• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

Page 821 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Ah. So this time they are gonna make the expensive X3Ds seem more attractive with exclusive features!

Could it be that the 9900X3D manages to be the first ever desirable x900X3D SKU???
 
Not the exact way I described. Epyc chips have an array of fuses that correspond to the OEM's/System Integrators and it will blow the fuse for them on first power on in their motherboard. This makes it so other brands of motherboard will refuse to post with the CPU.
But this is opt-in. It wont be used if system integrator/oem won't enable it. Usually it is also a vendor specific lock and it also applies to some ryzen pro systems too.
 
Strix Halo should have had X3D as part of its design intent. But AMD is not a consumer focused company, so expect perpetual disappointment (its not just them, careful what ya wish for, we'll be looking at ARM and low power/low core count/etc chips). But what if...

I think there is a chance that it happens. My guess is, the Strix Halo CCD will have ability to add V-Cache, but AMD will at first wait to see how much interest there is before committing to making it a product.

Just a speculation on my part.
 
It's still e-waste generation. AMD shouldn't enable OEMs like that but I understand they don't want to lose a deal to Intel because of some fuses.
I agree it would be best if they did away with this feature, just wanted to underline not every EPYC cpu has to be affected by this.
 
The only thing that could possibly make the dual CCD X3D SKUs desirable for gaming over the single CCD would be if they managed to unify both CCDs into a single 16 core CCX that shared the X3D die. This wont happen though, because it would be potent and sell like hotcakes. :tearsofjoy::tearsofjoy:
 
The only thing that could possibly make the dual CCD X3D SKUs desirable for gaming over the single CCD would be if they managed to unify both CCDs into a single 16 core CCX that shared the X3D die. This wont happen though, because it would be potent and sell like hotcakes. :tearsofjoy::tearsofjoy:
It would only cost 1000$ dollars. :fearscream:
 
I thought Strix Halo was going to have extra cache in the form of MALL or an L4 or something (shared by CPU and GPU) Wouldn’t 3D cache be a little redundant in that case?

Is there any solid info one way or the other if it will feature such a cache?
 
Has AMD fixed Zen 5 yet? By this I mean are they able to hit their performance targets prior to release of Zen 5? Will the x3D chips solve AMD's Zen 5 problems?
 
I thought Strix Halo was going to have extra cache in the form of MALL or an L4 or something (shared by CPU and GPU) Wouldn’t 3D cache be a little redundant in that case?

Is there any solid info one way or the other if it will feature such a cache?
Halo is going to have something to aid in the video memory bandwidth available to its GPU, as DDR-5 aint going to cut it for 40 CUs. Whether thats a large chunk of GPU L2 or L3 cache or on chip VRAM, nobody knows outside of AMD and TSM at this point.
 
Strix Halo should have had X3D as part of its design intent. But AMD is not a consumer focused company, so expect perpetual disappointment (its not just them, careful what ya wish for, we'll be looking at ARM and low power/low core count/etc chips). But what if...
The performance uplift is likely far lower than you think.
X3D yields so well on desktop parts due to low membw and IFOP bw.
Halo has far more membw, USR means the CCD is no longer bottlenecking and MALL to add another cache layer.
So it is likely not worth the cost.
 
MALL cache is there for a couple of reasons. Not only does it increase the effective bandwidth of main memory from the 256 bit 200+ GB/sec to about double that, it ALSO decreases power draw due to memory accesses as every bit that it serves is notably less costly, energy wise, than a call to DRAM.
 
Back
Top