Strix Halo's 40 CU GPU with enough power is way faster than Panther Lake's 12 core GPU. But it's RDNA3.5, so it's basically worthless at ray tracing, whereas Xe3 as an iGPU can actually play ray-traced games with okay settings at a playable frame rate.

TheinsanegamerNYes, it is. The latency from FG on multiplayer makes the game an absolute mess to play.
[...]

Never said it was forced on you. I dont know where people get the confidence to build arguments on things people didnt say. Reeding comprehension r hrd.

If it's not forced on you, then your argument falls apart. FG is not mandatory, hence the perceived latency is not "a huge problem" in BF6. If you choose to use thsi optional feature then any resulting issues are on you, not the game, Nvidia, or anybody else, and you have no right to complain about it.

This is the reason I brought it up. Somebody wheeling out the ol' "reading comprehension" trope should have enough general comprehension to be able to grasp this.

Looks like it is more or less at the RX 6600's performance level in some games, which should place it somewhere around GTX 1080 Ti performance level, at least for the newer games. Nvidia's mature and optimized DX11 driver should on paper still make the Pascal cards run those games better. And this is on shared memory. Remarkable performance improvements from Intel, IMHO. Let's just hope they don't lose momentum, even if they cannot deliver B770, I wanna see a C770 on gamers' rigs.

From what I recall Panther Lake uses quad channel memory and a tiled architecture, Strix Point is monolithic and dual channel, so no wonder that Panther Lake ends up much faster. It's Intels equivalent of Strix Halo, just cheaper because they didn't splurge on a full 40CU GPU tile.

And having 2x frames because it can do 4x frame gen is really not that interesting. It means they have better software stack, not better hardware.

It's marketing slides what did you guys expect? AMD and nvidia did similar things as well so nothing out of ordinary here. Nobody in their right mind would play CP2077 at those setting with those GPU.

I'm glad Intel did well in iGPU department, and their tile design isn't as bad as I thought. I bought Meteor Lake MSI Claw A1M recently, despite being 3 years old now, it was very well done to be honest, the iGPU driver works well today, performance in game is great and there isn't any graphical glitches or performance problems like earlier reported, their combination P-cores, E-cores and LP E-cores works well in Windows 11, when I disable some of the cores it just didn't perform as well when I just leave it as it is, I assume their Thread Director works here. And with boost disabled it can play old or low power games or emulator at around 9-15W of power so it can be efficient as well.

hsewAMD just refeshed Strix Point with Gorgon Point (CPU clock speed improvement, same 3yrs old GPU),

Mighty Intels iGPU beats 3 year old iGPU.
So brave.

hsew... and there really aren't much if any 8040S devices on the market.

There may or may not be any 8040S.
But there are 8050S and 8060S. Which are more powerful.

Ryzen AI Max+ may or may not be more expensive than X9 388H. But is more powerful. Both CPU and GPU wise.

New Xe3 iGPU is most powerful iGPU.
But only in its own league.
Between cheaper Strix Point and more powerful Strix Halo.

cinemawareNo, it isn't. Nobody forces you to switch FG on in a competitive shooter. And if you play SP, it's just fine if your base fps is sufficient.

Yes it is, Frame Gen sucks!! And who said anything about forcing it on....?

ymdhisFrom what I recall Panther Lake uses quad channel memory and a tiled architecture ...

Only 2 channels
DDR5 can address each 32 bit part of channel separately as sub-channels.
But its only 128 bit wide memory access in total. (Strix Halo uses 256 bit and that is why it can not be put into AM5)
And its DDR5 thing not Panther Lake thing. Even 12th gen Alder Lake can do that.

www.intel.com/content/www/us/en/products/sku/245526/intel-core-ultra-x9-processor-388h-18m-cache-up-to-5-10-ghz/specifications.html

melkor_unlimitedMighty Intels iGPU beats 3 year old iGPU.
So brave.


There may or may not be any 8040S.
But there are 8050S and 8060S. Which are more powerful.

Ryzen AI Max+ may or may not be more expensive than X9 388H. But is more powerful. Both CPU and GPU wise.

New Xe3 iGPU is most powerful iGPU.
But only in its own league.
Between cheaper Strix Point and more powerful Strix Halo.

This product clearly isn't targeting Strix Halo, it's a regular laptop processor with a standard mobile TDP of just about 45W. Heck my 5600H uses more power than this. You said it yourself, it's competing with an older product, which means AMD will either lower their prices or release a replacement for them. There's no RDNA 4 iGPU right now, so that's a golden chance to release such a chip.

LionheartYes it is, Frame Gen sucks!! And who said anything about forcing it on....?

If you can't already comprehend how inane your argument is from my previous post, then I suppose no amount of further explaining will help.

But to make it somehow even simpler (lol): if BF6 required FG to work then maybe you'd have a point. But guess what, it doesn't! If it "sucks" then just don't use it. It’s optional! No seriously, go into the game's settings and have a look, you will see it's a toggle you can flip. Poof! "the huge problem" is now solved.

Really, it's as daft as If I'd say that Epic preset is "a huge problem" in this game because it drags my framerate down and what can we do about it?

There, I tried. But I bet y'all still don't get it ;)

The CarnageThe graph min is on top and max on bottom

What kind of dumbass graphing method is this? Who the F does graphs top to bottom?!?!?!?!!!?

melkor_unlimitedOnly 2 channels
DDR5 can address each 32 bit part of channel separately as sub-channels.
But its only 128 bit wide memory access in total. (Strix Halo uses 256 bit and that is why it can not be put into AM5)
And its DDR5 thing not Panther Lake thing. Even 12th gen Alder Lake can do that.

www.intel.com/content/www/us/en/products/sku/245526/intel-core-ultra-x9-processor-388h-18m-cache-up-to-5-10-ghz/specifications.html

oh okay, I misinterpreted the HWInfo summary. It said quad channel there, while on my machine it just says dual channel.

SquaredStrix Halo's 40 CU GPU with enough power is way faster than Panther Lake's 12 core GPU. But it's RDNA3.5, so it's basically worthless at ray tracing, whereas Xe3 as an iGPU can actually play ray-traced games with okay settings at a playable frame rate.

This is the biggest problem I have with Strix Halo. It's using RDNA3.5 instead of RDNA4, which is much better in every way.
Then the price is also high, but that was kinda expected.

On the other hand I have to applaud Intel for the IGP, because they made a significant improvement there.
Of course we need to wait for official reviews, but from Notebookcheck.net article the improvement is very good. Now the question is the driver quality and price of Panther Lake systems.

THANATOSThis is the biggest problem I have with Strix Halo. It's using RDNA3.5 instead of RDNA4, which is much better in every way.

From what little I've read, 3.5 and 4 both have some of the same changes applied to it, so they are equal in some ways. What exactly are you referring to when you say RDNA4 is better in every way? The fact that it's 0.5 versions ahead?

AsRockThink i will skip the 50% +\- fake frames thank you.

Find it odd that there was a push for cloud gaming and it failed then shit like FSR and DLS popped up.

Soon it be look cloud gaming is as good now and it will be all due to the delay. Lets face it you play a multiplayer game it can be 80ms +\- delay and FSR and DLLS.

This isn't odd, its the way commerce 'keeps innovating'.

The same movement was reinforced by the mainstreamification of 4K.
The reality is, almost no gamer is getting true 4K either, if its 1080p internal, you can rub your hands.

And in the meantime, the price gets inflated because look this is clearly better. Moar pixels! Moar latency! Moar moar moar! Similarly, RT has added latency to every frame, and for good reason is even ommitted entirely in games like BF6. The penny is dropping: its a gimmick after all... who woulda thunk it! But, other games are a 'cinematic experience' anyway, so this is fine, even if its still a straight regression from decades of established games. 'Innovation'. Who needs twitch gaming in anything but a shooter. We'll just auto regenerate your health anyway, and to make it all feel seamless, it takes 6-10 seconds to take something out of your inventory as it plays the animation for you opening your backpack. Such immersion, please hold E to use!

This is the reason UE5 relatively performs like horse manure and gaming graphics have stalled since 2010. This shit was done 10-15 years ago, but money must be made. Its also the reason indies take off now. The baseline of graphics, in every art style, is just available and easy to run. Only triple A that makes a point out of crippling your hardware to keep the commercial clusterfuck maximized is a problem now.

But! Soon! You can just run it all off THEIR servers and they'll handle that performance for you, and its exactly like you say; you won't notice the difference and the slop doesn't care if it gets played at 120 fps or 15, it needs zero real input anyway. Your gaming experience is primarily what items you can buy from whichever store in whatever phase of your progression in an already paid for product. Want some Prime with that, or a few coupons or a bunch of virtual currency to be spent on shit you never needed? Aren't you having fun!

I'm distancing myself from this as much as possible.

cinemawareIf you can't already comprehend how inane your argument is from my previous post, then I suppose no amount of further explaining will help.

But to make it somehow even simpler (lol): if BF6 required FG to work then maybe you'd have a point. But guess what, it doesn't! If it "sucks" then just don't use it. It’s optional! No seriously, go into the game's settings and have a look, you will see it's a toggle you can flip. Poof! "the huge problem" is now solved.

Really, it's as daft as If I'd say that Epic preset is "a huge problem" in this game because it drags my framerate down and what can we do about it?

There, I tried. But I bet y'all still don't get it ;)

Reading comprehension is really lost on you after all. Look at the article we're in.



This is 65 avg FPS with 2x frame gen
are you telling me you're going to play on cloud or competitively with half the FPS? 32? This will dip to 25 or lower. That's Fallout 4 vanilla on an underpowered console, an experience nobody including Bethesda thought was OK - I can mention a few dozen other shitshows to match this performance to drive the point home if you like, such as Ubisoft telling us PC needs 30 FPS for a 'cinematic experience'. That's the argumentation you're in bed with now. Enjoy. Even SP is a dreadful experience, these are averages - it'll feel like heavy Vsync on a console game. So, yea, 'required to work', but YMMV.

hsewIt's no secret that Strix Halo doesn't offer much if any advantage over a mobile 5060 in gaming, yet for some reason AMD insist on charging such a premium for it.

Part of the reason for this, is that Strix Halo incorporates a 16 core CPU. Almost nobody needs this shit.
Fortunately AMD has announced cut down variants with 12 (Max+ 392) and 8 (Max+ 388) cores, so hopefully this'll translate into more reasonable prices.

hsewAt the end of the day, people need to realize Intel has become a force to be reckoned with in the GPU arena, and that AMD has been slacking in that same department. No, they aren't as bad as Intel was after SandyBridge, but they've seriously underestimated Intel in this space and they're about to pay the price for it.

Hue. In the laptop space, yes, and to a non alarming extent, given Intel's continuous stronghold in that segment.
But outside of this niche, all facts and rumours point to Intel being already dead. Otherwise they wouldn't have needed to seal an IP partnership with nVidia.
(Still, if they release NVL-HX next year, they should give AMD a run for their money... before transitioning to nVidia's GPUs).

ymdhisFrom what little I've read, 3.5 and 4 both have some of the same changes applied to it, so they are equal in some ways. What exactly are you referring to when you say RDNA4 is better in every way? The fact that it's 0.5 versions ahead?

RDNA 4 completely crushes any prior version when it comes to ray tracing.

TheinsanegamerN:roll::laugh::roll:

The company with 0% dGPU marketshare is a "force to be reckoned with"? Really? You cant be serious!

So long as intel cant make a dent in sales and has fantastic 1-2 year driver support for their xe iGPUs, there is no way in hell that AMD will be paying any sort of price.

Yes. I'm 100% serious. dGPU isn't going to be around forever. It's inevitably going to become a niche and disappear from the mainstream, just like dedicated sound cards, NICs, and other coprocessors did. There's also the threat of Intel/AMD deciding to provide less PEG connectivity to expansions, like precious dGPUs! They will become a casualty of the stalwart pillar of technological advancement known as integration. nVidia knows this (they aren't stupid), which is why they're making moves to partner with Intel and/or Mediatek as, oh, what was that again, oh yeah, AN IGPU PROVIDER :eek::eek::eek: to maintain a presence in arguably the more important half of the consumer market. Do you honestly not see this trend? Or understand the consolidating nature of technology as a whole?

Marketshare isn't a forecaster either, merely an indicator of how things have gone so far. Especially in tech, there's always a new product or competitor just around the corner. Don't sleep on iGPUs, even from Intel.

Intel's iGPU improvements should push AMD to implement RDNA5 across iGPU stack with Zen6 products. It is not clear at the moment whether they would manage to tape it out on time, depending on whem they are ready to release those products. No clear leaks showing what's goigm to happen.

hsewYes. I'm 100% serious. dGPU isn't going to be around forever. It's inevitably going to become a niche and disappear from the mainstream, just like dedicated sound cards, NICs, and other coprocessors did. There's also the threat of Intel/AMD deciding to provide less PEG connectivity to expansions, like precious dGPUs! They will become a casualty of the stalwart pillar of technological advancement known as integration. nVidia knows this (they aren't stupid), which is why they're making moves to partner with Intel and/or Mediatek as, oh, what was that again, oh yeah, AN IGPU PROVIDER :eek::eek::eek: to maintain a presence in arguably the more important half of the consumer market. Do you honestly not see this trend? Or understand the consolidating nature of technology as a whole?

Marketshare isn't a forecaster either, merely an indicator of how things have gone so far. Especially in tech, there's always a new product or competitor just around the corner. Don't sleep on iGPUs, even from Intel.

It depends really on where dGPU is going altogether. Could go left or right. What I do also see is that there is a growing resistance against cloud everything, as people start to realize how little control they have left. One thing is a certainty: the continued demand for (gaming) graphics accelerators. It frankly doesn't matter how they get there, as long as they do and we don't lose anything in the process.

So far, despite lots of 'massive changes' to the GPU landscape... we can use a dGPU, and even IGPs, to get full backwards compatibility while also adding more and more capability for new graphics tech. It underlines the strength of the demand. No company in their right mind is going to kill legacy hard. Look at Nvidia; Physx out... physx in, as recently as Blackwell. People will not accept a straight up regression.

I don't think its a bad thing to get your GPU with your CPU chip, if the result is a solid experience. It will allow for really compact systems and that's really opening doors, with the Steam Deck and alternatives as an early example.

cinemawareIf it's not forced on you, then your argument falls apart. FG is not mandatory, hence the perceived latency is not "a huge problem" in BF6. If you choose to use thsi optional feature then any resulting issues are on you, not the game, Nvidia, or anybody else, and you have no right to complain about it.

This is the reason I brought it up. Somebody wheeling out the ol' "reading comprehension" trope should have enough general comprehension to be able to grasp this.

Sooo when a feature is optional....you cannot share why you would not want to use it?
We cant talk about how Tesla's Autopilot causes accidents because you can choose not to use it?

This is the same as saying you cant critizise a game or other program because it happens to be free....

Im sorry man, you make no sense whatsoever, you're entire stance on this is odd.

I'm so fucking tired of these numbers with upscaling and framegen, this makes it a million times harder to get a good read on the actual performance.

And don't even get me started on the first image where they decided it was a good idea to show RT prowess by showing that all configurations are overwhelmed by the chosen settings.

Yes, there's something nice in what Intel's releasing, but I need a PHD in this shit and with the amount of tomfoolery going around everyone will interpret the information in a different way.

Sigh, since I'm not in the market for a new iGPU anyways guess I'll just wait for 3rd party numbers. Perhaps there's finally a worthy upgrade in cpu perf from my 12800H in my work laptop.

ymdhisFrom what little I've read, 3.5 and 4 both have some of the same changes applied to it, so they are equal in some ways. What exactly are you referring to when you say RDNA4 is better in every way? The fact that it's 0.5 versions ahead?

RDNA 3.5 is RDNA 3 with the memory controller tweaked for LPDDR/DDR memory instead of GDDR and power optimization. They did some bug fix and change from the original RDNA 3 but it's mostly the same uArch. RDNA4 brought dedicated RT and AI cores unlike previous RDNA uArch that used the same shaders cores for all those tasks.

This is why not only RDNA 3* can only run the INT8 model, but also running it much slower (since it use the same resources as the shaders units). Same thing for RT.


Anyway, AMD took their damn time and is milking RDNA 3.5 as much as they could but Intel now being awaken is putting a end to that. I mean they still released RDNA 2 iGPU chips in 2026. I heard that Intel said that AMD was selling Old silicon and they aren't far from the truth.

i think iGPU on those parts is becoming more and more relevant. Not only those APU can run more and more game from the huge PC catalog, they can also be used in handhelds. I mean i got a AMD ryzen 7840U laptop to my niece and she is gaming all the time on it on Minecraft and Roblox and it do the job flawlessly.

I think it's great that Intel is competing on this front and making it a sales argument. That will just force AMD to stop milking old GPU uArch on the Mobile iGPU and everyone will benefits from it.

I still can't understand why so many people are upset when a contender take the lead in a sector. That is generally when each product evolve the fasted. If AMD didn't take the lead on the CPU side, Intel probably would still milk us with 4 core CPU.

What are the chances that these Intel vs AMD charts are actually done based on equal configurations? There's always something that is not mentioned. TDP, memory speed, upscaling, etc.

TPUniquePart of the reason for this, is that Strix Halo incorporates a 16 core CPU. Almost nobody needs this shit.

Some of those nobodies used it in notebook, miniPC and even handhelds.

TPUniqueFortunately AMD has announced cut down variants with 12 (Max+ 392) and 8 (Max+ 388) cores, so hopefully this'll translate into more reasonable prices.

I am afraid it will keep its price high.
Yes. You may spare some $ on that one computing chiplet, but overpay tenfold more for memory :/

TPUniqueRDNA 4 completely crushes any prior version when it comes to ray tracing.

More ray tracing? Yes.
But the main difference is that ML part. AMD counterparts of the fancy features from DLSS set (ML Upscaling, ML denoising, ML frame gen, ML whatever) done using FP4 math.

Mack4285What are the chances that these Intel vs AMD charts are actually done based on equal configurations? There's always something that is not mentioned. TDP, memory speed, upscaling, etc.

It may or may not be zero.
Or very close to zero.

People forget the price tag 890M systems you can get sub 1000 The 388H will be nowhere near that pricetag.