AMD Ryzen 8000 "Strix Point" APU Leak Points to 16 RDNA 3.5 CUs

[Vayra86]

AMD can sell APUs, can't dictate where and how those APUs will be used.

LOL? What? Of course they can, they own the IP. Apple might have some tips and tricks to gain control over their value chain hm? Or Intel. Or Nvidia. Or... [insert pretty much any company].

This is a strategic choice, not impossibility. They can simply allow or disallow companies to release products with or without certain specs combined with their APUs. Intel created Ultrabook that way, for example. You set boundaries, you create product groups and force resellers to adhere.

AMD however is on a different form of logic, they think total freedom is the best way to get their product to better market share. It hasn't ever worked. But they keep at it, on some weird principle or something I don't know. Their GPU logic is the same wrt marketing. Its one WTF moment after another.

The only thing they have showing for that strategy is that they've always had revenue. Not profit, but revenue, certainly. Its also why after decades of pretty good products people still think of AMD as the lesser being of the 2 or 3 competitors. This is part of the reason why AMD is always competing on price even if they do have solid product. After all Zen isn't missing a featureset like GPUs do relative to Nvidia. It's missing its time to market, it dragged Vega along far too long, etc etc. All of this is timing and marketing strategy. And let's not even begin about the motherboard support fiasco that they keep choking on.

 

[Space Lynx]

So you can't figure out why their fastest integrated GPU isn't available in midrange laptops? And that's AMD's fault? Dumb as bricks, what?

Or are you calling AMD dumb as bricks just because they're not selling those APU's for a lower price? You know that they have loads of slower chips that they have to sell as well, right?

Notebooks mit Architektur AMD: Zen 4 (ab 2022), GPU - NVIDIA GeForce: (iGPU) Preisvergleich Geizhals EU

Preisvergleich und Bewertungen für Notebooks mit Architektur AMD: Zen 4 (ab 2022), GPU - NVIDIA GeForce: (iGPU)

geizhals.eu

amd makes exclusive deals all the time to push specific product lines. they could have teamed up with someone and sold it exclusives at best buy, which they have done with previous laptops, i don't really buy this is an excuse anymore as I have evidence AMD doesn't rely on other companies, they can push through something if they want to.

like the liquid metal 6800m laptop, best buy exclusive ^

 

[R0H1T]

The DGPU market has stagnated due to Nvidia's pricing, AMD's stance on that pricing and Intel's realistic pricing but less then optimal performance.

It's not necessarily due to these 3 doing nothing, it's just that new nodes are almost exponentially hard to turn to profits than even 5 years back! You have fixed costs that need to be recouped & then the prices of memory/PCB & other components just keeps on rising. There's also the fact that if you put 64bit wide bus on them & try to sell no one will probably buy them calling you cheapskates & if you go any higher you overshoot your budget. You also have to remember these things are competing with Apple these days & they have 1024(?) bit LPDDR5x memory at the top end. Everyone wants cheap GPU's for anaemic loss making prices but hey that's not how capitalism works! The simple fact is there's simply not much volume there as it was a decade back when the PC market was still booming. And you know what 4nm TSMC costs are

In an ideal world we could keep shrinking the Si down to infinite levels, but again that's not how reality works!

 

[john_]

LOL? What? Of course they can, they own the IP. Apple might have some tips and tricks to gain control over their value chain hm? Or Intel. Or Nvidia. Or... [insert pretty much any company].

This is a strategic choice, not impossibility. They can simply allow or disallow companies to release products with or without certain specs combined with their APUs. Intel created Ultrabook that way, for example. You set boundaries, you create product groups and force resellers to adhere.

AMD however is on a different form of logic, they think total freedom is the best way to get their product to better market share. It hasn't ever worked. But they keep at it, on some weird principle or something I don't know. Their GPU logic is the same wrt marketing. Its one WTF moment after another.

The only thing they have showing for that strategy is that they've always had revenue. Not profit, but revenue, certainly. Its also why after decades of pretty good products people still think of AMD as the lesser being of the 2 or 3 competitors. This is part of the reason why AMD is always competing on price even if they do have solid product. After all Zen isn't missing a featureset like GPUs do relative to Nvidia. It's missing its time to market, it dragged Vega along far too long, etc etc. All of this is timing and marketing strategy. And let's not even begin about the motherboard support fiasco that they keep choking on.

I would "LOL? What?" also if I was choosing to avoid reality. AMD is not a behemoth like Intel with it's own fabs. If Dell goes to Intel and asks if they can provide a million CPUs, Intel will reply "certainly". If Dell goes to AMD and asks AMD if they can provide a million CPUs, AMD will reply "Let us check with TSMC first. Also let us check if the allocation of wafers from TSMC for the specific product will be enough or we will need to reallocate more from another product line. And even then we don't know if it is in our interest to do so". Why do you think even in AMD's best years of Ryzen, big OEMs where insisting Intel? Why do they insist on 10nm Intel with limited P cores and questionable integrated graphics and worst battery efficiency? Because Intel can provide. On the other hand AMD seems to be providing every handheld manufacturer with Zen 4 and RDNA3 APUs, while big OEMs insist on older Zen 3 and Vega series. Why? Maybe because AMD can warranty huge quantities to OEMs of older APUs, but not the newest, meaning handheld makers with vast less needs get those easier.

If you have better explanation than "LOL, What? They are just stuck with a dumb strategy", please, be my guest.

I do agree on Vega. They waited Intel to become somewhat competitive before bringing RDNA to laptops. They could be selling RDNA2 based APUs from 2020. Consoles are proof that they could. Except if there was a contract or something with MS and/or Sony where AMD had to delay RDNA2 in APUs.

 

[R0H1T]

If you have better explanation than "LOL, What? They are just stuck with a dumb strategy", please, be my guest.

My theory's there's probably a lot more margins in the handheld gaming space right now?

I do agree on Vega. They waited Intel to become somewhat competitive before bringing RDNA to laptops.

That's actually wrong, AMD APU's always lagged their GPU by at least 1-2 gens in the past. To get their best GPU uarch & CPU's to be under one SoC was a decent move but they could only pull it off after their massive success with server chips, that's when they started turning profits in like forever.

 

[Vayra86]

I would "LOL? What?" also if I was choosing to avoid reality. AMD is not a behemoth like Intel with it's own fabs. If Dell goes to Intel and asks if they can provide a million CPUs, Intel will reply "certainly". If Dell goes to AMD and asks AMD if they can provide a million CPUs, AMD will reply "Let us check with TSMC first. Also let us check if the allocation of wafers from TSMC for the specific product will be enough or we will need to reallocate more from another product line. And even then we don't know if it is in our interest to do so". Why do you think even in AMD's best years of Ryzen, big OEMs where insisting Intel? Why do they insist on 10nm Intel with limited P cores and questionable integrated graphics and worst battery efficiency? Because Intel can provide. On the other hand AMD seems to be providing every handheld manufacturer with Zen 4 and RDNA3 APUs, while big OEMs insist on older Zen 3 and Vega series. Why? Maybe because AMD can warranty huge quantities to OEMs of older APUs, but not the newest, meaning handheld makers with vast less needs get those easier.

If you have better explanation than "LOL, What? They are just stuck with a dumb strategy", please, be my guest.

I do agree on Vega. They waited Intel to become somewhat competitive before bringing RDNA to laptops. They could be selling RDNA2 based APUs from 2020. Consoles are proof that they could. Except if there was a contract or something with MS and/or Sony where AMD had to delay RDNA2 in APUs.

They can still do it, they just don't want to. If they align their partners better, they can get it done. Especially in consumer channels. Also, they're providing for Enterprise now, how the hell does that fit into the 'please dear TSMC' make me chips? Apparently there is enough certainty. I won't deny they can't cover the market like Intel does, but they certainly can arrange halo product lines and they damn well should.

Every single company can do proper expectation management, because that's what this is. It takes effort though, continuously.

 

[john_]

AMD has already shown it's hand. It is obvious that the APUs with RDNA are an absolute success. It seems like every week some Electronics manufacturer is releaseing a hand held based on AMD APUs. I have been talking for months about how powerful the Desktop variant will be. I predict that in 2024 we will get a family of Desktop APUs that will fill the void that is the Budget GPU space.

AMD is enjoying a "no competition" period. Intel still needs time to make GPUs that will be competitive and compatible, maybe they will get there with the help of TSMC's 3nm and Nvidia is totally ignoring the idea of an Nvidia gaming platform for laptops and now handhelds. And why offer an alternative to highly expensive and huge profit making discrete GPUs? I believe an ARM based Ada SOC running SteamOS could kill the Ryzen based handhelds, thanks to much higher efficiency. Well, as long as Nvidia is happy with Nintendo Switch, AMD can sleep peacefully at night.

 

[Vayra86]

AMD is enjoying a "no competition" period. Intel still needs time to make GPUs that will be competitive and compatible, maybe they will get there with the help of TSMC's 3nm and Nvidia is totally ignoring the idea of an Nvidia gaming platform for laptops and now handhelds. And why offer an alternative to highly expensive and huge profit making discrete GPUs? I believe an ARM based Ada SOC running SteamOS could kill the Ryzen based handhelds, thanks to much higher efficiency. Well, as long as Nvidia is happy with Nintendo Switch, AMD can sleep peacefully at night.

You might see one tomorrow... the issue here is that Nvidia has Tegra and not x86. They can't make a proper APU to save their lives unless it guzzles power. And they're not competitive if they bandaid some ARM cortex with their own GPU.

 

[Deleted member 232534]

LOL? What? Of course they can, they own the IP. Apple might have some tips and tricks to gain control over their value chain hm? Or Intel. Or Nvidia. Or... [insert pretty much any company].

This is a strategic choice, not impossibility. They can simply allow or disallow companies to release products with or without certain specs combined with their APUs. Intel created Ultrabook that way, for example. You set boundaries, you create product groups and force resellers to adhere.

AMD however is on a different form of logic, they think total freedom is the best way to get their product to better market share. It hasn't ever worked. But they keep at it, on some weird principle or something I don't know. Their GPU logic is the same wrt marketing. Its one WTF moment after another.

The only thing they have showing for that strategy is that they've always had revenue. Not profit, but revenue, certainly. Its also why after decades of pretty good products people still think of AMD as the lesser being of the 2 or 3 competitors. This is part of the reason why AMD is always competing on price even if they do have solid product. After all Zen isn't missing a featureset like GPUs do relative to Nvidia. It's missing its time to market, it dragged Vega along far too long, etc etc. All of this is timing and marketing strategy. And let's not even begin about the motherboard support fiasco that they keep choking on.

so they should dictate what? demand that company can not offer dgpu with their cpus? they can not do that. think about it for a second.

maybe can dictate that b650 motherboards have bios flashback. thats it.

 

[john_]

My theory's there's probably a lot more margins in the handheld gaming space right now?

Well, handheld consoles are smaller than a typical, no discrete GPU, cheap and much bigger laptop and they cost the same or more. So maybe there is? Don't know. Maybe the rest of the components are of higher quality, meaning a ROG Ally could have the same profit margins with a typical $700 no discrete GPU ASUS laptop. Just throwing speculations here.
Then again, ASUS and Lenovo wouldn't be showing any interest if there where slim margins, except if they see this market as an extra market, meaning, they expect the person who will buy an ASUS or a Lenovo laptop could also be a buyer of a hand held gaming console, not buy the console in place of the laptop.

That's actually wrong, Intel APU's always lagged their GPU by at least 1-2 gens in the past. To get their best GPU uarch & CPU's to be under one SoC was a decent move but they could only pull it off after their massive success with server chips, that's when they started turning profits in like forever.

If I am not mistaken, AMD brought RDNA2 to APUs after Intel started offering it's latest iGPUs, after Intel starting coming close in performance to AMD's iGPUs, at least in synthetic tests. In games I think it's still much behind.

 

[Vayra86]

so they should dictate what? demand that company can not offer dgpu with their cpus? they can not do that. think about it for a second.

maybe can dictate that b650 motherboards have bios flashback. thats it.

What? They can easily force partners to sell APUs and GPUs of different segments in their proper segments. They can provide motivation to do so. There are many ways to approach this issue.
With DIY of course not, but AIOs, laptops... anything prebuilt...

Here's how Intel approached this

Ultrabook - Wikipedia

en.wikipedia.org

 

[john_]

They can still do it, they just don't want to. If they align their partners better, they can get it done. Especially in consumer channels. Also, they're providing for Enterprise now, how the hell does that fit into the 'please dear TSMC' make me chips? Apparently there is enough certainty. I won't deny they can't cover the market like Intel does, but they certainly can arrange halo product lines and they damn well should.

Every single company can do proper expectation management, because that's what this is. It takes effort though, continuously.

What I have seen with AMD, but it could easily be a false and moronic idea of mine, is that AMD doesn't take much risks. Even when they had the best CPUs, their financials couldn't jump and take by surprice analysts, like what we have seen with Nvidia. I think AMD sets a reachable target and tries to accomplish it. So, if Dell that was buying for example 50 CPUs, comes and asks 200, I doubt AMD can offer them straight away. Looking at financials in the past, Intel's and Nvidia's could jump way off their own predictions, up or down. AMD's financials in the end of a quarter where more in line with what they where predicting in that quarter's start, with the only exception when pandemic and mining gone away, where they also had to announce 1 billion less income.
So going and dictating to big OEM how to build their own products, it's probably out of the question. To much a risk to get an answer "What? Deal is off. Buying from Intel".

 

[persondb]

"This setup is quite similar to Intel's design."

Its quite far apart actually. Intel uses completely different core architectures which can lead to instructions set differences between P and E cores. AMDs solution is the same core design with just cut down cache.

Which makes no difference in the end. The big.little setup is about different performance target and not different or same cores. The ISA is the same for e-cores and p-cores anyway, even if Intel does it by artificially disabling AVX-512

 

[dyonoctis]

LOL? What? Of course they can, they own the IP. Apple might have some tips and tricks to gain control over their value chain hm? Or Intel. Or Nvidia. Or... [insert pretty much any company].

This is a strategic choice, not impossibility. They can simply allow or disallow companies to release products with or without certain specs combined with their APUs. Intel created Ultrabook that way, for example. You set boundaries, you create product groups and force resellers to adhere.

AMD however is on a different form of logic, they think total freedom is the best way to get their product to better market share. It hasn't ever worked. But they keep at it, on some weird principle or something I don't know. Their GPU logic is the same wrt marketing. Its one WTF moment after another.

The only thing they have showing for that strategy is that they've always had revenue. Not profit, but revenue, certainly. Its also why after decades of pretty good products people still think of AMD as the lesser being of the 2 or 3 competitors. This is part of the reason why AMD is always competing on price even if they do have solid product. After all Zen isn't missing a featureset like GPUs do relative to Nvidia. It's missing its time to market, it dragged Vega along far too long, etc etc. All of this is timing and marketing strategy. And let's not even begin about the motherboard support fiasco that they keep choking on.

nvidia for exemple doesn't allow their partners to sell an external GPU bundled with a desktop graphics card.

Freedom is part of AMD brand image; their customers want them to be and stay the absolute antithesis of Intel/Nvidia...which is good on some aspect, but annoying when their partners like to cut cost/don't share the same goals: While AMD did make the displayport using a USB-C plug a requirement on AM5, that made brands like MSI/Asrock not offering a displayport at all on some motherboards. Because they still want to offer that as a premium. AMD making the port mandatory would have probably make them groan

 

[AusWolf]

So that's basically a 6500 XT shoved into a mobile APU. I wonder how they managed to keep the TDP at 45 W. Is it clocked at 600 MHz or something?

 

[Minus Infinity]

It's not necessarily due to these 3 doing nothing, it's just that new nodes are almost exponentially hard to turn to profits than even 5 years back! You have fixed costs that need to be recouped & then the prices of memory/PCB & other components just keeps on rising. There's also the fact that if you put 64bit wide bus on them & try to sell no one will probably buy them calling you cheapskates & if you go any higher you overshoot your budget. You also have to remember these things are competing with Apple these days & they have 1024(?) bit LPDDR5x memory at the top end. Everyone wants cheap GPU's for anaemic loss making prices but hey that's not how capitalism works! The simple fact is there's simply not much volume there as it was a decade back when the PC market was still booming. And you know what 4nm TSMC costs are

In an ideal world we could keep shrinking the Si down to infinite levels, but again that's not how reality works!

And already the Strix and Sarlak specs have leaked and the bus widths are 128 bit and 256 bit respectively.

 

[Lew Zealand]

So that's basically a 6500 XT shoved into a mobile APU. I wonder how they managed to keep the TDP at 45 W. Is it clocked at 600 MHz or something?

It'll clock much higher than that.

The RX 6400 has 12 CUs and runs at 2300 MHz consuming 51W on a fatter "6nm" process. 33% more CUs gets you 68W before process improvements so I could see 2000+ MHz at "4nm" consuming 45W. Of course needing to share that power budget with the rest of the APU will cut that lower but 1800-2000 MHz could be reasonable and that's before increasing the TDP settings.

 

[AusWolf]

It'll clock much higher than that.

The RX 6400 has 12 CUs and runs at 2300 MHz consuming 51W on a fatter "6nm" process. 33% more CUs gets you 68W before process improvements so I could see 2000+ MHz at "4nm" consuming 45W. Of course needing to share that power budget with the rest of the APU will cut that lower but 1800-2000 MHz could be reasonable and that's before increasing the TDP settings.

That would be cool!

The next problem, though, is the system memory bandwidth. The 6400/6500 XT pair are already held back by the 64-bit GDDR6, not to mention regular DDR5.

And just a personal gripe: the naming. What the hell is going on with this "one generation, two different series" thing again? Or does it have to start the 8000 series because of the iGPU? I remember how confusing Ryzen 2000-5000 was/is with Zen 1+, 2 and 3 shoved into all over the place, and I don't miss it.

 

[Lew Zealand]

That would be cool!

The next problem, though, is the system memory bandwidth. The 6400/6500 XT pair are already held back by the 64-bit GDDR6, not to mention regular DDR5.

Yes, that's the real problem. It's not too bad for the 6400 as it has the same 768 cores clocked >12% higher than the 1050 Ti and 14% higher bandwidth, plus somewhat more efficient cores and manages considerably more performance than the 1050 Ti. So the memory there is OK, not great. Frankly having only 4GB is the real problem/limitation. But even with 128-bit LPDDR5 7500 (shared system memory), that's only 94 GB/sec, ~75% of the RX 6400's bandwidth for these 1024 cores. That's going to be enough-ish in some circumstances and wayyy not enough in others.

And just a personal gripe: the naming. What the hell is going on with this "one generation, two different series" thing again? Or does it have to start the 8000 series because of the iGPU? I remember how confusing Ryzen 2000-5000 was/is with Zen 1+, 2 and 3 shoved into all over the place, and I don't miss it.

I can't follow it. I have to look everything up every time and still sometimes get the CPU gen and iGPU gen wrong.

 

[JustBenching]

steam deck, rog ally, its changed something in the industry is my guess. i know a lot of people who love their Deck.

apu laptops can game really decently as well, no longer needed giant heavy dedicated gaming laptop unless you just want to play AAA titles of recent years

i think the market is trending towards more APU's.

APU laptops, that was the case a year ago. With the latest games launching the way they do, oh boy gaming on an apu laptop is... an adventire

 

[Kyan]

APU laptops, that was the case a year ago. With the latest games launching the way they do, oh boy gaming on an apu laptop is... an adventire

Valve is really dedicate to make them work as quickly as possible. As always, if you don't care about hype and are willing to wait/play other game for ~1 month, you really don't care at all, and with the number of game available these days, it's not a problem to wait while the dev fix the bad management schedule.

 

[Space Lynx]

Valve is really dedicating to make them work as quickly as possible. As always, if you don't care about hype and are willing to wait/play other game for ~1 month, you really don't care at all, and with the number of game available these days, it's not a problem to wait while the dev fix the bad management schedule.

agreed, i have so much backlog i don't mind waiting, in fact i always wait for patches and optimizations before playing any new game

 

[ViperXZ]

Wow slowly but surely this starts to rival console performance APUs as the GPU gets bigger and bigger. This approaches the Xbox Series S now and will shadow older consoles. This is great for efficient systems and especially hand held gaming which can only use APUs and have no power or room for additional discrete GPUs. Imagine a Steam Deck 2 with this, will have more than double, maybe even triple the performance as the current one. So with FPS limited to 60 it would have great battery life.

 

[AusWolf]

Wow slowly but surely this starts to rival console performance APUs as the GPU gets bigger and bigger. This approaches the Xbox Series S now and will shadow older consoles. This is great for efficient systems and especially hand held gaming which can only use APUs and have no power or room for additional discrete GPUs. Imagine a Steam Deck 2 with this, will have more than double, maybe even triple the performance as the current one. So with FPS limited to 60 it would have great battery life.

I doubt handhelds will get this, considering this is a 45 W APU.

 

[ViperXZ]

I doubt handhelds will get this, considering this is a 45 W APU.

Maybe not, but it's also possible to use this with lower wattage, as was always possible with Ryzen mobile.