Friday, August 23rd 2024

AMD Ryzen 9000X3D Processors with 3D V-Cache Arrive in January at CES 2025

AMD's upcoming Ryzen 9000X3D series of processors with 3D V-Cache have been reportedly scheduled to arrive in January 2025 and should make a debut at the CES event, a few months later than initially expected. While disappointing for eager enthusiasts, the delay could signify that AMD is taking extra precautions to ensure a smooth launch and deliver a product that meets the high-performance standards set by its previous 3D V-Cache offerings. Delaying the new product launch could also be a strategic move by AMD to avoid potential supply chain issues or to align the launch with other product announcements.

We previously reported that the series will maintain the same cache size configurations as the last generation with 3D V-Cache, and it will just be an upgrade to up the performance of the new Zen 5 design. The launch of the 9000X3D series follows a smaller six-month gap from the regular Ryzen 9000 series, where the previous Ryzen 7000X3D and 5000X3D followed seven and seventeen months after the launch of their regular SKUs, respectively.
Sources: HXL, via VideoCardz
Add your own comment

84 Comments on AMD Ryzen 9000X3D Processors with 3D V-Cache Arrive in January at CES 2025

#1
Outback Bronze
"AMD is taking extra precautions to ensure a smooth launch and deliver a product that meets the high-performance standards set by its previous 3D V-Cache offerings"

And if people can't be f'd waiting they will buy 7 series or 9 series. Win Win.
Posted on Reply
#2
JWNoctis
Figure looked like it was from the original 3D V-Cache whitepaper or something?

I think they are supposed to have a new V-Cache die this generation? Are there any further details yet?
Posted on Reply
#3
Nhonho
Too late Marlene...

AMD should rush the launch of the Ryzen 11000 generation, taking advantage of the crisis of Intel's current CPUs, to increase its sales.
Posted on Reply
#4
Legacy-ZA
I have said this from the 5000 series, just release the cached versions, nobody wants to have buyers remorse, we get that enough from nVidia already.
Posted on Reply
#5
Nhonho
Legacy-ZAI have said this from the 5000 series, just release the cached versions, nobody wants to have buyers remorse, we get that enough from nVidia already.
The production failures of these CPUs with a lot of cache must be much higher.. and, of course, they are more expensive to produce and only offer more performance in games and at 1080p resolution. At higher resolutions (1440p, 4K), what matters most is the GPU performance.

What AMD needs to do is place the memory controller on the same die of the x86 cores to minimize RAM access latencies.
Posted on Reply
#6
Bwaze
No doubt trying to somehow fix the error of 9000 launch, with promises of double figure performance uplift, but releasing something that's basically on the same level as previous generation.

But I doubt playing with Windows scheduler and power states will help - unless they somehow nerf Ryzen 7000 CPUs. And ignore Intel altogether, just declaring it undesirable due to other factors than performance.
Posted on Reply
#7
kondamin
NhonhoWhat AMD needs to do is place the memory controller on the same die of the x86 cores to minimize RAM access latencies.
Might as well use a 2 socket system if they were to start doing that
Posted on Reply
#8
JWNoctis
NhonhoThe production failures of these CPUs with a lot of cache must be much higher.. and, of course, they are more expensive to produce and only offer more performance in games and at 1080p resolution. At higher resolutions (1440p, 4K), what matters most is the GPU performance.

What AMD needs to do is place the memory controller on the same die of the x86 cores to minimize RAM access latencies.
And lose the one-chiplet-to-rule-them-all commonality and flexibility? AMD's approach put the L3 much closer to the cores than the competitors do, which resulted in good latency in L3, albeit I'm not sure if it is still better than Intel 13th/14th gens.

RAM access, on the other hand, is IF-bottlenecked which does cause problems. They may well end up doing something about it with Zen 6.
Posted on Reply
#9
mikesg
NhonhoThe production failures of these CPUs with a lot of cache must be much higher.. and, of course, they are more expensive to produce and only offer more performance in games and at 1080p resolution. At higher resolutions (1440p, 4K), what matters most is the GPU performance.
[...]
MS Flight Simulator really benefits from the extra cache, and a few other games at any resolution.
Posted on Reply
#10
R0H1T
NhonhoWhat AMD needs to do is place the memory controller on the same die of the x86 cores to minimize RAM access latencies.
That completely defeats the purpose of Zen, why would they do that just for "gamers" when the OEM market probably dwarfs them anyway?
Posted on Reply
#11
kondamin
R0H1TThat completely defeats the purpose of Zen, why would they do that just for "gamers" when the OEM market probably dwarfs them anyway?
Disable the igpu from one of their apu line of chips and gamers have their product
Posted on Reply
#12
R0H1T
The APU's have less cache, regular desktop chips have 2x & x3d around 4x IIRC, hence a gamer's delight!

Posted on Reply
#13
Assimilator
Here's hoping the delay is to fix their core scheduling mess.
AleksandarKDelaying the new product launch could also be a strategic move by AMD to avoid potential supply chain issues or to align the launch with other product announcements.
Probably RDNA4 launch.
Posted on Reply
#14
Bwaze
But the longer the delay, the more expectations will rise, because you kind of expect more progress with more time.
Posted on Reply
#15
Ruru
S.T.A.R.S.
Legacy-ZAI have said this from the 5000 series, just release the cached versions, nobody wants to have buyers remorse, we get that enough from nVidia already.
Personally I'm totally fine with my 5800X :D
Posted on Reply
#16
Totally
NhonhoToo late Marlene...

AMD should rush the launch of the Ryzen 11000 generation, taking advantage of the crisis of Intel's current CPUs, to increase its sales.
I'd rather they not only then to turn into Intel with bugs, exploits, and exploits in every CPU that they'd launch between then and now that could have been avoided.
Posted on Reply
#17
kondamin
R0H1TThe APU's have less cache, regular desktop chips have 2x & x3d around 4x IIRC, hence a gamer's delight!
there are just 3 ryzen AI 300 chips so far, they could make one with 6 zen 5 cores and a cache chip on top.
It probably would sell quite well
Posted on Reply
#18
Legacy-ZA
NhonhoThe production failures of these CPUs with a lot of cache must be much higher.. and, of course, they are more expensive to produce and only offer more performance in games and at 1080p resolution. At higher resolutions (1440p, 4K), what matters most is the GPU performance.

What AMD needs to do is place the memory controller on the same die of the x86 cores to minimize RAM access latencies.
Oh, I see, so what you are saying, we are paying too much for trash if we don't get the cached version, yes, yes, I know.
Posted on Reply
#19
Wirko
JWNoctisRAM access, on the other hand, is IF-bottlenecked which does cause problems. They may well end up doing something about it with Zen 6.
Indeed, there seems to be a lot of room for improvement here, and AMD should have updated the IOD for Zen 5 already, even if it remained on the same node. Chiplet-to-chiplet latency is far too high as well. Many times higher than memory CAS latency, for example!
Posted on Reply
#20
robert3892
From a marketing standpoint, I think this is a poor decision. It would have been better to have at least one model available to purchase during the holiday shopping season. That would also have benefitted system integrators/builders.
Posted on Reply
#21
Assimilator
robert3892From a marketing standpoint, I think this is a poor decision. It would have been better to have at least one model available to purchase during the holiday shopping season. That would also have benefitted system integrators/builders.
From a marketing standpoint it makes sense to launch Zen 5 vcache alongside RDNA4 so they can do one single, big bang, "gaming" event.

From a sales standpoint AMD is likely hoping to use holiday sales to draw down Zen 4 stock so that they can go all-in on Zen 5.
Posted on Reply
#22
Makaveli
JWNoctisAnd lose the one-chiplet-to-rule-them-all commonality and flexibility? AMD's approach put the L3 much closer to the cores than the competitors do, which resulted in good latency in L3, albeit I'm not sure if it is still better than Intel 13th/14th gens.

RAM access, on the other hand, is IF-bottlenecked which does cause problems. They may well end up doing something about it with Zen 6.
This.

I also think the Infinity Fabric bottleneck is the biggest issue in Zen currently which they will be fixing with Zen 6.
R0H1TThat completely defeats the purpose of Zen, why would they do that just for "gamers" when the OEM market probably dwarfs them anyway?
Agreed but gamers have one track minds and think people don't to anything else on a PC.
Posted on Reply
#23
Nhonho
kondaminMight as well use a 2 socket system if they were to start doing that
It would be enough to make the CPUs this way in a single socket:

Posted on Reply
#24
AusWolf
It's a delight to see so many people knowing things way better than any AMD engineer. I just wonder why they all aren't employed by AMD already. :wtf: /s
Posted on Reply
#25
Assimilator
NhonhoIt would be enough to make the CPUs this way in a single socket:

Except then AMD has a duplicate memory controller on every CCX. And the socket needs to have a whole bunch of unused pins for the <= 8-core parts. And motherboards need to support either dual- or quad-channel memory configurations for the > 8-core parts. And AMD doesn't want to bring quad-channel to mainstream. And so many other reasons.

Far easier to build a half decent memory controller to begin with (well you'd think, but AMD has consistently managed to not do this).
Posted on Reply
Add your own comment
Nov 21st, 2024 08:58 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts