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

Page 888 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

yottabit

Golden Member
Jun 5, 2008
1,527
592
146
The dual CCD parts are mostly productivity chips and given the (surprisingly honestly) gains the 9800X3D has in some productivity workloads, I think lends some credence to the idea they may be dual VCache this time around. At least one can hope!
 
Reactions: biostud

Joe NYC

Platinum Member
Jun 26, 2021
2,672
3,839
106

AMD Marketing Brain(Dead) Thrust in full force, again.

They are still pushing brain dead 12 core SKU and pair it with the highest end iGPU configuration. While at the same time, they pair the most desirable 8 core SKU with lower end, partially disabled iGPU.

Also, AMD does not fail to disappoint by lowering max memory speed from 8533 to 8000.

And also, not using Memory on Package, which would allow the fastest speeds on the inside of the CPU package, without any changes needed on the laptop mini-mobo is not used. So AMD already tied its hands, preventing future improvements.
 

Joe NYC

Platinum Member
Jun 26, 2021
2,672
3,839
106
You raise an interesting point; Will all the Strix Halo SKUs have the same 256 bit memory bus?

For the Ryzen Ai Max Pro 380's 16 CU GPU, 128 bit LPDDR5X-8000 and 32 MB SLC would be sufficient to feed it.

But still use a massive and super expensive N3e node?

I have to wonder how many SoC rejects AMD will have, so bad that they need to disable 60% of the GPU to make this CPU necessary. There is no way to make money on such a costly SoC die, unless this is the absolute garbage bin.
 
Last edited:

LightningZ71

Golden Member
Mar 10, 2017
1,910
2,260
136
It looks like they are attempting to recover every possible poor bin combination. I doubt that the 6 core/16 cu part has either a full 256 but memory Implementation or a full 32 MB MALL cache. It's likely an IOD recovery part for ones with a bad memory controller or SRAM cell. They are likely throwing CCDs at it that have at least one failed core or can't speed/power bin for the 12 core package.

It's a cost of production reduction part and will likely be an OEM/lower cost market model that will help total OEM platform volume numbers.
 

Joe NYC

Platinum Member
Jun 26, 2021
2,672
3,839
106
That's the fastest (currently) qualified jedec speed for SO-DIMMs. As for the earlier comment, not everyone will need full implementation AVX-512. The one on Strix is no worse than Zen4, which was praised for it's performance as well. In addition, AMD has stated that the same AGESA changes that improved the cross CCX latencies for the desktop chips also help Strix.

In the end, we already see that Strix in laptops is plenty good enough for most games, with or even (with appropriate settings) without a DGPU. AMD's APUs have had better IMCs than their regular desktop parts and there's no reason to believe that Strix is an exception.

It would make a great ITX chip for compact chassis with good RAM. For an inexpensive 1080p monitor or running a TV, it would be just fine.

I just don't think it will be cheap. The alternative is Kraken, but they appear to have notably reduced it's PCIe capabilities to the point that it doesn't make sense in anything but low end laptops.

It really depends on what kind of PC are you trying to achieve.

If looking to a PC with dGPU, then none of the APUs make sense. It is better to just get a desktop CPU + dGPU.

If you decide on using iGPU, then what is the point of building full PC, when MiniPC can do the same job, in a smaller package, at lower price. And they don't use socketed CPUs, they use notebook package CPU.

Strix Point is not ideal configuration in any case. Actually, quite sub-optimal. 2 ring buses, one with 16 MB L3 and only 4 cores, and 2nd ring bus with 8 MB of L3 and 8x Zen 5c cores.

Kraken will apparently fix this and (if binned equally) will perform better than Strix Point (except in Cinebench). With smaller die, lower cost.

Then, on the high end, Strix Halo will offer superior performance all around.
 

StefanR5R

Elite Member
Dec 10, 2016
6,056
9,106
136
If an item in the rumored product lineup looks braindead to you, then it's not because AMD marketing is braindead.
Rather, either it's because the rumor is false, or it's because AMD marketing and one or more of their customers came to an agreement which looks braindead to you, but not to them.¹

________
¹) At least not to the customer, that is.
 
Last edited:

LightningZ71

Golden Member
Mar 10, 2017
1,910
2,260
136
It really depends on what kind of PC are you trying to achieve.

If looking to a PC with dGPU, then none of the APUs make sense. It is better to just get a desktop CPU + dGPU.

If you decide on using iGPU, then what is the point of building full PC, when MiniPC can do the same job, in a smaller package, at lower price. And they don't use socketed CPUs, they use notebook package CPU.

Strix Point is not ideal configuration in any case. Actually, quite sub-optimal. 2 ring buses, one with 16 MB L3 and only 4 cores, and 2nd ring bus with 8 MB of L3 and 8x Zen 5c cores.

Kraken will apparently fix this and (if binned equally) will perform better than Strix Point (except in Cinebench). With smaller die, lower cost.

Then, on the high end, Strix Halo will offer superior performance all around.
Not everyone wants a fully fixed configuration MiniPC that, for the good ones, is more expensive than a decent spec miniITX box. Having an x16 slot that can be used for anything is VERY appealing. Need special I/O? Need a bunch more SSDs? Want a 10Gbps nic? It's all possible. Not everyone needs the highest end GPU.
 
Reactions: MangoX

CouncilorIrissa

Senior member
Jul 28, 2023
575
2,256
96
AMD Marketing Brain(Dead) Thrust in full force, again.

They are still pushing brain dead 12 core SKU and pair it with the highest end iGPU configuration. While at the same time, they pair the most desirable 8 core SKU with lower end, partially disabled iGPU.
I couldn't disagree more.
The 12-core SKU seems like the most sensible one and addresses one of the most popular complaints about AMD APUs, that the only way to get the full iGPU config is to go for the highest core count option. This one retains the full iGPU config but has fewer cores. It slots nicely above HX 370 in the lineup in terms of CPU performance and gets the full iGPU.

Furthermore, the reason that 12-core SKUs have started being frowned upon since Zen 4 is AMD raising the PPT to 230W (Zen 4)/200W (zen 5). The Zen 2/3 12-core SKUs were pretty good (in fact, the 3950X arrived months later than 3900x). Strix Halo is likely to operate at much lower power envelopes where these 4 additional cores would be relatively power starved.
 
Last edited:
Reactions: r.p

ToTTenTranz

Member
Feb 4, 2021
182
312
106
I wonder if down the line we could have:

- Halo IOD with full iGPU and 1x Granite Ridge CCD Zen5 8C with VCache (extreme gaming edition?)

- Halo IOD with 1x Turin Dense CCD Zen5C 16C for highest power efficiency (perhaps with some Zen 5C cores disabled).



Ryzen AI Max Pro 380 is overlapping with the Ryzen AI HX 370 in terms of specifications.

Same 16 CU RDNA3.5 GPU, but only 6 CPU cores?

Max Pro 380 will have a massive gaming performance advantage due to 256bit LPDDR5X and 32MB Infinity Cache.
Strix Halo is choking pretty hard on a GPU bandwidth bottleneck.
 

poke01

Platinum Member
Mar 8, 2022
2,581
3,409
106
Strix Halo GPU series is 8000S but it’s actually RDNA3.5. The mobile variants will be 8000M but they will be RDNA4.

AMD needs to fire their pathetic marketing department unless they keep doing it to on purpose to mislead customers.

I hate that Nvidia does this well but AMD using it S/M suffix is even worse since it’s different architectures.

RDNA4 will be much much better than 3.5 and I hope Strix Halo buyers won’t be fooled into thinking they are buying an SoC with RDNA4. I don’t even know Strix Halo why is releasing with RDNA3.5 instead of 4 unless it was internally delayed.

AMD Marketing Brain(Dead) Thrust in full force, again.

They are still pushing brain dead 12 core SKU and pair it with the highest end iGPU configuration. While at the same time, they pair the most desirable 8 core SKU with lower end, partially disabled iGPU
 

jpiniero

Lifer
Oct 1, 2010
15,365
5,884
136
RDNA4 will be much much better than 3.5 and I hope Strix Halo buyers won’t be fooled into thinking they are buying an SoC with RDNA4. I don’t even know Strix Halo why is releasing with RDNA3.5 instead of 4 unless it was internally delayed.

Strix Halo is going to be mainly about AI AI AI... so I don't think it'll be too much of a problem.

As for the delay, probably used more N3E wafers on Turin Dense than originally anticipated.
 

eek2121

Diamond Member
Aug 2, 2005
3,135
4,479
136
IIRC that leak showed the same 5.2ghz Fmax boost for the 9950x3d, which may hint that both CCDs are v-cached, or that the 2nd CCD is trash bin
Given the headroom we are seeing on the 9800X3D, I suggest ignoring that “leak”. I would be extremely surprised if the 9950X3D has a peak clock that low.

Remember: these parts are no longer thermally limited. The only thing that matters is how fast they can go with a given TDP, which is 170W for the x950 chips.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |