Do you mean they will be low ? or high ? or something else ?
It's not? Neither -H nor -U is 4+4+4 for the full product.Someone on Intel thread posted that Panther Lake will be 4+4+4, which sounds like a laptop chip. Not something that would challenge Zen 5 desktop.
It's not? Neither -H nor -U is 4+4+4 for the full product.
EDIT: oh you meant 4P + 4LP-E + 4Xe
Yeah you wait until NVL-S.Maybe @adroc_thurston could be right and Intel is out of the competition on desktop until 2026
Zen 5 is a CPU core µArch, not a product in of itself.I fear for Zen 5 sales.
Closed captioned unavailable. Can you give a brief summary please ?Mlid tripling down on Zen 5 performance.
Consider the source. Probably just useless noise.Can you give a brief summary please ?
A certain someone already said that Cougar isn't a big improvement/optimization over Lion. So indeed PTL won't compete with Strix/Kraken in ST and MT.Even if only for laptops it seems underpowered to face Zen 5, even if Strix was to use Zen 4 cores it would still be challenging for Intel given that they seem to remove HT and also due to either the sheer amount of cores in Strix or the augmented IPC relatively to Zen 4 in Kraken.
That's called Medusa.Coming in volume at CES 26 will be a bad idea given it would have to fight against STX+1 and KRK+1.
Wait...You mean Z6 will come to mobile first in 26? Huh, interesting to see AMD accelerating the cadence on mobile.That's called Medusa.
But yeah Intel client position is kinda tough, they bet on expensive underwhelming parts.
I'm pretty sure the point of Strix Halo is to beta test their advanced packaging and LP cores so they can dip their toes in the future of mobile.Wait...You mean Z6 will come to mobile first in 26? Huh, interesting to see AMD accelerating the cadence on mobile.
RDNA 5?So Z6 Core IP + RDNA 5 GFX and XDNA 3 for STX/KRK + 1? That's quite scary (For the competition😁)
Oh, it does. I was just surprised that they'll go directly into Zen 6 in 2026. Was expecting a Zen 5 refresh. AMD is being very aggressive with their timelines (As they should be).Makes sense that they shoot for mobile first after that when Z6 comes out.
RDNA 4 will be a dGPU only generation*. So STX+1 at least should implement RDNA 5, no? Specially given Intel will go with Celestial in PTL.RDNA 5?
In APUs?
If memory serves the point of Z5 is to be the big Tock, while the point of Z6 is to take care of the uncore. Clearly between new packagings, LP cores mixins, whatever multiple packagings they will have for server, client and mobile, it makes sense to go for the full package and for Z6 core to be the tick directly.Oh, it does. I was just surprised that they'll go directly into Zen 6 in 2026. Was expecting a Zen 5 refresh. AMD is being very aggressive with their timelines (As they should be).
I guess...but the rumour was RDNA 3.5 until 2027. It is the new Vega, in every way. Little CUs all over the APUs.RDNA 4 will be a dGPU only generation*. So STX+1 at least should implement RDNA 5, no? Specially given Intel will go with Celestial in PTL.
Right, that make sense. Specially with the unification of DT and Mobile into Client.If memory serves the point of Z5 is to be the big Tock, while the point of Z6 is to take care of the uncore. Clearly between new packagings, LP cores mixins, whatever multiple packagings they will have for server, client and mobile, it makes sense to go for the full package and for Z6 core to be the tick directly.
Was the rumor that AMD would use only RDNA 3.5 until 2027 or that 3.5 will keep appearing in the SKU line-up until 27? There's a important difference here.I guess...but the rumour was RDNA 3.5 until 2027. It is the new Vega, in every way. Little CUs all over the APUs.
I also haven't heard a single thing regarding RDNA 4.5 anymore. Feels like it went up in smoke, like tears in the rain.
It was a brief tweet in Korean, little details. But there is no reason to think that it will be ONLY RDNA 3.5 indeed.Right, that make sense. Specially with the unification of DT and Mobile into Client.
Was the rumor that AMD would use only RDNA 3.5 until 2027 or that 3.5 will keep appearing in the SKU line-up until 27? There's a important difference here.
If they get faster RAM, it could easily justify dropping the older uarch.For example,Vega GFX IP is still part of AMD "current" line-up with Barcelo-R (7730, 7530, etc). But AMD introduced newer GFX IPs in the meantime. I would be very surprised if AMD stand still.
But if it's the case that Medusa keeps 3.5 IP, it won't be the end of the world. 3.5 will still be a class leading IP and with adoption of LPDDR6/LPDDR5X 10700, AMD could certainly afford to go wider to counteract newer competitor GFX IP. Wouldn’t be surprised if Medusa brings a 20 CU RDNA (Whatever the IP version is) iGFX.
IF we are just talking CPU ignoring GPU its super workload specific , generally expect 0-5% more ipc. there will be outliers , gaming or specific games / engines being one of those.What should we expect the difference in ST integer and floating point performance to be between Strix Point and Halo? Would its cache and extra bandwidth also help those kinds of operations, besides NPU and GPU workloads?
So besides the larger RAM, there would not be much point in choosing Halo over Strix Point for professional applications. Thanks!IF we are just talking CPU ignoring GPU its super workload specific , generally expect 0-5% more ipc. there will be outliers , gaming or specific games / engines being one of those.
Depends , if the working set is large it can have a big impact. Like if your a guy who when you go CTL alt f9 in excel your waiting for minutes then it probably will make a noticeable difference.So besides the larger RAM, there would not be much point in choosing Halo over Strix Point for professional applications. Thanks!
Why, though? Excel is never RAM-bound even in the most obscene workbook. It is easily and often compute-limited. Why would Halo make a difference in that scenario?Depends , if the working set is large it can have a big impact. Like if your a guy who when you go CTL alt f9 in excel your waiting for minutes then it probably will make a noticeable difference.
Arrowlake = zen 5 + good chunkClosed captioned unavailable. Can you give a brief summary please ?
I little cryptic for me. Arrow lake is better than Zen 5 ???Arrowlake = zen 5 + good chunk