point being there's more comp than just Intel in the market now.N1X isn’t a threat IMO, it comes with stock ARM cores from 2024.
It has the green badge and actually useful GPU IP, plus Mediatek is doing the SoC engineering as NV is not good at that.N1X isn’t a threat IMO, it comes with stock ARM cores from 2024.
Medusa (mobile) is CES'27, desktop and server is H2'26Glymur at the least CPU side is going to very good but it won’t come out till H12026. Medusa will likely launch mid by 2026.
The interwebs say PCIe 5.0 x16 gives us 63 GB/s (minus overhead; twice that if TX and RX are summed up).before CXL can be bottlenecked by the PCIe5 x16 BW of 256 GB/s, the system RAM bottleneck is already hit 128GB/s on 8000MT/s DDR5.
Need wider buses
Why not? Intel currently holds the #1 position in x86. Seems like an AWFUL lofty claim. Please explain.They do not care about Intel.
Venice D is just a bunch of CCD's using Zen 6c cores. What makes you think that it wouldn't make sense for AMD to put one of them on the desktop along side a full Zen 6 (12c) and the new IOD? I figure AMD will NEED this to compete with Nova Lake's 54 core variant since there is no possible way a 24c Zen 6 is going to compete in MT against a 54 core Nova Lake. Do the math.Venice-D CCD is, for, well, Venice-D.
They're not doing dense spam on DT.
Why not? What practical reason would AMD have NOT to do this?Dense CCDs will never be in client, only cheap mono parts.
Ironically the Venice-D CCD is the priciest CCD ever.
Except it is not likely that a Threadripper system can compete in price with a Nova Lake system. Not even close. How would the economics of that work?Enough big boy cores is enough, want more than 48 threads, you need more memory channels.
Just buy Threadripper.
It is interesting to me how many people feel that the design OEM's are simply lazy and/or stupid. Ever meet any of these guys? The reason innovation is slowing down is because foundry technology advances are slowing down. MOST of the big performance improvements have come by taking advantage of vastly larger transistor budgets gen over gen. That is no longer the case.Nothing like a loss of market share or revenue to move these companies. If QC/NV can manage to win share in PC there will be some innovation.
Because.Why not?
well the world extends beyond the walls of x86.Intel currently holds the #1 position in x86. Seems like an AWFUL lofty claim. Please explain.
no it's a product with a designated CCD (well, stack, really).Venice D is just a bunch of CCD's using Zen 6c cores
Because client doesn't need any of that.What makes you think that it wouldn't make sense for AMD to put one of them on the desktop along side a full Zen 6 (12c) and the new IOD?
No they don't.I figure AMD will NEED this to compete with Nova Lake's 54 core variant since there is no possible way a 24c Zen 6 is going to compete in MT against a 54 core Nova Lake
The world is more than just cinememe.there is no possible way a 24c Zen 6 is going to compete in MT against a 54 core Nova Lake
who cares.I guess AMD could simply concede the MT performance crown that they have been touting for years now to Intel and become the single thread king? Kinda a wierd role reversal don't you think?
Completely divergent roadmaps.Besides, I think it way more likely that AMD will continue to improve the IOD capabilities through trickle down from server learnings. I expect more of this cross-pollenation, not less.
Well it's just a 2t 8+16 config.They need to focus on this and not creating 52 core CPUs that will most likely get bottlenecked by mem bandwidth.
Yes, the older Genoa (by that time) will be affordable, with up to 96 cores and 12 channel memory and avx-512. Turin may or not be affordable, but it has full width avx-512, and a year or two from now, who knows.Here is the thing most DIY sales are for gaming. So in that in mind, the "52" core NVL part has to outperform the 10800X3D and the 10950X3D in gaming.
Its also likely going to very expensive. So 99.99% of gamers will go for Zen6 X3D. What the NVL part will excel at is Cinebench but excelling in those don't move units otherwise Arrow lake would be selling like crazy.
----
The core of client business is laptops and here Intel just doesn't have a good product in 2026. What is Intel's response to AMD/QC/NV in the laptop space in late 2026? Its currenly a 8+4+4 NVL-H part.
They need to focus on this and not creating 52 core CPUs that will most likely get bottlenecked by mem bandwidth.
Cuz it's better.
yeah it's better better.Hah, no. They picked it for another very BIG reason.
Venice CCD being very expensive, presumable volume and margin of such a hybrid would be too low (~5000 buyers at $2K price would still be only 10 mil rev, probably not even enough to cover the engineering cost).Why not? What practical reason would AMD have NOT to do this?
Remember, AMD needed *something* to show in the slides - and in case of Zen 1, Zen 1+, and Zen 2 it really couldn't be their single-thread performance. So in case of Zen 1/1+ they were fine with the server-derived 8c choice. With Zen 2 they needed something more, hence the easy-enough 2 CCD config.I figure AMD will NEED this to compete with Nova Lake's 54 core variant since there is no possible way a 24c Zen 6 is going to compete in MT against a 54 core Nova Lake. Do the math.
I guess AMD could simply concede the MT performance crown that they have been touting for years now to Intel and become the single thread king? Kinda a wierd role reversal don't you think?
Dense version trades max clocks for density, so having 4-ish ghz peak is hardly going to be attractive on desktopVenice D is just a bunch of CCD's using Zen 6c cores. What makes you think that it wouldn't make sense for AMD to put one of them on the desktop
Who cares if you get 32 cores at 4 ghz instead of 12 at 5.x ghz? Its roughly double the wrought Mt performance. Your primacy CCD should still be a 12c for peak st and lightly threaded loads.Dense version trades max clocks for density, so having 4-ish ghz peak is hardly going to be attractive on desktop
The late Gene Myron Amdahl would have cared.Who cares if you get 32 cores at 4 ghz instead of 12 at 5.x ghz?
I worded it poorly. Who cares if the 32 cores run at only 4ghz?The late Gene Myron Amdahl would have cared.
Well games for starters will not like that for sure, better question is who needs 32 cores at desktop on new process and prepared to pay $3-4k for them?Who cares if you get 32 cores at 4 ghz instead of 12 at 5.x ghz?
Links to some data please. I don't think this is correct.Here is the thing most DIY sales are for gaming.
No, they wont. This would be about maintaining dominance over Intel's Nova Lake offerings.tl;dr: They won't waste money just to make a small group of desktop enthusiasts' wet dreams come true.
See below (as this is nearly word for word how I was about to reply).Dense version trades max clocks for density, so having 4-ish ghz peak is hardly going to be attractive on desktop
I saw another rumor yesterday that suggested a dual CCD with 12 full Zen 6 and another with 16 Zen 6c. I don't see the point in this configuration since the 16 core CCD would be unique to that usage while the 32 core CCD would be shared by EPYC D Zen 6.... and it would also bring the performance of AMD's Zen 6 offering in MT up to snuff with the rumored Nova Lake.Who cares if you get 32 cores at 4 ghz instead of 12 at 5.x ghz? Its roughly double the wrought Mt performance. Your primacy CCD should still be a 12c for peak st and lightly threaded loads.
My invoice is in the post...See below (as this is nearly word for word how I was about to reply).
This is the funniest joke I've seen in a while.1) Gaming (likely <10% of desktop users)
2) Workstations for editing video/graphics (my guess is a big percentage of the remaining 90%)
Please back up your humor with some data.This is the funniest joke I've seen in a while.
There are like a zillion data points for stores that sell DIY CPUs and 9800X3D is like >80% of the volume.Please back up your humor with some data.
Gaming and javascript perf is literally DT's bread and butter.1) Gaming (likely <10% of desktop users)
2) Workstations for editing video/graphics (my guess is a big percentage of the remaining 90%)
Seems to me like MT is way more important than ST for desktop. Someone steer me straight if I am off base here.