Discussion AMD SoC Halo series GPU discussion

Page 40 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

adroc_thurston

Diamond Member
Jul 2, 2023
5,493
7,681
96
I didn't exactly mean price, more like the number of available models, but price is also high, It starts at 1 749 €, but this one is with AI 9 HX 370 + RTX 4060, the rest are business and most of them are from Asus.
Premium part starts at premium money? nooooooooooo waaaaaay.
Yeah, overpriced corporate stuff.
Kinda cool tho.
 
Reactions: lightmanek

LightningZ71

Platinum Member
Mar 10, 2017
2,134
2,586
136
KRK is for when you SPECIFICALLY want the ST improvements available with Zen5 mobile and don't care in the slightest in any sort of beyond casual gaming performance. Unfortunately, that won't be much of an uplift at all because the top KRK SKU boosts to only 5.0Ghz, whereas HWK boosted to 5.2GHZ. Further more, all of HWKs cores are P cores of the Zen4 vintage, where half of KRK cores are E cores. Hawk likely has a better iGPU with 12CU instead of only 8, likely also with the cut down parts. HWK also has more PCIe lanes.

Honestly, as long as HWK is on the market, there's almost zero point to ever buy a KRK part.

If AMD had really wanted to do something special there, KRK would have had 2xP cores, 6xE cores, and 12CU. The P cores would have been specced to boost to 5.4Ghz and the RDNA3.5 would have been able to run the iGPU faster than the one in HWK. It would have made a better portable game console chip with ST and GPU performance improvements over the Z1 extreme. It would have sacrificed a tiny bit of MT performance to HWK, but that completely didn't matter in that market anyway. Oh, and leave at least 4 more PCIe lanes for better I/O.
 

adroc_thurston

Diamond Member
Jul 2, 2023
5,493
7,681
96
KRK would have had 2xP cores, 6xE cores
that's not gonna work layout-wise.
Honestly, as long as HWK is on the market, there's almost zero point to ever buy a KRK part.
lmao no KRK has more CPU perf and the same gfx perf as hpt1.
It would have made a better portable game console chip with ST and GPU performance improvements over the Z1 extreme.
wow a tiny irrelevant market when the KRK target is explicitly commercial laptop.
Which it, well, won. Kind of.
 

Trovaricon

Member
Feb 28, 2015
30
51
91
that's not gonna work layout-wise.

lmao no KRK has more CPU perf and the same gfx perf as hpt1.

wow a tiny irrelevant market when the KRK target is explicitly commercial laptop.
Which it, well, won. Kind of.
How can Krackan 512SPs match Phoenix 768SPs when the GPU uarch is basically the same and frequency is not that much higher? Krackan has also less ROPs - half (16 vs 32), doesn't it?
 

adroc_thurston

Diamond Member
Jul 2, 2023
5,493
7,681
96
How can Krackan 512SPs match Phoenix 768SPs when the GPU uarch is basically the same and frequency is not that much higher?
Turns out real workload frequency between RDNA3 and 3.5 differs by quite a lot.
Especially at 20/25W.
Krackan has also less ROPs - half (16 vs 32)
who cares weenie iGPs are never fillrate limited to any relevant degree.
 
Reactions: Tlh97 and madtronik

reaperrr3

Member
May 31, 2024
89
288
86
How can Krackan 512SPs match Phoenix 768SPs when the GPU uarch is basically the same and frequency is not that much higher? Krackan has also less ROPs - half (16 vs 32), doesn't it?
- I've seen Phoenix iGPU's real clocks drop as low as somewhere around 1-1.2 GHz under very heavy load + lower TDPs.
- Some of the RDNA3.5 uArch improvements should at least slightly improve IPC per CU, too.
- Full 12CU PHX IGP may also be a little more cache/bandwidth-constrained sometimes, limiting its real-world advantage further.
- I think in most real-world scenarios, 12CU PHX was already only ~20% faster than 8CU SKUs. That's the average perf advantage of the 8700G vs. 8600G according to Computerbase, at least.

Basically, KRK's IGP probably only needs to clock around ~30% higher on average under load to catch up in real-world perf, which shouldn't be that hard with smaller spec + uArch polished towards more efficiency with feedback from Samsung from their former Exynos collab.
 

LightningZ71

Platinum Member
Mar 10, 2017
2,134
2,586
136
that's not gonna work layout-wise.
- with the current floorplan, definitely, but that could have been a floorplan design decision. It's not a show stopper.
lmao no KRK has more CPU perf and the same gfx perf as hpt1.
- the 890m is barely faster in real game benchmarks than the 780m in HWK. Synthetics show a better speedup, but it's never more than 25-30%. Chopping half of that off and clocking it 20-30% higher than HWK isn't going to magically make it faster.
wow a tiny irrelevant market when the KRK target is explicitly commercial laptop.
Which it, well, won. Kind of.
- it's a growing market, and given AMD's historically low mobile APU volume numbers, every little bit helps. It was a design time decision that doesn't even make sense on paper.

Just taking HWK and simply doing what needed to be done to take that RTL to N3P would have better you a better product than KRK all around and would have even rivaled SPT in many areas. Yes, copilot+ is a thing, but it has precious few uses. If they were feeling REALLY froggy, they could have uprated the NPU at the same time. It was never going to happen, but it was a possible choice.
 

adroc_thurston

Diamond Member
Jul 2, 2023
5,493
7,681
96
Chopping half of that off and clocking it 20-30% higher than HWK isn't going to magically make it faster.
yeah it does.
it's a growing market, and given AMD's historically low mobile APU volume numbers, every little bit helps.
bro please KRK won commercial. Stop coping, it's embarassing.
Just taking HWK and simply doing what needed to be done to take that RTL to N3P would have better you a better product than KRK all around and would have even rivaled SPT in many areas
just flat out embarassing poasting.
 

LightningZ71

Platinum Member
Mar 10, 2017
2,134
2,586
136
I feel fairly strongly that the publicly available benchmarks for Strix Point and Hawk Point support my position. Strix Point is not much faster than Hawk Point when using the iGPU. You are proposing that Kraken Point, built on the same node, with the same logical blocks, with it's 8CUs will be as fast as or faster than Hawk Point's 12CU RDNA3 implementation when the Ryzen AI 9 365, with it's 12 CUs, is neck and neck with Hawk Point in many cases.

As for winning commercial...

Here's all 15 design wins that it has as of two weeks ago, built on about 9 different actual motherboard platforms. How much of this "winning" is really impacting AMD's bottom line? That's a TON of "winning" from some of the absolute giants in commercial sales, oh, it's got some Lenovo to be sure, but that's one Ideapad and some Yogas that I barely see in corporate. Yes, we do expect to see some HP and Dell, but I'm not holding my breath...


Face it, Kraken point is an exercise in cost optimization and not value. The gains from N3P would have the Zen4 core doing high enough clocks to keep pace with mobile Zen5 at the released clocks. Zen 5's big gains outside of DC were in AVX-512 throughput, which the mobile cores only have about half of. N3P would have helped RDNA3 achieve or sustain higher clocks as well.

I'm going to let the final benchmarks, once they get run by 3rd parties, prove this all out. KRK is going to be a fine chip, but, it's not going to be much of an improvement over HWK.
 

adroc_thurston

Diamond Member
Jul 2, 2023
5,493
7,681
96
I feel fairly strongly that the publicly available benchmarks for Strix Point and Hawk Point support my position. Strix Point is not much faster than Hawk Point when using the iGPU. You are proposing that Kraken Point, built on the same node, with the same logical blocks, with it's 8CUs will be as fast as or faster than Hawk Point's 12CU RDNA3 implementation when the Ryzen AI 9 365, with it's 12 CUs, is neck and neck with Hawk Point in many cases.
you need to benchmark more then.
Face it, Kraken point is an exercise in cost optimization and not value
it's both.
The gains from N3P would have the Zen4 core doing high enough clocks to keep pace with mobile Zen5 at the released clocks
why the hell would you put old old IP on a shiny new very expensive node. Are you insane?
but, it's not going to be much of an improvement over HWK.
It is a very tangible improvement over HPT1 in many ways.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |