr a 5900x is currently on sale for $445 ($5 cheaper than 5800x3d), so would that match or beat a 5800x3d in gaming still you think?

 

[Taraquin]

"Cache is king in most games. Generally higher cache matters more for fps than how many cores you have"

"Games care about memory / cache latency first, and then single thread performance."

Any proof of this? I've seen a lot of Ryzen overclocking, optimisation videos, and while you can benefit from lower memory latency and higher memory speed, it's usually very cherry picked results, like 0.1% lows in a couple of carefully selected games. Any benchmarking with wider array of games has shown very little performance uplift from cheap RAM and stock Infinity fabric clock to ultra expensive ones and with huge OC in FCLK.

It depends on the game and what the game scales with. Many newer games like Troy and Cyberpunk prefers BW over latency. Many older games like latency more.

@btarunr a 5900x is currently on sale for $445 ($5 cheaper than 5800x3d), so would that match or beat a 5800x3d in gaming still you think?

No, in some games they will be close, in other games 5800X3D will be 20% faster, AMD compared 5900X with 5800X3D in their marketingslides.

 

[Fouquin]

So much cache but little uplift.. shows l2 cache is more important thats why intel will increase it for raptor lake

Remember when generational improvements between entirely different architectures could barely scrape together a 9% improvement? I do. This chip isn't even a new architecture, it's literally a downclocked Zen 3 with a cache slice glued on, and it's putting up measurable improvements. Whatever fantasy land you want to live in doesn't negate that this strategy clearly works.

 

[Bwaze]

I'd really wait for the gaming benchmarks.

It makes no sense to be expecting a 20% gaming uplift when single core Geekbench score (the result that usually quite well represents speed in games) shows no uplift, even regression.

 

[Space Lynx]

I'd really wait for the gaming benchmarks.

It makes no sense to be expecting a 20% gaming uplift when single core Geekbench score (the result that usually quite well represents speed in games) shows no uplift, even regression.

yeah I am in no rush, I will just wait for the reviews and go from there.

 

[Dragokar]

So I might be able to run UT99 GOTY inside the X3D Cache alone

 

[fevgatos]

"Cache is king in most games. Generally higher cache matters more for fps than how many cores you have"

"Games care about memory / cache latency first, and then single thread performance."

Any proof of this? I've seen a lot of Ryzen overclocking, optimisation videos, and while you can benefit from lower memory latency and higher memory speed, it's usually very cherry picked results, like 0.1% lows in a couple of carefully selected games. Any benchmarking with wider array of games has shown very little performance uplift from cheap RAM and stock Infinity fabric clock to ultra expensive ones and with huge OC in FCLK.

Because most of the benchmarks will be GPU limited. If you want to see the impact of ram go down to 720p (or even lower depending on the game) and test. For example in cyberpunk 2077 with a 3090 @ 720 ultra settings + RT and DLSS ultra performance I go from 170 to 247 fps in the exact same scene just from 4800 stock ram no xmp to 6000c32.

 

[napata]

I'd really wait for the gaming benchmarks.

It makes no sense to be expecting a 20% gaming uplift when single core Geekbench score (the result that usually quite well represents speed in games) shows no uplift, even regression.

It doesn't though. Just look at the 11900k vs 10900k. The former has a much higher single core score but this doesn't translate to games at all. The reason is cache & latency.

 

[birdie]

Faster?

Here's a comparison with a bog standard non-OC'ed 5800X.