Monday, December 23rd 2024
AMD Ryzen 9 9950X3D and 9900X3D to Come with Clock Speeds Resembling non-X3D SKUs
The AMD Ryzen 7 9800X3D launched this November remains the fastest processor for PC gaming, but with just an 8-core/16-thread configuration, its performance in multithreaded productivity pales in comparison to the alternatives, including from AMD's own camp, such as the Ryzen 9 7950X3D. The company is planning to expand the Zen 5 X3D line of desktop processors in Q1-2025, with the introduction of its high core-count Ryzen 9 series SKUs, the Ryzen 9 9950X3D, and the Ryzen 9 9900X3D. HXL, a reliable source with hardware leaks, says that the Ryzen 9 X3D SKUs will lack the kind of "frequency debuff" we've seen with the older generation Ryzen 9 X3D chips such as the 7950X3D and 7900X3D.
On the 7950X3D, only one of the two 8-core CCDs comes with 3D V-Cache, the other is a regular CCD with 32 MB on-die L3 cache. The CCD with the 3D V-Cache has its frequency "debuffed" compared to the other CCD, mainly to deal with the thermal limitations of the way the 3D V-Cache is stacked on top of the CCD. With the Ryzen 9000 X3D generation, AMD has redesigned this CCD + L3D stacking such that the CCD is now above, directly interfacing with the STIM and IHS; with no conductive structural silicon along the way. This means that the CCD now thermally behaves like a regular Zen 5 CCD; and for this reason, not only will the CCD with 3D V-Cache on the 9950X3D/9900X3D have the same clock speeds and boosting behavior as the one without 3D V-Cache; but also the 9950X3D and 9900X3D are expected to ship with similar, if not identical clock speeds to the 9950X and 9900X.
It is highly likely that AMD will retain the software-based OS scheduler management it's been using on the 9950X and 9900X, which ensures that gaming workloads are allocated to the CCD with 3D V-Cache. On multithreaded productivity workloads that can scale up to 32 threads, this scheduling could be more relaxed compared to the one on the 7950X3D, since both CCDs will boost up to similar frequencies. HXL also dropped a hint about possible performance of the 9950X3D in Cinebench R23, a benchmark that scales well across 16 cores, but doesn't quite benefit from 3D V-Cache (frequency matters). HXL says that the 9950X3D ends up with nearly identical 1T and nT scores to the 9950X. If you recall, the 7950X3D falls slightly behind the 7950X due to frequency debuffing on its CCD with 3D V-Cache.
Sources:
HXL (Twitter), HXL (Twitter), VideoCardz
On the 7950X3D, only one of the two 8-core CCDs comes with 3D V-Cache, the other is a regular CCD with 32 MB on-die L3 cache. The CCD with the 3D V-Cache has its frequency "debuffed" compared to the other CCD, mainly to deal with the thermal limitations of the way the 3D V-Cache is stacked on top of the CCD. With the Ryzen 9000 X3D generation, AMD has redesigned this CCD + L3D stacking such that the CCD is now above, directly interfacing with the STIM and IHS; with no conductive structural silicon along the way. This means that the CCD now thermally behaves like a regular Zen 5 CCD; and for this reason, not only will the CCD with 3D V-Cache on the 9950X3D/9900X3D have the same clock speeds and boosting behavior as the one without 3D V-Cache; but also the 9950X3D and 9900X3D are expected to ship with similar, if not identical clock speeds to the 9950X and 9900X.
36 Comments on AMD Ryzen 9 9950X3D and 9900X3D to Come with Clock Speeds Resembling non-X3D SKUs
Same as 7950X3D.
I think there's little to no point to having the v-cache on both CCDs. If both CCDs have v-cache, then the cost to make them just about doubles for AMD (over the non-v-cache variants). So whatever cost premium the X3D carries over the non-X3D now, go ahead and double that, give or take.
I'm going to go out on a limb and say that the performance gains it would get from a second v-cache CCD over one would be nowhere near the increase it sees from one v-cache CCD over none. In many cases, I could actually see it being margin of error (so let's say no) difference. I don't think that's a stretch to assume since the Ryzen 9 CPUs often see very minimal if any performance gain in games over the Ryzen 7 despite their extra cache, and that is probably because it is split across two CCDs just as it would be here.
So if it had an even further disproportionately higher price premium relative to the performance gain, how many people would truly be interested in it? With most people already choosing the Ryzen 7 x800X3D over the Ryzen 9 X3Ds, it's clear AMD would see it as a poor business choice to make the value of attractiveness of that lesser chosen model even worse.
Maybe I'm wrong about all that (I very well could be as this is just how I see things), and I'm not saying we'll never see it happen; maybe AMD thinks differently either now or in the future and decides to test the water. But right now, I don't see it as an attractive change to make.
Instead of these having v-cache on both CCDs, I think increasing the individual CCDs from 8 cores to either 12 or 16 (and then the v-cache amount per local CCD increasing as well) would likely be more meaningful and reasonable changes to (eventually) expect.
Plus having two CCD's with cache won't solve the latency penalty anyway for crossing CCD's. So in the situations that do, it'll still be slower than it should be. In situations that don't (which is the ideal situation) it'll be just as fast as having cache on one CCD and the second CCD's extra cache will be unused.
So why put it on two CCD's for measly gains in suboptimal conditions while jacking up the price? To all the proponents wanting cache on two CCD's, I really don't get why they want it in the first place.
Given the likely application domain, my question is whether there are other computationally intense activities that benefit from dual v-cache. I think the answer to that question is yes.
My hope is that when stacking the CPU on top it's easier to make both stacks the same height by putting v-cache under each. If yields are better with both CPUs sitting on a v-cache that miracle would lead to a product that differentiates itself enough to merit an upgrade from the 7950X3D. Otherwise, maybe not.
And if one is, they realize the small difference between the x800X3D and the x950X3D is more than offset by having the much higher multi-threaded performance.
In other words, I don't imagine this is a big deal to most of the actual target audience of those CPUs. It only seems to be a big deal to those who are only gamers wanting the Ryzen 9 when they don't even need it.
You can say that it won't chase off potential buyers but I disagree with that, personally. This isn't comparable to GPUs. An RTX 4090/5090 are, simply put, always faster and by quite a lot than an RTX 4080/5080 respectively. So to those people, it's worth paying disproportionately for those faster ones.
CPUs are different to GPUs in that regard because Ryzen 9s are generally meaningfully faster than Ryzen 7s only for cases where the extra cores are meaningfully used, and not "by default in everything" as is the case with GPUs. That's the big difference. For CPUs, the time where that's the case is in highly threaded things, such as productivity, and not really in games. How many highly threaded productivity users would be willing to spend some hundreds more for probably single digit (if any!) performance increases? Once you move the price closer to Threadripper, it starts losing more appeal for those low end workstation users in my mind.
Adding v-cache to both CCDs just seems to be beneficial to too niche of a case (at least for now) that would lose it appeal for its actual target audience just to appease... gamers who don't even need it, and can already get that out of a x800X3D for less money.
It's also possible that AMD might experiment with adding X3D to their ZenC-cores and see if that provides uplift to warrant increased use of ZenC-type cores for some SKUs, providing more cores without having to actually squeeze in more cores per standard Zen chiplet.
The CPU speeds up everything, even small increases are important.Threadripper gaming performance is poor. You don't seem to be thinking about the people who want a CPU that can do everything well. The vast majority of people aren't all productivity or gaming, most do both. At the end of the day it's about pluses and cons and for the top bracket of the market a higher cost isn't much of a con.
But I guess even if scheduling was 100%, fixed there will still be people claiming "if only AMD had cache on both CCD's it would be even faster!"