News Unannounced Ryzen 9 9950X3D dominates Ryzen 7 9800X3D in Factorio benchmark – Ryzen 9000X3D flagship up to 18% faster than current fastest gaming CPU

[Admin]

The 16-core Ryzen 9 9950X3D is much faster than the 8-core 9800X3D according to one benchmark site that has a result for the as-of-yet unannounced chip.

Unannounced Ryzen 9 9950X3D dominates Ryzen 7 9800X3D in Factorio benchmark – Ryzen 9000X3D flagship up to 18% faster than current fastest gaming CPU : Read more

 

[-Fran-]

The dual VCache'd CCD cope has begun!

Regards

 

[SonoraTechnical]

The dual VCache'd CCD cope has begun!

Regards

Is that true? I thought it was like the previous X3D generation where 8 cores got the cache and 8 didn't. If they truly have the V-Cache on both of the 8 core CCDs that is a major change.

 

[SonoraTechnical]

modified the above comment to reflect surprise if V-Cache were present on BOTH of the 8 core CCDs...

 

[AngelusF]

modified the above comment to reflect surprise if V-Cache were present on BOTH of the 8 core CCDs...

Wasn't the point of having the second CCD without VCache that the extra cache silicon on top of the CPUs hobbled them so much? That's no longer the case, so surely the second CCD will have VCache as well now, won't it? Otherwise you run into the same old problems with Windows having a hard time trying to allocate threads to the appropriate CCD.

 

[torbjorn.lindgren]

Is that true? I thought it was like the previous X3D generation where 8 cores got the cache and 8 didn't. If they truly have the V-Cache on both of the 8 core CCDs that is a major change.

It's not been confirmed AFAIK but it's a frequent speculation/leak and there's been lots of hints that AMD could choose to do this.

Remember that the reason why AMD combined one core with and one core without V-Cache in the older 12/16-core models was because adding the V-Cache on top really hurt the thermals and thus frequency of the chip with V-Cache. Hence AMD tried to split the difference so the multi-die cpus had both some high frequency cores for productivity apps and some high cache cores for gaming.

Managing these dissimilar cores requires careful juggling and sometimes goes wrong (similar problems as having E and P-cores) which is one of the reasons why 7800X3D tended to be be the gaming champion.

With 9800X3D AMD has demonstrated that this limitation is mostly gone so reason for it is gone - and thus a 9950X3D where both dies have V-Cache is very likely to dominate one where only one has it in pretty much every workload (even the ones prefer frequency over cache is likely to be better off with frequency and cache!).

Yes, using two V-Cache enabled dies in the 12/16-core cpus will cost AMD a bit more to produce but the 12 & 16-core models are the consumer CPUs with the highest profit margin for AMD so shortchanging these users would be outright stupid by AMD. Not that AMD (and Intel) hasn't made plenty of stupid decisions, but, well, lets hope sanity prevails.

 

[Makaveli]

Why did you remove the 5800X3D numbers which show it producing higher performance than the 7800X3D. The source articles says this benchmark isn't reliable.

 

[mac_angel]

why not show Intel's for some sort of comparison?

 

[salgado18]

why not show Intel's for some sort of comparison?

At this point, Intel is almost irrelevant in gaming

 

[Mama Changa]

Fingers crossed the 9900X3D is now going to be a good gaming cpu as well as show the same uplifts as 9800X3D did in productivity and is as strong at least as the 9900X. 9800X3D is on average faster than 9700X in productivity, which other than price sort of renders the 9700X useless. I would gladly pay the extra for the 9800x3D and in 3 months AMD will drop the price $50 once the pent-up demand has eased. This is what happened with 7800X3D and this time you'll have options for gaming like the 9950X3D and possibly 9900X3D if you also do a lot of productivity work too. So demand could be shared across the entire X3D range this time.

 

[Gillerer]

The reason the 7950X3D didn't have V-Cache on both CCDs was that it didn't give better performance (as per AMD's internal testing), but not (mainly) because of lower frequency, but the added latency from CCD-to-CCD hops.

If a game ever gets scheduled on cores on two different CCDs, the two hops thru the I/O die will kill performance, so you've already lost. Double the V-Cache could *maybe* scrape back some of the deficit, but would be sure to make the product more expensive. They thought it better to just try to do some trickery to make sure games are scheduled on a single CCD.

 

[YSCCC]

with ARL being basically a flop, I can foresee it being priced unobtainably for any sane person..

 

[ilukey77]

This is why ive waited .. ive bought the 5800x3d and the 7800x3d at launch this time ive waited for the 9950x3d looks like a smart choice!!

 

[andrewoid]

256GB at DDR5-6000? I call fake news.

 

[TheHerald]

why not show Intel's for some sort of comparison?

At this point, Intel is almost irrelevant in gaming

Intel is at the top of the factorio benchmarks in 50k maps. Which is what matters, because the game is locked at 60 ups / fps anyways, the goal in factorio is to run the biggest base possible while maintaining 60 fps. RPL and Alderlake are dominating there.

 

[YSCCC]

Intel is at the top of the factorio benchmarks in 50k maps. Which is what matters, because the game is locked at 60 ups / fps anyways, the goal in factorio is to run the biggest base possible while maintaining 60 fps. RPL and Alderlake are dominating there.

cough, and then IIRC you are eyeing on the 9950X3D...

 

[TheHerald]

cough, and then IIRC you are eyeing on the 9950X3D...

So? Im not playing factorio so I don't really care that Intel is dominating there.

 

[mac_angel]

Intel is at the top of the factorio benchmarks in 50k maps. Which is what matters, because the game is locked at 60 ups / fps anyways, the goal in factorio is to run the biggest base possible while maintaining 60 fps. RPL and Alderlake are dominating there.

I kinda suspected that would be the case with Intel having many more physical cores. Which is why I believe they should have put one or two Intel CPUs for reference.

 

[TheHerald]

I kinda suspected that would be the case with Intel having many more physical cores. Which is why I believe they should have put one or two Intel CPUs for reference.

It's not about the cores, memory latency.