systemBuilder22 said:

Irrelevant. The 780m (Steam Deck, Z1 Extreme) is 18 months old (Jan 2023). Xe2 also beats the ATI 9600 but the news of that fact is just about as unimportant as this entire article ...

What matters is that the AMD Strix Point GPU is slated to go into The Asus Rog Ally 2, which virtually guarantees that its better ...

Click to expand...

But Strix point isn't out, so what are they meant to compare it to? Possibly the currently available hardware? Seems like these results aren't irrelevant, they are actually very relevant, as it lets us know how it performs to current hardware in the hands of people, like the Ally in my hands. I hope it's as good as this person claims. I want AMD and Intel to slog it out for efficient and powerful gaming devices.

J95qRP7 said:

But Strix point isn't out.

Click to expand...

Strix Point is out, it's Strix HALO that is not out yet. Strix HALO does a 4x on the memorybus (256-bits, 4-way interleave) so that GPU can achieve 4070-class performance.

The most suspicious benchmarks I've seen in a long time.

Well that sure sounds promising, but let's not forget that Intel Lunar Lake is on TSMCs N3B node whilst the Z1 extreme is on N4 and the Steam deck APU is on N7 (N6 for the OLED model).

So simply moving to the latest proces node should have AMD relatively close.

Very good news though, it feels like AMD is half assing it a bit. They can do with some competition and intels previous attempt was underwhelming.

We know the X3D CPUs are very successful
We know APUs scale well with memory bandwidth
We know AMDs infinity cache introduced with the RX 6000 series works great and would offer great bandwidth.

Now how about X3D chips for handheld gaming and have it tap in there for the infinity cache. It feels like AMD has had the cards in hand for ages to make amazing APUs that would obsolete all dedicated 1080 and even somer 1440p cards but then decides not to. Why try to compete with NVIDIA whilst they could just offer something that Nvidia cannot.

systemBuilder22 said:

Irrelevant. The 780m (Steam Deck, Z1 Extreme) is 18 months old (Jan 2023). Xe2 also beats the ATI 9600 but the news of that fact is just about as unimportant as this entire article ...

What matters is that the AMD Strix Point GPU is slated to go into The Asus Rog Ally 2 (Z2 Extreme APU), which virtually guarantees that its better ...

Click to expand...

Performance per watt is what matters for handhelds.

And the reason AMD lost the Switch 2 deal (also because FSR was too bad compared to DLSS).

Nvidia delivered better PPW and better upscaling and therefore got the deal for the second time. Nintendo was begging Nvidia back in 2015-2016 to use Tegra for Switch, because nothing else would deliver that kind of performance per watt back then.

Axeia said:

Well that sure sounds promising, but let's not forget that Intel Lunar Lake is on TSMCs N3B node whilst the Z1 extreme is on N4 and the Steam deck APU is on N7 (N6 for the OLED model).

So simply moving to the latest proces node should have AMD relatively close.

Very good news though, it feels like AMD is half assing it a bit. They can do with some competition and intels previous attempt was underwhelming.

We know the X3D CPUs are very successful
We know APUs scale well with memory bandwidth
We know AMDs infinity cache introduced with the RX 6000 series works great and would offer great bandwidth.

Now how about X3D chips for handheld gaming and have it tap in there for the infinity cache. It feels like AMD has had the cards in hand for ages to make amazing APUs that would obsolete all dedicated 1080 and even somer 1440p cards but then decides not to. Why try to compete with NVIDIA whilst they could just offer something that Nvidia cannot.

Click to expand...

While this is true, I'd say there's more considerations: AMD has had trouble keeping supply for Strix Point so far with reports of being very expensive for manufacturers to get a hold of, which is understandably why some people might even think it's not out yet (The other part is the naming convention of course)

Now there is a chance that AMD was intentionally conservative with the allocation of Strix Point in order to make room for Strix Halo but the most likely scenario is that it will also look very good just on paper and will be far too difficult to actually find on products.

The best chance is to hope that the big players like Asus and Lenovo actually are capturing most of the allocation giving raise to the current availability issues but it's not guaranteed: there's a significant chance they might themselves experience delays, shortages and be pushed into increasing prices if they go with Strix Point/Halo.

I think given all these there's more than just promise for Lunar Lake here, but only if intel can manage to keep up in the software side and that so far, hasn't been the case: testing of Lunar Lake I've seen seems wildly inconsistent like you would expect from a subpar driver optimization: some games are great with a clear win vs Halo Point, some are literally half the performance and some games just flat out aren't working on Lunar Lake yet.

Most gamers will probably be understandably cautious about intel still because of this, but given their battery life numbers I'm willing to give them a chance if I can find a decently priced 2 in 1 laptop which is what I'm in the market for right now: Already own a Legion Go and it's good enough for most of my gaming and windows tablet needs, just need a laptop daily driver that can replace my desktop rig, 13-14" class and has a reasonable keyboard.

Rapob said:

Performance per watt is what matters for handhelds.

And the reason AMD lost the Switch 2 deal (also because FSR was too bad compared to DLSS).

Nvidia delivered better PPW and better upscaling and therefore got the deal for the second time. Nintendo was begging Nvidia back in 2015-2016 to use Tegra for Switch, because nothing else would deliver that kind of performance per watt back then.

Click to expand...

Again, there's more considerations: Nvida is a lock for someone like Nintendo who's already a third party outsider with a tight knit closed garden ecosystem (Pirates notwithstanding) so they don't really need any kind of CPU strength for their devices, they just need performance per watt as you mention and Nvidia is very clearly ahead (Again, high end offerings of theirs notwithstanding)

However for basically everybody else but Nintendo (And maybe Apple) they can't just switch to Nvidia for their entire solution since Nvidia just doesn't develops any x86 CPUs. Remember that even Sony is x86 now so there's far too much trouble to try and convince devs to develop for anything else no matter how miraculously efficient the GPU part is, doing native ARM is either too costly to develop or too costly of a performance hit.

For this reason there's still a healthy market for x86 APUs and nobody is jumping at partnering with Nvidia for this since they know they'll get the worst end of the deal so, things like this will continue to matter.

A laptop is faster than a handheld? Oh, the Shock of it all!!!

Dimitriid said:

Again, there's more considerations: Nvida is a lock for someone like Nintendo who's already a third party outsider with a tight knit closed garden ecosystem (Pirates notwithstanding) so they don't really need any kind of CPU strength for their devices, they just need performance per watt as you mention and Nvidia is very clearly ahead (Again, high end offerings of theirs notwithstanding)

However for basically everybody else but Nintendo (And maybe Apple) they can't just switch to Nvidia for their entire solution since Nvidia just doesn't develops any x86 CPUs. Remember that even Sony is x86 now so there's far too much trouble to try and convince devs to develop for anything else no matter how miraculously efficient the GPU part is, doing native ARM is either too costly to develop or too costly of a performance hit.

For this reason there's still a healthy market for x86 APUs and nobody is jumping at partnering with Nvidia for this since they know they'll get the worst end of the deal so, things like this will continue to matter.

Click to expand...

I would not be surprised if PS6 will go ARM instead of x86.

Im waiting for strix point halo for handhelds that'll be a bigger upgrade

Dimitriid said:

While this is true, I'd say there's more considerations: AMD has had trouble keeping supply for Strix Point so far with reports of being very expensive for manufacturers to get a hold of, which is understandably why some people might even think it's not out yet (The other part is the naming convention of course)

Click to expand...

This is untrue and is spread by Intel to discredit AMD. The honest fact is, AMD gave an exclusive to Asus until October on the Strix Point APUs, and it was easy to get one of the ~9 models of Strix Point APU laptops from Asus (I know I own a Strix Point Zenbook S16 with hx365 CPU). AMD gave the exclusive to practice their launch ramping technique, and it worked splendidly.

Now that the exclusive is finished, Strix Point APUs are available from other makers (HP, and Lenovo very soon) at all price points, with no ordering delays, although it's clearly not a cheap APU because the laptops start at $1000, even on sale ...

Also, AMD really handicapped themselves by limiting bus speed to 7500 Mhz, whereas Intel and Qualcomm support 8533 Mhz which is about 14% faster. The 7500 Mhz is a FAIL when your 880m APU is already memorybus bottlenecked and your 890m APU achieves only a 9% speedup over the 880m with 33% more CUs because the APU is memorybus bottlenecked. So this week AMD increased the memorybus spec to 8000 Mhz, a 6.7% speedup.

Also, AMD had planned to put an infinity cache into their APU but the VLSI real estate was gobbled up by microsoft demanding a 40+ TOPS NPU (AMD delivered a 50 TOPS NPU). If the AMD APU had incorporated the infinity cache, I think it would have easily bested the Intel APU.