AMD Strix Halo APU Leak Reports the Performance of a laptop 4070, which is equivalent to the RX 6700 XT

[My_Computer_Is_Trash]

Summary

AMD's Strix Halo Mobile APU has been leaked and it appears to have the graphics performance of a laptop 4070. (The 4070 Laptop has the performance benchmarks of a Radeon RX 6700 XT) Estimations of its arrival are in the second half of 2024. This looks like AMD attempt to hit three birds with one stone. The birds being Intel, Nvidia, and Apple.

 

Quotes

Quote

"In an early Cinebench R23 result, AMD's Strix Halo processor is said to be 25% faster than their existing 16-core Dragon Range processor when both CPUs are limited to under 90 watts. Not only does Strix Point reportedly have a strong GPU, it also has a strong CPU. Add on AMD's rumoured 40 TOPS AI engine, and AMD are set to deliver strong competition for AMD, Nvidia, and Intel in 2024." "AMD are aiming to release a "Halo" product in 2024 that aims to take the mobile CPU market by storm. With Strix Halo, AMD are reportedly aiming to deliver a 16-core Zen 5 mobile processor will ship with a RDNA 3+ GPU, a dedicated AO processor and a wide 256-bit LPDDR5X memory controller. With this chip, AMD are challenging Apple's M-series of mobile processors, Nvidia's Max-Q series graphics chips, and Intel's mobile processors." "AMD are reportedly preparing multiple SKUs that will combat Nvidia on several levels. Their 40 CU model should deliver performance levels that are similar to sub-95W RTX 4070 Max-Q GPUs, their 32 CU model should compete sell with sub-65W RTX 4060 Max-Q laptops, and their lower-end 24CU and 20CU models should compete well with Nvidia's RTX 4050 laptops" (Overclock3D.net)

 

"The 40 CU RDNA 3.5 discrete GPU chiplet part is interesting as this would enable Strix Halo to rival a <95 W RTX 4070 Laptop GPU. Strix Halo SKUs will also offer 32 CU (rivals <65 W RTX 4060 Laptop GPU), 24 CU, and 20 CU (rival an RTX 4050 Laptop GPU) variants" "Strix Halo chiplet APUs targeting Apple silicon efficiency with RTX 4070-class graphics along with new Strix Point and Fire Range APUs"(NoteBookCheck). 

 

"Further backing up that claim is the discrete, 20WGP (work group processor), 40CU, 32MB infinity cache, RDNA 3.5 graphics chiplet that would afford the “Strix Halo” enough graphical prowess to rival a sub-95w Nvidia GeForce RTX 4070 mobile GPU" (LaptopMag). 

 

"it's said it will come as a “large chiplet APU” to rival the Apple M-series chips."(kitguru.net)

 

My thoughts

I hope AMD delivers at their normal budget price point. I would like to see this in desktops eventually.

 

Sources

https://overclock3d.net/news/cpu_mainboard/amd_s_next-gen_zen_5_strix_halo_apu_should_worry_both_apple_and_nvidia/1

https://www.laptopmag.com/news/amd-ryzen-8000-series-leak-macbook-tier-efficiency-4070-graphics-incredible-performance

https://www.kitguru.net/components/cpu/joao-silva/amd-zen-5-based-ryzen-8000-mobile-cpu-specs-reportedly-leak/

https://www.notebookcheck.net/Massive-AMD-Ryzen-8000-Zen-5-leak-Strix-Halo-chiplet-APUs-targeting-Apple-silicon-efficiency-with-RTX-4070-class-graphics-along-with-new-Strix-Point-and-Fire-Range-APUs.709446.0.html

 

More

The video that leaked these APUs has much more information and is here:

 

[jaslion]

Not sure I'd fully believe this but on the other hand that asus handheld was doing a lot in what is probably a 15w package.

 

Curious to see where this goes. It would for sure fit since amds best selling stuff has been apu's for over a decade.

[Agall]

12 minutes ago, My_Computer_Is_Trash said:

Summary

AMD's Strix Halo Mobile APU has been leaked and it appears to have the graphics performance of a 90 Watt 4070 Max-Q. (The 90 Watt 4070 Max-Q Has the performance benchmarks of a 1660ti) Estimations of its arrival are in the second half of 2024.

 

Quotes

 

My thoughts

I hope AMD delivers at their normal budget price point. I would like to see this in desktops eventually.

 

Sources

https://overclock3d.net/news/cpu_mainboard/amd_s_next-gen_zen_5_strix_halo_apu_should_worry_both_apple_and_nvidia/1

https://www.laptopmag.com/news/amd-ryzen-8000-series-leak-macbook-tier-efficiency-4070-graphics-incredible-performance

*salivating at the thought of a new APU build for my TV that I don't need*

 

I almost jumped on the 4700g, 5700g, but deferred it by using a GT1030 spare card for my home theater setup, but I prefer APU builds for the simplicity. Its really just a way for me to justify building an APU system for the real purpose of overclocking it.

Hoping we get an AM5 version of the Asrock Desk mini or equivalent.

[DrMacintosh]

If its not ARM its not a competitor to Apple Silicon.

[porina]

Trying to understand the claims here. While it is an "APU" in AMD speak, it is nothing like the socketed APUs we know. If this rumour is correct you'll get something like a current gen console in a chip. Soldered ram only. No upgrade path. It wont be cheap by itself but could open up interesting SFF gaming desktops and laptops.

 

Looking up some other ball park comparisons for a 4070 Max-Q which may better illustrate where it may fall, it is 3050, 1070, Vega 56. So kinda upper mid range dGPU from over 5 years ago level.

[Brooksie359]

53 minutes ago, porina said:

Trying to understand the claims here. While it is an "APU" in AMD speak, it is nothing like the socketed APUs we know. If this rumour is correct you'll get something like a current gen console in a chip. Soldered ram only. No upgrade path. It wont be cheap by itself but could open up interesting SFF gaming desktops and laptops.

 

Looking up some other ball park comparisons for a 4070 Max-Q which may better illustrate where it may fall, it is 3050, 1070, Vega 56. So kinda upper mid range dGPU from over 5 years ago level.

Yeah I feel like the comparison is pretty garbage. Who can actually think of the performance of a card like the 4070 95 watt variant. Why just compare it with a normal card? I hate it when people make a comparison to something very obscure making it so you don't know any more than you did prior to the comparison. 

[My_Computer_Is_Trash]

4 minutes ago, Brooksie359 said:

Yeah I feel like the comparison is pretty garbage. Who can actually think of the performance of a card like the 4070 95 watt variant. Why just compare it with a normal card? I hate it when people make a comparison to something very obscure making it so you don't know any more than you did prior to the comparison. 

I did. I compared it to the 1660ti

 

1 hour ago, My_Computer_Is_Trash said:

AMD's Strix Halo Mobile APU has been leaked and it appears to have the graphics performance of a 95 Watt 4070 Max-Q. (The 95 Watt 4070 Max-Q Has the performance benchmarks of a 1660ti) 

[Brooksie359]

Just now, My_Computer_Is_Trash said:

I did. I compared it to the 1660ti

 

 

I know but if you had to compare it to the 1660ti then you are basically confirming that the original comparison is bad. Why include the first comparison then at all?

[My_Computer_Is_Trash]

1 minute ago, Brooksie359 said:

I know but if you had to compare it to the 1660ti then you are basically confirming that the original comparison is bad. Why include the first comparison then at all?

Could you provide more clarity?

[Brooksie359]

6 minutes ago, My_Computer_Is_Trash said:

Could you provide more clarity?

What isn't clear? The 4070 comparison is bad and it would have been better to just compare it to the 1660ti and be done with it. 

[My_Computer_Is_Trash]

18 minutes ago, Brooksie359 said:

What isn't clear? The 4070 comparison is bad and it would have been better to just compare it to the 1660ti and be done with it. 

"The 40 CU RDNA 3.5 discrete GPU chiplet part is interesting as this would enable Strix Halo to rival a <95 W RTX 4070 Laptop GPU. Strix Halo SKUs will also offer 32 CU (rivals <65 W RTX 4060 Laptop GPU), 24 CU, and 20 CU (rival an RTX 4050 Laptop GPU) variants" (NoteBookCheck). 

I did some more research. And as it turns out, the GPU comparison website I used gave me false results when comparing the laptop 4070 to the 1660ti. I will edit my post.

[My_Computer_Is_Trash]

1 hour ago, DrMacintosh said:

If its not ARM its not a competitor to Apple Silicon.

It competes with performance per watt.

[DrMacintosh]

4 minutes ago, My_Computer_Is_Trash said:

It competes with performance per watt.

Yeah but if it's still on the traditional architectures that isn't going to last more than a generation and will eventually run into a wall. All the existing legacy processes are running up against a wall. The industry needs to move away from legacy support and start innovating. 

[HenrySalayne]

There is a single source in the wild and it's MLID - so basically no source at all. ¯\_(ツ)_/¯

Where is the Titan RTX he promised in December? Or the $750 4070?

 

There is little to no concrete and new information here and it's coming way later than the dates MLID hallucinated. When we figure out how to use all the hot air MLID creates, mankinds energy problems are a thing of the past.

[starsmine]

1 hour ago, HenrySalayne said:

Where is the Titan RTX he promised in December? Or the $750 4070?

Things he never promised for 1000 alex.

Between
https://en.wikipedia.org/wiki/Analytic_confidence
and 

final decisions not being made yet. I don't know what more I can say.

don't come back with "oh he gets a free pass if he is wrong"
Again, I will point you to studying up on Analytic confidence because no, you don't get a free pass. 

https://www.cisecurity.org/ms-isac/services/words-of-estimative-probability-analytic-confidences-and-structured-analytic-techniques

If you are regularly missing your buckets, you are not doing a good job. 

Not to defend the Titan one, but again, what did he say? that Nvidia was working on prototypes, or that Nvidia was planning to launch one in the next quarter.

You act like the correct way to live life is to not adjust your analysis as more information comes in. Just be 100% confident and be 100% right.

[porina]

2 hours ago, Brooksie359 said:

Who can actually think of the performance of a card like the 4070 95 watt variant.

There may be two reasons for choosing a 4070 Max-Q 90W:

1, people who don't know what Max-Q 90W means just see 4070 and think, it will perform like a $600 dGPU of today.

2, It might overlap in market uses if it is ever released in this claimed form, although given the timescales involved it'll be getting close to going against 50 series.

[GOTSpectrum]

In all honesty, I would love AMD to just go ham with a crazy APU, give it 6-8 cores, and whatever silicon they can muster for a GPU. I mean go big or go home kinda deal. 

 

I know it would be used in something, as a console/mini gaming PC chip or even if they can as an AM5 chip. Although on am5 they would probably need to add a stack of HBM. Look at what apple has done, think of it as an M1 moment for AMD, they are probably one of the few companies in the world rn that has the tech, expertise and customer base for it. Intel is catching up fast in regards to packaging and GPUs and Nvidia are working on an enterprise combined solution, but if they tried they could get it out in 6-12 months leaning on the semi-custom team. 

 

I always wanted them to make a threadripper with a decent CPU and half the cores... this could be a spiritual successor or hell even make it a threadripper. (not I think AMD really care about that segment rn) But just think if they made a rendering/workstation/gaming beast line. 8-24 cores, and have the rest of the space filled with a GPU and HBM. Aim it at super high-end, a drop in complete SoC that just works. 

I know they will never do it but I can dream 

[HenrySalayne]

5 minutes ago, starsmine said:

Not to defend the Titan one, but again, what did he say? that Nvidia was working on prototypes, or that Nvidia was planning to launch one in the next quarter.

We talked recently about it in another thread. I'm sure you can find it.

MLID is not a reliable source, he is filling the generally huge gaps in his "leaks" with made-up sensationalism. Nobody should give his kind of "news" a stage.

 

Do you know that cults always have armageddon just around the corner? They will never set a date far in the future. It's at best a few years away. This creates urgency and makes the followers more susceptible for the preachings.

 

A Zen 5 mobile APU in Q1 2024 is so unlikely, no reputable news outlet would even mention it. But here we are with MLIDs preachings...

[Mark Kaine]

7 hours ago, DrMacintosh said:

If its not ARM its not a competitor to Apple Silicon.

why not... small form factor, good perf/w... of course it competes... its an alternative,  theoretically a better one (cheaper,  lower power usage etc)

[leadeater]

6 hours ago, starsmine said:

You act like the correct way to live life is to not adjust your analysis as more information comes in. Just be 100% confident and be 100% right.

The issue is the efficacy and credibility of the "analysis". Put it this way, people believe in tea leaf readings.

 

When one dabbles in the ridiculous one must accept the reputation that comes along with it. There are plenty of alternative sources of information that do not suffer from this issue, there is a good reason for that. Not everyone just outright believes the information or thinks theirs is 100% accurate but since they have proven to be credible and reliable enough what they say gets given the time of day. If MLID is your jam all the power to you, I however don't believe in tea leaf readings.

[Zodiark1593]

16 hours ago, DrMacintosh said:

Yeah but if it's still on the traditional architectures that isn't going to last more than a generation and will eventually run into a wall. All the existing legacy processes are running up against a wall. The industry needs to move away from legacy support and start innovating. 

If Ryzen didn’t exist, I’d have probably have finally given some merit here. However, this is not at all the case. The so-called legacy architectures are not standing still, with fierce competition driving heavy investment, and the future of x86-64 (noting in particular AMD’s Epyc gaining server marketshare, and the current consoles using them) seems more solidified than ever. 
 

And, this doesn’t include the possibility of further extensions down the road. 
 

At this point, the only way that moving from x86-64 would make sense, is if pursuing a RISC-V (I no longer see possibility of ARM succeeding x86-64) architecture with fast x86-64 emulation, is more cost effective. 
 

Edit: A bit off topic, but calling the cores Apple uses ARM, is a bit of a disservice to what they’ve created. Really, they’re Apple cores, that happen to be ARM compatible. 

[xAcid9]

<95w 4070 laptop is equal to RX 5700 XT/RTX 3060 at best. 

Assuming it runs on LPDDR5X@8000MT/s, the bandwidth will be like 256GB/s?

If that the case it probably will perform like RX6600 or slightly better unless it come with significantly improve infinity cache. 

 

 

also MLID as source = weak hearsay.

Someone make a spreadsheet of MLID accuracy for AMD Zen3+4, he's just over 50% accurate. Laughable for a "reputable" "leaker" 
https://docs.google.com/spreadsheets/d/1a7OWkeyHcdFgoQ395zmi7RlWluAKCv9bG_2silBQU5Q/edit#gid=0

[WallacEngineering]

I seriously doubt it. That would be like 7-8 generations worth of improvement in a single generation when we are getting close to the edge of not being able to shrink transistors any smaller without significant engineering issues or changes. Theres just no way. I said the same thing on the YT video that features the news. Its BS

[WereCat]

MLID... uh...

[RVRY]

I find it hard to believe AMD would release this unless it's quite literally a Ryzen CPU and RDNA 3 GPU stapled together on the same package. I'd only expect to see a chip like this on a console or high-end NUC because it'd be pretty massive...