AMD Ryzen 9 9950X3D Carries 3D V-Cache on a Single CCD, 5.6 GHz Clock Speed, and 170 Watt TDP

[AleksandarK]

Recent engineering samples of AMD's upcoming Ryzen 9 9950X3D reveal what appear to be the finalized specifications of the top-tier AM5 chip. The 16-core, 32-thread processor builds upon the gaming success of the Ryzen 7 9800X3D while addressing its core count limitations. The flagship processor features AMD's refined cache design, combining 96 MB of 3D V-Cache with 32 MB of standard L3 cache. Unlike its predecessor, the 7950X3D, the new Zen 5 architecture incorporates a redesigned CCD stacking method. The CCD now sits above the cache, directly interfacing with the STIM and IHS, eliminating thermal constraints that previously required frequency limitations. The processor features asymmetric cache distribution across its dual CCDs—one die combines 32 MB of base L3 cache with a 64 MB stacked V-Cache layer, while its companion die utilizes a standard 32 MB L3 cache configuration. In total, there is a 128 MB of L3 cache, with 16 MB of L2.

This architectural advancement enables the 9950X3D to achieve a 5.65 GHz boost clock across both CCDs, matching non-X3D variants. The processor maintains a 170 W TDP, suggesting improved thermal efficiency despite the additional cache. AMD's software-based OS scheduler will continue to optimize gaming workloads by directing them to the CCD with 3D V-Cache. Early leaks indicate the 9950X3D matches the base 9950X in Cinebench R23 scores, both in single and multi-threaded tests—a significant improvement over the 7950X3D, which lagged behind its non-X3D counterpart due to frequency limitations. AMD plans to expand the Zen 5 X3D lineup in Q1-2025 with both the 9950X3D and 9900X3D models. Full performance benchmarks and pricing details are expected at CES 2025, where AMD will officially unveil these processors alongside their RDNA 4 GPUs.



View at TechPowerUp Main Site | Source

 

[Cheeseball]

@AleksandarK

I probably won't upgrade to this one until after 1 or 2 years. If this was like a 30% performance jump (its not, comparing 7800X3D to 9800X3D), then sure, but not $700 sure.

 

[Vincero]

Sigh.... Still with the assymentric cache arrangement....

Was kind of hoping they would do that as a 'lower' tier Ryzen 9 part and release a dual-CCD X3D top-tier Ryzen 9 CPU....

 

[AleksandarK]

@AleksandarK

View attachment 377301

I probably won't upgrade to this one until after 1 or 2 years. If this was like a 30% performance jump (its not, comparing 7800X3D to 9800X3D), then sure, but not $700 sure.

Not sure what are you pointing to? My rendering looks fine?

 

[sephiroth117]

Sigh.... Still with the assymentric cache arrangement....

Was kind of hoping they would do that as a 'lower' tier Ryzen 9 part and release a dual-CCD X3D top-tier Ryzen 9 CPU....

It's not as simple, there are limitations still on 3D CCD, even if the 2nd gen X3D are much better, heat less etc, they still want one CCD for fast Ghz and the other for gaming/3D applications

There's a reason, also with dual CCD, would having 32+32MB be as good as one CCD with 64MB 3D extra L3 cache ? if no, wouldn't 64+64 be too expensive ?

I think there are genuine cost and technological obstacles for dual CCD, it's not just them wanting to add a software director and more complexity, maybe further down the line

 

[Carillon]

Not sure what are you pointing to? My rendering looks fine?

probably typo in the cpu name.
I'll be very disappointed if this turns out to be the final sku

 

[AleksandarK]

probably typo in the cpu name.

Yeah, thanks for pointing out!

 

[evernessince]

It's not as simple, there are limitations still on 3D CCD, even if the 2nd gen X3D are much better, heat less etc, they still want one CCD for fast Ghz and the other for gaming/3D applications

There's a reason, also with dual CCD, would having 32+32MB be as good as one CCD with 64MB 3D extra L3 cache ? if no, wouldn't 64+64 be too expensive ?

I think there are genuine cost and technological obstacles for dual CCD, it's not just them wanting to add a software director and more complexity, maybe further down the line

Read the article, the CPU boosts to top clocks on both CCDs this time around. X3D no longer limits frequency or heat.

This architectural advancement enables the 9950X3D to achieve a 5.65 GHz boost clock across both CCD

I suspect the reason AMD is not releasing a dual X3D chip is a lack of competition from Intel.

 

[Cheeseball]

Yeah, thanks for pointing out!

I forgot to circle it with the Snipping Tool before I replied. My bad.

 

[Dr. Dro]

Guess it's another generation of stopping at 8 cores to get the full benefits of your purchase. I was almost excited for this when the first rumors that it would finally be dual X3D emerged. And of course, the 9900X3D is set up to be another remarkably poor processor if you happen to do video games.

 

[sephiroth117]

Read the article, the CPU boosts to top clocks on both CCDs this time around. X3D no longer limits frequency or heat.



I suspect the reason AMD is not releasing a dual X3D chip is a lack of competition from Intel.

Not really, it's easy to check on TPU even since they have the reviews: Whilst some overclocked the 9800X3D to achieve slightly higher clocks, the 9700X can reach up to 5.5Ghz boost clock, vs the advertised 5.2Ghz max boost clock of the 9800X3D.

There are still some limits like I said, they are not as big as gen 1, but there are still differences between 3D and non-3D clearly.

 

[Vincero]

It's not as simple, there are limitations still on 3D CCD, even if the 2nd gen X3D are much better, heat less etc, they still want one CCD for fast Ghz and the other for gaming/3D applications

There's a reason, also with dual CCD, would having 32+32MB be as good as one CCD with 64MB 3D extra L3 cache ? if no, wouldn't 64+64 be too expensive ?

I think there are genuine cost and technological obstacles for dual CCD, it's not just them wanting to add a software director and more complexity, maybe further down the line

By that same token, even the single die CCD X3D parts are actually not great - outside of gaming and cache limited / memory bandwidth / thread limited scenarios the higher boosting higher TDP normal CCD CPUs win in average productivity.

Anyway, as for the dual X3D part.... Someone will always pay for it even if the benefit is only 1% (or even less) over the next competitive product in the lineup, witness the 14900KS...

 

[evernessince]

Not really, it's easy to check on TPU even since they have the reviews: Whilst some overclocked the 9800X3D to achieve slightly higher clocks, the 9700X can reach up to 5.5Ghz boost clock, vs the advertised 5.2Ghz max boost clock of the 9800X3D.

There are still some limits like I said, they are not as big as gen 1, but there are still differences between 3D and non-3D clearly.

You attribute to limits that which can be explained via product segmentation or other sources.

Heat dissipation wise 2nd gen X3D cache enabled chips are now identical to non-X3D chips:

RIP Intel: AMD Ryzen 7 9800X3D CPU Review & Benchmarks vs. 7800X3D, 285K, 14900K, & More | GamersNexus

CPUs RIP Intel: AMD Ryzen 7 9800X3D CPU Review & Benchmarks vs. 7800X3D, 285K, 14900K, & More November 7, 2024 Last Updated: 2024-12-03 We test the 9800X3D’s frequency, thermals, gaming performance, efficiency, and more The Highlights The 9800X3D is a follow up to the 7800X3D but runs on Zen 5...

gamersnexus.net

View attachment 1735242643211.webp

The above chart demonstrates that at a given heat load they are now about the same and thus the single limitation has been removed. Frequency has thus scaled according to that.

Lower binned parts having lower frequencies isn't evidence that said heat limit still exists, it merely demonstrates the obvious in the PC world that lower end parts often have lower frequency.

 

[RogueSix]

I suspect the reason AMD is not releasing a dual X3D chip is a lack of competition from Intel.

While there might be some product policy considerations, I think that technical hurdles are still a thing. As far as product policy, it might not be so much no competition from Intel but the fact that AMD wants to sell dual 3D cache Threadrippers. Who would buy a pro platform if you can have the same on a much cheaper consumer platform? Right. No one. That is probably the main reason why AMD is keeping dual 3D Cache Threadripper (and above) exclusive.

The technical hurdles for gaming (as opposed to the professional stuff with the Threadrippers) are not to be underestimated, however. Inter CCD latencies are still a thing and you would have to make sure that the scheduler always "fills up" a single CCD first (eight cores) before switching to the second CCD if more cores are needed (which will be the case in less than 1% of games btw).

AMD is doing the right thing here imo. A dual cache 99xx CPU would be very expensive and the real world returns would be minimal or even detrimental, depending on the game. The most critical factor for gaming is inter CCD latencies and you can only mitigate those latencies to a minor extent by having cache on both CCDs. You would still want any game to utilize only a single CCD for as long as possible before spilling over onto CCD #2.
The challenges with regard to scheduling would, in fact, become even more difficult. With 7950X3D it is "easy" because the 2nd CCD is simply put to sleep (core parking) but if you had dual cache on a 9950X3D, well, wow... that would open an all new palette of cans of worms when it comes to scheduling .

Games remain thread-limited. You can't just throw more cores at the problem and hope that all tasks are automagically split up between the cores. That's not how it works. Multithreading is very complex and most of the time other threads are waiting for one thread to finish their work.

Anyone who believes that they need more cores for gaming (should be very few people as only very few games effectively utilize more than four to six cores) is better off waiting for AMD to put more cores onto their CCDs. We'll have to wait and see if they will go straight from 8 to 16 or if there will be an intermediary step but for gaming that is the (our) ticket to success.
Dual CCDs will always remain a compromise for a home/gaming plus light productivity setup. Anyone with serious needs will go Threadripper/EPYC and anyone who is just gaming will always be better off with a single CCD CPU (9800X3D/7800X3D at the high end especially).

 

[john_]

Guess it's another generation of stopping at 8 cores to get the full benefits of your purchase.

It's worst with Intel from this generation and who knows how many more. 8 P cores and NO Hyperthreading.

 

[Bruno Vieira]

disappointing

 

[ZoneDymo]

boring af, well, here is waiting for the next generation

 

[A Computer Guy]

Sigh.... Still with the assymentric cache arrangement....

Was kind of hoping they would do that as a 'lower' tier Ryzen 9 part and release a dual-CCD X3D top-tier Ryzen 9 CPU....

On the flipside they have left themselves an opening for a future SKU 9952X3D

X3D only for 1 CCD blah, so are they still paring 1 good CCD and 1 worse CCD because that is the real crime.

 

[bonehead123]

Well, I for one won't be buying it, simply because the TPD is less than 600w....hehehehe

Give me hot or give me not !




j/k
/s

 

[dgianstefani]

It's worst with Intel from this generation and who knows how many more. 8 P cores and NO Hyperthreading.

FYI Zen 5 performs better in games with SMT off, so I wouldn't be so quick to criticise Intel for being ahead of the curve.

 

[Wirko]

You would still want any game to utilize only a single CCD for as long as possible before spilling over onto CCD #2.

That's true as long as the game runs in a single process. I'm wondering if any games are split into multiple processes with the purpose of overcoming the limits that slow inter-CCD communication imposes.

Multithreading is very complex and most of the time other threads are waiting for one thread to finish their work.

That one critical thread should be left to run alone on a core and not be burdened by SMT - but I'm not sure there are provisions in Windows to make that possible.

 

[evernessince]

FYI Zen 5 performs better in games with SMT off, so I wouldn't be so quick to criticise Intel for being ahead of the curve.

AMD looses some 2.6% gaming performance with SMT off. Not a crazy amount.

Someone really needs to do a performance per square mm (factoring in just the core size) to figure out how efficient Intel's big little is as compared to AMD. Should probably do one for AMD's dense cores as well.

 

[lexluthermiester]

FYI Zen 5 performs better in games with SMT off, so I wouldn't be so quick to criticise Intel for being ahead of the curve.

True!

AMD looses some 2.6% gaming performance with SMT off. Not a crazy amount.

That depends on the game.

 

[evernessince]

That depends on the game.

It always does, goes without saying.

 

[Geofrancis]

If they were going to do a dual 3d cache chip I suspect it would be launched as a AM5 EPYC rather than a Ryzen, then the price could be justified.