• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Discussion AMD SoC Halo series GPU discussion

Page 17 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
adroc_thurston said:
Not really, no. There's that one subtest.
Click to expand...
No, Zen 3 and Zen 3+ were competitive... They're falling behind since then and Strix Halo on the CPU side only solves what was already solved: good MT performance which the Ranges already offered.
 

Attachments

  • Gb7ac-3aMAAuBwE.jpeg
    Gb7ac-3aMAAuBwE.jpeg
    28.9 KB · Views: 21
adroc_thurston said:
No, Cezanne versus M1 wasn't that. Again, omnetpp.
Click to expand...
Zen 3 desktop was ahead of M1. Zen 3 laptop was 10% lower.
Zen 5 desktop is behind M4. Zen 5 laptop is 35% lower. Strix Halo might move that to only 30% lower. And that's spun as a huge and revolutionary change for the x86 world...

They're becoming less competitive. And the people coping with that are those stuck in the x86 ghetto.
 
ToTTenTranz said:
🙁




Not true.

View attachment 117522

View attachment 117523

View attachment 117524

View attachment 117513








If the 5070 mobile performs close to the 4070 mobile (which seems like a possibility due to SM performance regression), that's also not true.

View attachment 117520


BTW here's how the 1% performance looks like on the 4070m desktop equivalent (4060 Ti 8GB) running Indiana Jones, at the same 1080p Ultra settings:

View attachment 117526
Click to expand...
Those two Indiana Jones graphs are at two different graphical presets. Different graphics settings affect the average and minimum FPS that a GPU can achieve.

edit: just to clarify. The 65W 8GB 4070 mobile mobile achieves 30% better FPS in notebookcheck's testing than the 16GB 4060 ti desktop does in ComputerBase's testing. There's no way that's possible unless the testing conditions are vastly different.
 
Last edited:
insertcarehere said:
And a vast, vast majority of the people would look at the pricing, then go buy the Switch 2/Steam Deck instead and keep the change. The super high end gaming handheld market does not have enough volume to tap on.
Click to expand...

You seem awfully concerned about the market volume for high-end handhelds.

Worry not, if there are enough people that justify the R&D and manufacturing for all these 18" 4-5Kg hedious bricks that go for 4000-5000€:

1739904579748.png


Then I'm sure there's enough market for 1500-2000€ high-end handhelds.

And it's not like you have to buy one for yourself, either. Others being able to get one shouldn't be a problem for the ones who don't.


insertcarehere said:
In the theoretical dreamscape that such a cut down SKU would preserve the whole iGPU, maybe.
Click to expand...
The iGPU is in a separate chip from the ones with the CPU cores. They can mix and match whatever combination they want to.
In this case, all they have to do is take out the second CCD from the 395 Max and keep the rest the same.

insertcarehere said:
In the whole history of AMD APU releases we have never seen such a SKU offered.
Click to expand...
You never saw AMD offering SKUs like any of the Ryzen AI Max either, so we're past that barrier as well.
 
gdansk said:
Zen 3 desktop was ahead of M1. Zen 3 laptop was 10% lower.
Zen 5 desktop is behind M4. Zen 5 laptop is 35% lower. Strix Halo might move that to only 30% lower. And that's spun as a huge and revolutionary change for the x86 world...

They're becoming less competitive. And the people coping with that are those stuck in the x86 ghetto.
Click to expand...
Shouldn’t we wait till Zen6 and N2? Would be a fairer comparison, with the only difference being the ISA and design then.

adroc_thurston said:
and Apple is becoming less efficient
Click to expand...
Only at the high load and that’s cause more P cores than previous gen’s. But their battery life for light loads has improved with the M4 generation.
 
poke01 said:
Shouldn’t we wait till Zen6 and N2? Would be a fairer comparison, with the only difference being the ISA and design then.
Click to expand...
Don't need to wait. AMD already said it's a 10% generation. Even with a big clock rate increase it's behind.

But hey that'll look great in servers (still the only thing they design for despite having separate CCDs now). And I quite like having compromised server cores for desktop. But throughput monsters like the Zen 5/6 family are the wrong direction for client devices.
 
Abwx said:
Not throttling for a first run, but what about after 10mn, there s nothing on those graphs that say that it wont throttle after a run or two and for much of the 30mn.
Click to expand...
IMG_1690.png
The 14” M4 Pro CPU barely throttles after a 10 min.
 
jpiniero said:
People buy Macs to get away from Windows because Windows is so crappy now.
Click to expand...

Apple's laptop market share disagrees with this statement. It's been on a slow decline from when the M1 launched, pretty much, constantly sitting below 10%. Actually if memory serves me correctly, it's been like 8.5-9% for the last year?

Apple's user base isn't really growing.
 
Like someone said it's looking better than I expected, but the ability to get it into lower price range matters. First you have the Nvidia brand you have to go against and second the reality is you can't ignore $800 "cheap plastic" gaming laptops and the gaming market either. They are only $800 for a reason, so those that want a cheap one would get those.

It's not a vacuum.
poke01 said:
But sweesh is Intel missing out; they are always late..
Click to expand...
Not really. Their attitude of ignoring everything other than CPU over decades of time meant when they needed a good GPU, no one really cared and they couldn't bother to make a good product out of it. An ARC equivalent of Strix Halo wouldn't make them that more relevant.
Meteor Late said:
It means if AMD intends to use a similar packaging technology for all their 2027 laptop chips (which it's what I understand will do) it means battery life will never be great at this rate, it will be merely good.
Click to expand...
It's not just packaging. A larger die requires lot more work to keep static/idle power low. And it has lots of them.
uzzi38 said:
Apple's laptop market share disagrees with this statement. It's been on a slow decline from when the M1 launched, pretty much, constantly sitting below 10%. Actually if memory serves me correctly, it's been like 8.5-9% for the last year?
Click to expand...
9% for $1000+ laptops. That's a lot. They have a significant share of the high end ones.

It's same as their Smartphones. Only 20% share but 70%+ of revenue. They have the perfect business strategy.
adroc_thurston said:
They really aren't, they're skinny ass cores relative to comp. rofl.
In some ways Skymont is bigger, and that's a real favela core.
Click to expand...
Actual mm2 matters so no it's smaller.
 
Last edited:
poke01 said:
[View attachment 117497
This is a better comparison for creative tasks as it’s 50watts.
It’s close, against the M4 Pro. Really what saved Apple was that extra increase in P cores from 6 P cores on the M3 Pro to 10.

If the M4 Pro stayed at 6 P cores then Strix Halo would have destroyed the M4 Pro in CPU nT related tasks. But Apple did well with 10 P cores despite the core/thread advantage that AMD has and N3E helped Apple too.

Also looking at this I don’t see how M4 will be able to compete in 2026. Thinking M4 is enough to compete with Medusa Halo is crazy talk.

What’s more interesting is the GPU side of things. But I’ll do that later but AMD has no competition here.
Click to expand...
Given that the M5 design is already done, that’s what Medusa Halo will face.
 
poke01 said:
View attachment 117532View attachment 117533

That’s a big difference. This is the biggest advantage Apple has. Not throttling when unplugged and achieving better battery life.

How can AMD achieve the same? Is it a x86 Windows issue?
Click to expand...

Others don't show the same thing, so not sure (the battery life part, not the throttling part)? Might come down more to what power plan you have set while running on battery.

1739911896029.png

1739911601664.png
 
uzzi38 said:
Apple's laptop market share disagrees with this statement. It's been on a slow decline from when the M1 launched, pretty much, constantly sitting below 10%. Actually if memory serves me correctly, it's been like 8.5-9% for the last year?

Apple's user base isn't really growing.
Click to expand...

That might be true... but people opting for Windows are buying cheap stuff.

As in Strix Point is completely irrelevant because of the price.
 
jpiniero said:
Go look on Amazon and see what consumers actually buy. You'll see what I mean.

Strix Halo was absoultely designed for gaming first... and got repurposed for AI because of the hype.



People buy Macs to get away from Windows because Windows is so crappy now.
Click to expand...
I don’t think Strix Halo was ever for gaming and is not now. It’s going to cost more than alternative gaming solutions. People who buy Strix Halo for gaming are wasting money.

The main point of this chip is unified memory and GPU access to a huge pool of RAM. Fire Range is the gaming line.
 
The Hardcard said:
I don’t think Strix Halo was ever for gaming and is not now
Click to expand...
It is.
The Hardcard said:
It’s going to cost more than alternative gaming solutions
Click to expand...
Cost is a function of the volume.
The Hardcard said:
The main point of this chip is unified memory and GPU access to a huge pool of RAM
Click to expand...
The point is the halo APU.
The Hardcard said:
Fire Range is the gaming line.
Click to expand...
no that's a luggable line. actual gaming laptops anyone uses are STX1, HPT1 or soon nuff KRK1.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top