Dunno how they proceeded given that AT also use GCC and had allegedly 0% in INT,Interesting that Spec with GCC shows that much higher IPC gain versus test with Clang. I wonder what flags they were using as well.
Dunno how they proceeded given that AT also use GCC and had allegedly 0% in INT,
their number in this matter is about the same as to what AMD displayed in the few INT based tests.
AT didn’t normalize for clock speed and apparently didn’t notice that the CPU in their laptop was thermally throttling during the ST tests, so you can’t get any IPC data from AT’s testing, unfortunately.
Well, using the latest datas we can at least deduct by how much their CPU was throttling, assuming of course that it was the only cause.
Well if you insist.Hey, bring that back. This thread can't become any more of a dumpster fire than it already is.
@FlameTail already pointed out It's bigger and also using a newer process.Strix Point laptops are expensive because there s no competition in this segment but since the chip is the same size or so as Hawk Point it will gradually be cheaper as time goes by.
Strix Halo uses N4 + N3 and the size is 2x ~66mm² + ~300mm² IOD including the IGP according to leak.Strix Halo cost more to manufacture but that s surely less than say a 8840 APU + RX 7600 chip, and AMD will cash on both the CPU and GPU, so on the mid term it will also be substancially cheaper once some RD cost is amortized.
Just looked at the version history, 13.3 came out May 21, 2024 and the latest 14.2 came out yesterday. The version history isn’t sequential which is interesting (as there were two releases in between with lower version numbers). Y’all probably know why better than me. But in terms of versioning it was the highest at the most recent time.Even though they are using GCC, they are using a much older version than the linked twitter post, so you won't get a super accurate prediction that way either. Bad data is just bad data*, not much you can do with it. I previously estimated around 5% throttling based on the other tests AT ran, but it's a very rough estimate and Spec is a much longer test than everything else, so it's very possible, if not probable, that there was more throttling during at least some of the Spec tests than the others.
(*I'm saying it's bad data to try and calculate IPC, I have no reason to believe it's not fine as a measurement of the performance you get from STX in that particular laptop).
Just looked at the version history, 13.3 came out May 21, 2024 and the latest 14.2 came out yesterday. The version history isn’t sequential which is interesting (as there were two releases in between with lower version numbers). Y’all probably know why better than me. But in terms of versioning it was the highest at the most recent time.
How much would that affect the results? I can’t imagine it would be materially.
The Zenbook S 16 reviews were all rushed.Guess that we need a few more tests to definitly have an accurate picture, anyway there s only 5 day left before everything being cristal clear.
Not comparable. Compiler patches revealed the horrible double issue restrictions pretty early. It was "wtfed" pretty very early in the hype cycle.>RDNA double-issue flashbacks.
Z4 already has dual SDP per CCD (EPYC GMI wide), so at the very least with dense fanout interconnect they can enable both SDPs and still consume less than half the energy compared to DT CCD.
So it would have double the BW of the DT CCD if they do this at least.
But that sounds like bare minimal effort, some new innovation should be there.
The other interesting aspect is the MALL, which can do aggressive prefetching for hiding memory latencies (as done on MI300), the one on RDNA2/3 is not capable of this.
I again hope this is the one they use not the one from RDNA3.
Strix Halo is really big. Anyone hoping It It will be relatively cheap should forget about that, just look at what they ask for a Strix Point laptop.
My prediction is >2000 euro and for that only 4070 80W level of performance is not very good.
In my opinion Strix Halo is not aimed for gamers, that's just secondary. The main selling point is the 16C32T CPU paired with 64-128GB RAM.
This is it. It’s not a gamer part. It comp will be the M3 Max and ML usage. The 128GB SKU will likely be >$2500The main selling point is the 16C32T CPU paired with 64-128GB RAM.
MoP means SKU spam.
Exactly, if you’re going to solder RAM might as well use the best implementation. MoP saves board space and enables higher busses on laptops. AMD just didn’t want to go all out I guess.Intel has so far managed with only 2 SKUs, IIRC.
Strix Halo uses N4 + N3 and the size is 2x ~66mm² + ~300mm² IOD including the IGP according to leak.
178 mm² 8840 uses N4 + 204mm² RTX 7600 using N6.
Yet you think It will be cheaper to manufacture than that CPU+dGPU combo. Not happening.
I'm pretty sure it is a gamer part by design. Why else would it use an RDNA variant which still doesn't have good ROCm support. They'll have to market it as something else because it isn't competitive where it was aiming.This is it. It’s not a gamer part. It comp will be the M3 Max and ML usage. The 128GB SKU will likely be >$2500
@FlameTail already pointed out It's bigger and also using a newer process.
So It will likely get somewhat cheaper with more manufacturers offering them not just Asus, but not as cheap as Phoenix was(is).
Strix Halo uses N4 + N3 and the size is 2x ~66mm² + ~300mm² IOD including the IGP according to leak.
178 mm² 8840 uses N4 + 204mm² RTX 7600 using N6.
Yet you think It will be cheaper to manufacture than that CPU+dGPU combo. Not happening.
Have you seen the LNL SKU list?Intel has so far managed with only 2 SKUs, IIRC.
Because that's the only GFX IP AMD has that draws triangles.Why else would it use an RDNA variant which still doesn't have good ROCm support
It's a lot more performant than that.178 mm² 8840 uses N4 + 204mm² RTX 7600 using N6.
So since people who so far presented SPEC results are using different environments and different compilers with different compiler options, comparing their results is more like comparing apples to oranges than apples to apples.
Have you seen the LNL SKU list?
It's a lot more performant than that.
thats intel binning by clock and gpu count. If AMD binned by RAM there should only be 3 SKUs.Have you seen the LNL SKU list?