Alleged AMD Desktop CPU Roadmap Leaks Out: Ryzen 7000 3D V-Cache In Early 2023

Not sure where it says early 2023 for X3D? It clearly says "TBA" and it has a 2023 slot - but there's not specific definition.

Anyway, it is hilarious to me how everyone now is pro-5800X3D - when it launched everyone poo poo'd the core count and the productivity numbers and the dead end platform. Now - we've got the 5800X3D (gaming performance-wise) PLUS 16 cores, DDR5, PCIE5, and more in the 7950X and everyone is like meh. lol...
 
Not sure where it says early 2023 for X3D? It clearly says "TBA" and it has a 2023 slot - but there's not specific definition.

Anyway, it is hilarious to me how everyone now is pro-5800X3D - when it launched everyone poo poo'd the core count and the productivity numbers and the dead end platform. Now - we've got the 5800X3D (gaming performance-wise) PLUS 16 cores, DDR5, PCIE5, and more in the 7950X and everyone is like meh. lol...
It all depends on two things:
1. Who's reading/posting at the time
2. What just came out, because the "other" thing is better every time :p
 
https://wccftech.com/amd-ryzen-7000...00-ryzen-7000g-apus-desktop-cpu-roadmap-leak/

"The next major update to the family will come in the form of the X3D or 3D V-Cache parts which are mentioned but it looks like these chips might slip into early 2023"

The article mentions Threadripper 7000 model numbers appearing in some database somewhere and 7000G APUs coming next year.

It's WCCFTech, so you know, take it as seriously (or not) as you wish.
Zen4 X3D in Early 2023 has been leaked since a couple of days before Zen 4 announcement/presentation.
 
Not sure where it says early 2023 for X3D? It clearly says "TBA" and it has a 2023 slot - but there's not specific definition.

Anyway, it is hilarious to me how everyone now is pro-5800X3D - when it launched everyone poo poo'd the core count and the productivity numbers and the dead end platform. Now - we've got the 5800X3D (gaming performance-wise) PLUS 16 cores, DDR5, PCIE5, and more in the 7950X and everyone is like meh. lol...
I bought the 5800x3d several months back and it is a beast in gaming. Anyway the main reason is to let the first new generation and all the issues, extra costs etc. pass. The Zen 4 CPU is fine but everything else from motherboard prices, ram and features such as pcie 5 at this stage is not that beneficial is not worth it to me.

Zen 4 x3d might be interesting, Threadripper will be interesting depending upon total costs. To me it looks like Raptor lake will outperform Zen 4 anyways. Meteor Lake maybe my buy in stage. Zen 5?
 
I bought my 7900X last week fully intending to sell it in a few months for a 7950X3D. Bring them on
 
I was planning a build this fall, but now I'm just going to hold out for these. I figure we should (hopefully) have PCIE 5 drives and PSUs on shelves by then, too. Plus the B650/E mobos seem right up my alley in terms of features and cost. DDR5 prices should keep falling, too. I just have to cross my fingers and hope my wife's machine doesn't die between now and whenever these become available.
 
I bought my 7900X last week fully intending to sell it in a few months for a 7950X3D. Bring them on

are they gonna do a full lineup/more than a 7800X3D?

Edit: oh yes supposedly, we'll see
 
I agree that the timing and lack of clarity on TR 7000 is annoying, but it being the high-end gaming solution has been dead since TR1 - it very quickly moved to the prosumer marketplace and folks with oddball workloads like I have (hence why I have both x399 and TRX4). It's all about content creation, mass storage, almost-server, etc - while being a highly competent normal desktop machine - and those first focus points are far more important than the latter, as the consumer and gaming side will always move faster (and has different needs, that are steadily (sadly) diverging from the consumer space). Do I really hope that there's a prosumer-level Threadripper? Oh you're damned right I do. I LOVE mine to death - best damned system I've ever owned. But... I also recognize that it's not likely. The folks doing true professional studio/rendering work have different needs than those of us that straddle the two worlds, and I think our market is dying.

They come out with a sub-2k TR with 16-24 cores and I'll buy it - otherwise, I'll be riding this system down till someone has a replacement, or I'll have to get creative.

I guess the question then is why they're bothering to keep it around at all , if its so heavily leaning towards the "pro" side of "prosumer" both in exclusivity of function and price? Why not simply make "cheap" EPYCs or whatnot? I'm a big fan of HEDT going back to my Intel X58 and then X99 platforms, and I'd be happy if AMD brought back a Threadripper w/ X3D proc series that could clock single/few cores to the same level or beyond Ryzen due to binning + multicore potential + more PCI-E lanes / quad-channel RAM, other high end I/O and chipset features. Hell, I think there would be a market for people to buy as you said even a an 8 or 16 core chip but with all the other benefits of HEDT, provided it could handle the single/few core performance and OC features like my old X99 / Haswell-E 8c/16t that keeps trucking along. It just seems like they aren't planning to do that and instead will continue to focus heavily on the massive multithread workload side, I'm not sure why its worth it to have a whole separate platform vs just making "lower end" EPYC .

Though at this point if they are having technical issues getting 7000 series X3D fab'd at the frequency they desire (A few with knowledge of the fab process in one of these threads mentioned this was the case, I cannot independently confirm it offhand), then I hope they can focus on that ASAP because unless Raptor Lake is just a horrid dud across the board then AMD will likely need X3D to recapture the performance crown for some use cases that are NOT heavily multitthreaded and benefit significantly in those games and other tasks that respond to increased cache. I hope they can sort it out and bring especially the 7950X3D and 7800X3D to market as soon as quality allows them; ideally this year, but if it comes to next year and really thrives that will be better than a fast release with massive limitations to frequency or other settings, performance hold-ups and the like.
 
I guess the question then is why they're bothering to keep it around at all , if its so heavily leaning towards the "pro" side of "prosumer" both in exclusivity of function and price? Why not simply make "cheap" EPYCs or whatnot?
That's literally what it is now. Remove the connections/disable multi-socket, boost core speeds a bit since you've got quiet parts of the chip not making heat now, and sell it for a bit cheaper. Xeon-W is just a Xeon chip with the multi-socket turned off and memory limits. TR Pro is the same - but works in a "simpler" motherboard with more consumer features, since server boards suck for desktop or workstation.
I'm a big fan of HEDT going back to my Intel X58 and then X99 platforms, and I'd be happy if AMD brought back a Threadripper w/ X3D proc series that could clock single/few cores to the same level or beyond Ryzen due to binning + multicore potential + more PCI-E lanes / quad-channel RAM, other high end I/O and chipset features. Hell, I think there would be a market for people to buy as you said even a an 8 or 16 core chip but with all the other benefits of HEDT, provided it could handle the single/few core performance and OC features like my old X99 / Haswell-E 8c/16t that keeps trucking along. It just seems like they aren't planning to do that and instead will continue to focus heavily on the massive multithread workload side, I'm not sure why its worth it to have a whole separate platform vs just making "lower end" EPYC .
I'd buy it too - for sure - but see above about the Epyc part. It's the same cores.
Though at this point if they are having technical issues getting 7000 series X3D fab'd at the frequency they desire (A few with knowledge of the fab process in one of these threads mentioned this was the case, I cannot independently confirm it offhand), then I hope they can focus on that ASAP because unless Raptor Lake is just a horrid dud across the board then AMD will likely need X3D to recapture the performance crown for some use cases that are NOT heavily multitthreaded and benefit significantly in those games and other tasks that respond to increased cache.
This is one of the many reasons I build a separate gaming box now - the venn diagram of gaming needs and workstation needs is not a circle.
 
I guess the question then is why they're bothering to keep it around at all , if its so heavily leaning towards the "pro" side of "prosumer" both in exclusivity of function and price? Why not simply make "cheap" EPYCs or whatnot? I'm a big fan of HEDT going back to my Intel X58 and then X99 platforms, and I'd be happy if AMD brought back a Threadripper w/ X3D proc series that could clock single/few cores to the same level or beyond Ryzen due to binning + multicore potential + more PCI-E lanes / quad-channel RAM, other high end I/O and chipset features. Hell, I think there would be a market for people to buy as you said even a an 8 or 16 core chip but with all the other benefits of HEDT, provided it could handle the single/few core performance and OC features like my old X99 / Haswell-E 8c/16t that keeps trucking along. It just seems like they aren't planning to do that and instead will continue to focus heavily on the massive multithread workload side, I'm not sure why its worth it to have a whole separate platform vs just making "lower end" EPYC .

Though at this point if they are having technical issues getting 7000 series X3D fab'd at the frequency they desire (A few with knowledge of the fab process in one of these threads mentioned this was the case, I cannot independently confirm it offhand), then I hope they can focus on that ASAP because unless Raptor Lake is just a horrid dud across the board then AMD will likely need X3D to recapture the performance crown for some use cases that are NOT heavily multitthreaded and benefit significantly in those games and other tasks that respond to increased cache. I hope they can sort it out and bring especially the 7950X3D and 7800X3D to market as soon as quality allows them; ideally this year, but if it comes to next year and really thrives that will be better than a fast release with massive limitations to frequency or other settings, performance hold-ups and the like.
The reason you don't see any cheap Threadrippers and won't for the foreseeable future is that they sell 95% of the ones they currently make at that high price, AMD does not have the Silicon resources to create a new lineup, and selling them cheaper than they are now with them flying off the shelves would just get AMD an investor lawsuit and lower revenue for their troubles. Prosumer HEDT is just not a profitable market and both AMD and Intel are more than happy to let it die.
.
Regarding the X3D parts and their TBA release date which is less technical and more mechanical, TMSC is having a hard time with the accuracy of their stacking equipment, they need it to be accurate to under 200nm to physically line the chips up for their 5nm process. However in practice, the variance on it is causing abysmal failure rates, and it seems only consistent around 500nm instead. TSMC is working with its partners to resolve this but it is delaying AMD launch schedules. Fortunately, TSMC has stated they have a new formula for the bonding agent they use between the layers and it has lowered electrical resistance a significant amount, that lowered resistance means it generates less heat and lets them increase the voltage accordingly.
In the 5000 series, the reason they had to lock the voltage down so strictly is that the resistance of their earlier formula is high enough that the heat it generates is enough to melt the bonding agent and cause the chips to delaminate. The thermal limit of the agent is set in stone because any higher than it currently is would damage the chips during the application process leaving it a very delicate process. The engineering behind the whole process is actually quite extraordinary, I still have a hard time wrapping my head around it.

And I really don't think that Raptor Lake will be a horrid dud, but even if it does edge out the AMD lineup at this stage the AMD platform has more than enough strengths to make it a viable contender, they just need to adjust their pricing a little but given they have a margin upwards of 40% in their Ryzen lineup they have lots of wiggle room there.
 
There was that 34 core raptor lake making the rumor round last week
I read that and it screams Xeon-W part to me, the lack of the E cores makes me think it's not intended as a normal desktop part and is intended for servers or workstations where you really don't want to be dealing with the big.little core designs. The voltage required to run that many P cores.... unless they tone the clock speeds down significantly doesn't seem like something that would fit inside their desktop power envelopes.
I mean if I am wrong and it is a consumer part that F-YEAH! That's the kind of monster chip we want to play with, it just seems like a big departure from their desktop lineup and strikes me as odd.
 
I read that and it screams Xeon-W part to me, the lack of the E cores makes me think it's not intended as a normal desktop part and is intended for servers or workstations where you really don't want to be dealing with the big.little core designs. The voltage required to run that many P cores.... unless they tone the clock speeds down significantly doesn't seem like something that would fit inside their desktop power envelopes.
I mean if I am wrong and it is a consumer part that F-YEAH! That's the kind of monster chip we want to play with, it just seems like a big departure from their desktop lineup and strikes me as odd.
That’s sapphire rapids for sure, not raptor lake. And it’s branded Xeon-w or so all the rumors say- but they also say there will be a HEDT line and a prosumer workstation line of those. Fingers are crossed!
 
Back
Top