News Intel may have finally caught up to AMD in iGPU performance — Meteor Lake's Arc iGPU edges out AMD Phoenix's RDNA 3 iGPU in gaming and synthetic be...

[Admin]

Intel's Core Ultra Meteor Lake CPUs have seemingly dethroned AMD's Ryzen 7040 Phoenix APUs in integrated graphics.

Intel may have finally caught up to AMD in iGPU performance — Meteor Lake's Arc iGPU edges out AMD Phoenix's RDNA 3 iGPU in gaming and synthetic be... : Read more

 

[Bikki]

To say the Intel iGPU win AMD iGPU on those games is not correct. If you was following game benchmarking then you know LoL, Dota 2, Mount & Blade are one of those titles that are cpu bounded. Intel chip wining here is not surprising given raptor lake already had higher single thread cpu performance.

To be fair, Intel is finally getting real close to AMD in term of performance, but not surpassing them just yet.

 

[Roland Of Gilead]

I don't know. IMO Intel have done a good job here. As history tells us, Intel's iGPU have been rubbish over the years. We all know that. This time around the iGPU is more akin to AMD's APU's level iGPU. Certainly now close in performance. Arc release wasn't bad either. I think this bodes well for the general consumer, who will have better battery life, along with half decent gaming in a nice Thermal envelope (not currently). The only downside is the shared ram/vram. Still, it's a good show.

 

[abufrejoval]

AMD started a competition there, where they aren't well positioned to win any more, now that Intel has caught up.

When AMD split their chips into CCDs and IODs for a far more efficient way to produce a large range of desktop and server chips, that obviously caught Intel offguard.

But that flexibility was lost on the APU side, where AMD for the longest time and almost systematically today only has a single physical die.

Sure, producing different designs in different process sizes has given them some leeway, but they still basically don't have a response to Intel's far more fine-grained mix-and-match chips cookery. And they surely can't just multiply their engineering teams to match Intel's size.

Intel can now produce a vast variety of product across a giant range of performance, efficiency and price points, most of which won't actually be necessary, but bargain hunting consumers just revel in choices, little matter if they are meaningful or not. And it's worst case when Intel can claim the top performer in each variant.

AMD will only be able to hit a few of these spots and for lack of 50% performance gaps in any of those main categories of performance, efficiency or price), vendors will most likely go with an Intel design that allows them to cover the entire range. With so many (somewhat artificially inflated) choices to deal from Intel, where is the motivation and gain to deal with AMD offers? AMD just can't afford contra revenues like team Blue to bribe vendors.

Unless AMD has a Foveros 2 hidden in their labs, I am afraid that AMD's only choice is to go with design much like Apples M series, APUs that can be aggregated in S, M, L using 1/2/4 APU slices with fixed core, GPU, cache and RAM channel allocations. I'd still want one external RAM channel allowed for CPU RAM expansion, though, because I can rarely have enough RAM.

P.S. I am also worried that consoler makers are eying Intel with far more interest today...

 

[fla56]

I don't know. IMO Intel have done a good job here. As history tells us, Intel's iGPU have been rubbish over the years. We all know that. This time around the iGPU is more akin to AMD's APU's level iGPU. Certainly now close in performance. Arc release wasn't bad either. I think this bodes well for the general consumer, who will have better battery life, along with half decent gaming in a nice Thermal envelope (not currently). The only downside is the shared ram/vram. Still, it's a good show.

AMD is literally 30% more efficient at power levels that mater for laptops

Unless you are claiming people buy Intel iGPUs to game at 50W?

 

[fla56]

AMD started a competition there, where they aren't well positioned to win any more, now that Intel has caught up.

When AMD split their chips into CCDs and IODs for a far more efficient way to produce a large range of desktop and server chips, that obviously caught Intel offguard.

But that flexibility was lost on the APU side, where AMD for the longest time and almost systematically today only has a single physical die.

Sure, producing different designs in different process sizes has given them some leeway, but they still basically don't have a response to Intel's far more fine-grained mix-and-match chips cookery. And they surely can't just multiply their engineering teams to match Intel's size.

Intel can now produce a vast variety of product across a giant range of performance, efficiency and price points, most of which won't actually be necessary, but bargain hunting consumers just revel in choices, little matter if they are meaningful or not. And it's worst case when Intel can claim the top performer in each variant.

AMD will only be able to hit a few of these spots and for lack of 50% performance gaps in any of those main categories of performance, efficiency or price), vendors will most likely go with an Intel design that allows them to cover the entire range. With so many (somewhat artificially inflated) choices to deal from Intel, where is the motivation and gain to deal with AMD offers? AMD just can't afford contra revenues like team Blue to bribe vendors.

Unless AMD has a Foveros 2 hidden in their labs, I am afraid that AMD's only choice is to go with design much like Apples M series, APUs that can be aggregated in S, M, L using 1/2/4 APU slices with fixed core, GPU, cache and RAM channel allocations. I'd still want one external RAM channel allowed for CPU RAM expansion, though, because I can rarely have enough RAM.

P.S. I am also worried that consoler makers are eying Intel with far more interest today...

Is this post parody / comedy?

Which console maker will be buying this chip exactly?

Which gamers are buying Intel to game at 50W exactly?

 

[gilgamex101]

AMD started a competition there, where they aren't well positioned to win any more, now that Intel has caught up.

When AMD split their chips into CCDs and IODs for a far more efficient way to produce a large range of desktop and server chips, that obviously caught Intel offguard.

But that flexibility was lost on the APU side, where AMD for the longest time and almost systematically today only has a single physical die.

Sure, producing different designs in different process sizes has given them some leeway, but they still basically don't have a response to Intel's far more fine-grained mix-and-match chips cookery. And they surely can't just multiply their engineering teams to match Intel's size.

Intel can now produce a vast variety of product across a giant range of performance, efficiency and price points, most of which won't actually be necessary, but bargain hunting consumers just revel in choices, little matter if they are meaningful or not. And it's worst case when Intel can claim the top performer in each variant.

AMD will only be able to hit a few of these spots and for lack of 50% performance gaps in any of those main categories of performance, efficiency or price), vendors will most likely go with an Intel design that allows them to cover the entire range. With so many (somewhat artificially inflated) choices to deal from Intel, where is the motivation and gain to deal with AMD offers? AMD just can't afford contra revenues like team Blue to bribe vendors.

Unless AMD has a Foveros 2 hidden in their labs, I am afraid that AMD's only choice is to go with design much like Apples M series, APUs that can be aggregated in S, M, L using 1/2/4 APU slices with fixed core, GPU, cache and RAM channel allocations. I'd still want one external RAM channel allowed for CPU RAM expansion, though, because I can rarely have enough RAM.

P.S. I am also worried that consoler makers are eying Intel with far more interest today...

AMD is enormous nowadays compared to where it used to be at its peak down-and-out era of Late Phenom II' into the FX Chip line nightmare, poor Bulldozer...... In any case, they have tremendous market share and are in almost every console, if not AMD, then ARM. They have built in-roads, business relationships, trust and a very interdependent connection over support of all kinds. They are extremely competitive, and AMD's graphics division is no slouch whatsoever. While Nvidia is far and away the leader of all 3 in that department, there is a push towards more open source technologies, I feel. An inevitable trend towards open source especially in light of manufacturing process delays which have been a complete nightmare for Intel over the last couple of years. Delays after delays, hardly inspires confidence. They tried the low power notebook market, netbooks, tablets, you name it. They crashed out hard, not that they might not return one day. I just personally think proprietary hedgemons are going to find themselves with increasing stiff competition moving forward. RISC-V anyone?

 

[brandonjclark]

"In a review by Golden Pig Upgrade, the Core Ultra 7 155H could barely overtake AMD's Ryzen 7 7840HS in gaming and synthetic iGPU benchmarks."

To say, "could barely overtake" is poor phrasing. It implies that the Intel product had a position to defend when in fact the AMD product was the one defending the position.

A better editorial phrase choice would be, "was able to barely overtake".

In this sense, "was able to" implies that a new feat was accomplished.

 

[abufrejoval]

Is this post parody / comedy?

Which console maker will be buying this chip exactly?

Which gamers are buying Intel to game at 50W exactly?

AMD can currently offer pretty much cut & paste at chip design time--for the entire ASIC.
The ability to create bespoke designs for console makers with CPU, GPU, NPU, VPU and AI blocks of any size is attractive to console makers, who want to hit a very specific price/performance point for their products.

That was a lot of flexibility and it won Sony and Microsoft.

But what if you want several products? Or if you want to adapt to the market mid-generation?

With Foveros the cut & paste capabilities move a lot nearer to the assembly time, a much later integration point with lower hurdles for change. The barrier to variations is much lower.

It obviously all depends on the cost of Foveros vs. a single chip design, but that may not be a static cost, too many engineers at Intel are working on lowering that barrier for that.

Intel is betting the company on being able to create much more fine grained chip assemblies for a far wider range of products than AMD.

AMD is currently executing at the die carrier level and they do their V-cache stuff, but those are two rather specific use cases that do not cover the personal computing range very much.

They better have something up their sleve!

 

[GoldenBullet]

For igpu, the upscaling abilities is key for the best gameplay experience. Unless Intel has something as good as amd and Nvidia, amd is still the clear winner.

 

[D1v1n3D]

Intel's Core Ultra Meteor Lake CPUs have seemingly dethroned AMD's Ryzen 7040 Phoenix APUs in integrated graphics.

Intel may have finally caught up to AMD in iGPU performance — Meteor Lake's Arc iGPU edges out AMD Phoenix's RDNA 3 iGPU in gaming and synthetic be... : Read more

But there is a new Gen well half a Gen that is coming very very soon this will be a mild upgrade but when things get even smaller more kinds of combo chiplets designed to do specific tasks will fit under the same space the 5to 7nm chiplets lay then adding AMD 3d stacking tech Amd seems to be ready to launch some crazy stuff but why when what is out there now is outselling Intel all day long lately I bet there are some top secret chips in design and testing phases that may never make it to market for one reason or another. Like a lot of big engineer teams that break off into a few different teams have three different directions and the best one of the three makes it the other two for whatever reason don't see the light of day for now at least.

 

[GoldenBullet]

• Strix Point IGP is based on RDNA 3.5 architecture #
• 16 CUs in IGP #

Above is coming next year, so 25 percent more CUs.

"Strix Halo" is a mobile SoC with 16 Zen 5 cores, 40 CU GPU, 256-bit LPDDR5X memory, 120 W TDP

Above is coming in 2024 but looks like 2025 now.

Amd is grabbing the igpu market but the balls. Personally I want the strix halo, but a year is a year. I recently purchased 3 7840hs mini pcs, 2 for myself and one for dad.

https://www.techpowerup.com/review/future-hardware-releases/#ryzenphoenix2

 

[thestryker]

I didn't see the system they were using identified, but if it's anything like the pre-release Hardware Canucks got I'm not sure about the reliability of benchmarking right now. HC's was bad enough at performance consistency they called their numbers a preview. No matter what it's good to see Intel actually improving efficiency and IGP performance on mobile finally.

 

[palladin9479]

Eh that's more about comparing one laptop to another, we can discuss further if AMD ever releases another desktop APU for a real comparison. This ... reference ... is also missing a key detail, Intel's drivers. Intel has done amazing things to play catch up but they still aren't close to what nVidia / AMD (ATI) have in driver maturity.

 

[Bikki]

I was wondering why the iGPU portion of intel chip is always so petty. Intel may finally chose to up their iGPU game this generation and increase it's die size, in jealous of AMD success in portable market. The trade off is it's cpu core take a back seat, evident by the lack of core count increase on intel 4 (previously known as intel 7nm).

Intel alder lake/raptor lake m iGPU has just a fraction of the chip to work with

In contrast to AMD foenix

Just look at the montrous iGPU on apple M2 pro

 

[TerryLaze]

But there is a new Gen well half a Gen that is coming very very soon this will be a mild upgrade but when things get even smaller more kinds of combo chiplets designed to do specific tasks will fit under the same space the 5to 7nm chiplets lay then adding AMD 3d stacking tech Amd seems to be ready to launch some crazy stuff but why when what is out there now is outselling Intel all day long lately I bet there are some top secret chips in design and testing phases that may never make it to market for one reason or another. Like a lot of big engineer teams that break off into a few different teams have three different directions and the best one of the three makes it the other two for whatever reason don't see the light of day for now at least.

Article from november of this year...intel is outselling amd 8 to 2 in the mobile market, and any other market for that matter.
https://www.tomshardware.com/pc-com...s-market-share-in-laptops-pcs-and-server-cpus

Eh that's more about comparing one laptop to another, we can discuss further if AMD ever releases another desktop APU for a real comparison. This ... reference ... is also missing a key detail, Intel's drivers. Intel has done amazing things to play catch up but they still aren't close to what nVidia / AMD (ATI) have in driver maturity.

Eh, people used to promote amd for their "fine wine" driver/hardware situation...this is in the same boat, intel will keep improving their drivers just as they have until now and will unlock more of the hardwares potential.

 

[Roland Of Gilead]

Unless you are claiming people buy Intel iGPUs to game at 50W?

I'm not claiming anything. Only that the general consumer when purchasing these laptops that don't have an dGPU, are able to game lightly. It's always been a downside with the Intel HD iGPU's/laptops.

 

[Order 66]

I don't know. IMO Intel have done a good job here. As history tells us, Intel's iGPU have been rubbish over the years. We all know that. This time around the iGPU is more akin to AMD's APU's level iGPU. Certainly now close in performance. Arc release wasn't bad either. I think this bodes well for the general consumer, who will have better battery life, along with half decent gaming in a nice Thermal envelope (not currently). The only downside is the shared ram/vram. Still, it's a good show.

I had one of those intel CPUs that would bottleneck it’s own IGPU, so I never got to experience the full power of my rubbish intel IGPU. It was at 1366x768.

 

[-Fran-]

I have to say... This is a positive first look at Meteor Lake, but there's not been proper benchmarks from reputable sources to the point where you can scrutinize the numbers.

I'm trying my best to stay positive, because competition is good, but what I'm reading here is the same disservice the "press" gives AMD from time to time hyping the products up and then falling flat on their face (most recent: see RDNA3).

Let's wait and see until more notebooks and benchmarks are in the wild, since I've also seen proof of the opposite, where Meteor Lake loses badly.

Regards.

 

[palladin9479]

Eh, people used to promote amd for their "fine wine" driver/hardware situation...this is in the same boat, intel will keep improving their drivers just as they have until now and will unlock more of the hardwares potential.

ATI's drivers have always been behind nVidia, that didn't change when ATI become a subdivision of AMD. Intel devoted just enough resources to have their drivers accelerate general purpose 2D GUI interaction and media encoding / decoding and that's it. Intel GMA's were perfectly fine for Office / Web work and notoriously bad for "gaming", Intel didn't care since it primarily sold it's products to OEM's who then sold PC's to business's.

What's changed is that "Gaming" is a big thing now and building PC's is so easy that people see it as a rite of passage. Intel wanted some of that action but was decades behind in driver development. Hardware, while not easy, isn't really that hard for a company who already knows how to make complex highly parallel processors, so I expect Intel's hardware to be pretty competitive next generation. But gaming requires lots of driver support, especially with Vulkan / DX12 moving most of the rendering control out of the OS and closer to the metal. Intel's drivers are going to struggle for at least another three to four years as they build up experience and a solid codebase.

 

[TerryLaze]

Intel's drivers are going to struggle for at least another three to four years as they build up experience and a solid codebase.

So the iGPU of meteor lake going to be better than it is now in three to four years!
That's what I'm saying.
Unless intel decides to drop the GPU market but that would be extremely unlikely since that's going to be a huge cash cow for the foreseeable future.

 

[palladin9479]

So the iGPU of meteor lake going to be better than it is now in three to four years!
That's what I'm saying.
Unless intel decides to drop the GPU market but that would be extremely unlikely since that's going to be a huge cash cow for the foreseeable future.

Eh what are you comparing it to exactly? AMD hasn't released an APU in a few generations, I know because I haven't rebuilt my HTPC since their last APU release. AMD's APU's are just Radeon GPU's with cut down dies embeded into their CPUs.

 

[gruffi]

"Caught up" compared to a one year old chip? LOL. Who sees the mistake? If it's about iGPU and mobile gaming then Intel will have to compete with Dragon Range and Strix Point next year, not Phoenix.

 

[Quirkz]

Intel didn't get a win here. These chips are different performance classes - The AMD part is designed to run at lower wattages for battery powered handhelds. IT's sweet spot is ~20. The meteor lake is designed to run at higher at higher TDPs for things like laptops, not handhelds.