Strix has 5 times the design volume of Hamoa so they don't need to.I am not sure AMD would be able to get something like that going.
With Strix Halo I'm thinking models will be more the $999-$1499 segment.
APU only models don't seem like they're going much over that. Anything priced toward $2000+ range is Strix Point + dGPU (4070 tier), so I'd expect strix halo designs to be priced in that area since the dGPU performance should probably be somewhere around 4070 range.Strix Point is like the top end of that range and higher. Halo's gonna be more than that.
I mean maybe $1999 for the 128gb models, $999 for 16gb units. With all sorts of various form factor pricing within.Strix Point is like the top end of that range and higher. Halo's gonna be more than that.
Maybe, krk is the cheap part though? To me I'd prefer strix GPU less, the point of it for me is that it has very decent GPU performance and stellar CPU performance with decent battery. Why bother with the GPU, I want powerful glue though (in the laptop space, it's pointless in desktop)I mean maybe $1999 for the 128gb models, $999 for 16gb units. With all sorts of various form factor pricing within.
I mean when I look at amazon, 4080 laptops are under 2k, thats a lot silicon you get on one device, I see no reason to assume they're gonna price the chip so high when it really shouldn't be too big area wise at the end. 2x CPU chiplets and a IO/GPU chiplet the size of a midrange GPU isn't a crazy cost factor to AMD.
Its like, they could price the 9950x at $999, but they won't because they know it wouldn't sell. Why would AMD or the OEMs try to sell 4060 level performance at 4090 prices? they wouldn't.
Also strix point is definitely the successor to phoenix, as time goes on its gonna get quite cheap.
Kraken 1and 2(bin) are more really the true successors to Hawk Point/Phoenix SKUs. Strix Point is meant to be a cut above.Also strix point is definitely the successor to phoenix, as time goes on its gonna get quite cheap.
DT CPUs remain at AMD, so ASUS could not leak themKinda strange its only these these laptops scores thats leaking online, Ryzen 9000 series have been out in the wild 2-3 weeks already and so far airtight
Guess they haven't learned that geekbench is a big no no since you dont get a result if score are not uploaded uploaded automatically
(except GB3)
No they are not... If that were the case people would be in for lots of gray hairs come releaseday thanks to hordes of gemlings in RC agesaDT CPUs remain at AMD, so ASUS could not leak them
The gains seems tangible, looking to beat 7950X scores with Strix 5.1G but in the current context not great either.Another GB6 ES run.
ASUSTeK COMPUTER INC. ASUS Zenbook S 16 UM5606WA_UM5606WA_00014511B - Geekbench
Benchmark results for an ASUSTeK COMPUTER INC. ASUS Zenbook S 16 UM5606WA_UM5606WA_00014511B with an AMD Eng Sample: 100-000000994-37_Y processor.browser.geekbench.com"processor_frequency": {
"frequencies": [
4822,
4824,
4824,
4825,
4808,
4824,
4824,
4815,
4814,
4824,
4824,
4819,
4696,
4825,
4826,
4827,
4830,
4823,
4814,
4828,
4830,
4830,
4826,
4831,
4829,
4830,
4828,
4830,
4830,
4828,
4829,
4830,
4830,
4829,
4830,
4830,
4741,
4822,
4809,
4824,
4797,
4815,
4823,
4819,
4823,
4810,
4831,
4888,
4952
]
}
Frequency is reported properly this time, so we're looking at production samples to score around 2950-3000, just like Huang said.
So yeah, if you want ST, Apple is still the only way to go.
I only superficially followed the discussion about David Huang's posts, and now I realize I may have gotten the wrong impression based on incomplete reading. Assuming David's info is accurate, how many Zen5 cores are there? Are we talking about Zen 5 desktop, Zen 5 mobile, and then two corresponding variants for Zen 5C?I think based on David Huangs assessment, some unknowns wrt DT in terms of halved SIMD throughput
There is also the case of the reduced op cache, additionally the compiler patches indicated reduced FADD cycles but seems David is saying increased FADD latencies.
Just scaling the clock would put the GB6 DT scores of ~3500@5.7G without any contributions from the elimination of reduced SIMD throughput.
For the record, here is the Google translated David's Huang blog page.I only superficially followed the discussion about David Huang's posts, and now I realize I may have gotten the wrong impression based on incomplete reading. Assuming David's info is accurate, how many Zen5 cores are there? Are we talking about Zen 5 desktop, Zen 5 mobile, and then two corresponding variants for Zen 5C?
My initial impression was that Zen 5C was the one with reduced SIMD throughput and different cache implementation.
I am not a Chinese speaker so Google Translate could have failed meMy initial impression was that Zen 5C was the one with reduced SIMD throughput and different cache implementation.
Youth Edition Big core
- CCX0: 4 cores with 16MB L3 cache
Zen 5 Youth Edition Big Core: Compared with the full version of Zen 5 for desktop/server, its maximum frequency is reduced from 5.7 GHz to 5.1 GHz, SIMD throughput is halved, and the corresponding L1 vector load bandwidth is also halved- CCX1: 6 or 8 cores with 8 MB L3 cache
Zen 5c small core: uses exactly the same microarchitecture as the large core, and continues to reduce the Fmax target at the back-end physical design level to reduce the area, with a frequency not exceeding 4 GHz.
Observable reduction in macro-op cache. Zen 4's throughput drop in the 8-12KB range (corresponding to 4k-6k instructions) is more gradual than Zen 5's, and it is speculated that Zen 5 reduced the macro-op cache to the same 4k instructions as Zen 3.
When the core runs two SMT threads, each can monopolize a decoder so that the x86 decoding throughput limit of the entire core reaches 8 in most cases
Yeah we know ES perf already.Another GB6 ES run.
ASUSTeK COMPUTER INC. ASUS Zenbook S 16 UM5606WA_UM5606WA_00014511B - Geekbench
Benchmark results for an ASUSTeK COMPUTER INC. ASUS Zenbook S 16 UM5606WA_UM5606WA_00014511B with an AMD Eng Sample: 100-000000994-37_Y processor.browser.geekbench.com"processor_frequency": {
"frequencies": [
4822,
4824,
4824,
4825,
4808,
4824,
4824,
4815,
4814,
4824,
4824,
4819,
4696,
4825,
4826,
4827,
4830,
4823,
4814,
4828,
4830,
4830,
4826,
4831,
4829,
4830,
4828,
4830,
4830,
4828,
4829,
4830,
4830,
4829,
4830,
4830,
4741,
4822,
4809,
4824,
4797,
4815,
4823,
4819,
4823,
4810,
4831,
4888,
4952
]
}
Frequency is reported properly this time, so we're looking at production samples to score around 2950-3000, just like Huang said.
So yeah, if you want ST, Apple is still the only way to go.
It seems mobile gets halved SIMD throughput which means equal to Zen4 more or less.
This is his observationThe throughput of 128/256bit SIMD integer addition operations is halved compared to Zen 4, but 512bit remains unchanged. It is speculated that this problem only exists on the Zen 5 core with halved SIMD, which may be related to port allocation;
Only 20 more daysYeah we know ES perf already.
Retail Si on retail software is what matters, still no good info there.
Individual subtests are pretty consistent with a non-ES run from a few weeks ago.Yeah we know ES perf already.
Retail Si on retail software is what matters, still no good info there.
STX is running at 4.85 GHz. But you're right in the sense that it looks mid.Comparing that 5.1GHz Strix ST result to a random Hawk Point@5.1GHz doesn't look mighty impressive:
Is this fine? Different TDPs shouldn't matter for ST, right?
What are the frequencies of this oneIndividual subtests are pretty consistent with a non-ES run from a few weeks ago.
STX is running at 4.85 GHz. But you're right in the sense that it looks mid.
I've linked this a few pages ago.
Over 9000!!!!!