Oh, this is a good start, Intel seems to be active in ironing out issues. I hope they continue with this effort, they will definitely appeal to me more with future products.

Legacy-ZAOh, this is a good start, Intel seems to be active in ironing out issues. I hope they continue with this effort, they will definitely appeal to me more on future products.

I mean if they ever really want the compete they kinda have to....my main problem is that they are so quite now, I want a roadmap regarding drivers, I want them to set targets per month or bi-monthly or so and then update with vids that they hit those targets or if they did not, why not etc.
They were reasonably transparant so far, (apart from only measuring themselves against Nvidia and not also AMD) so yeah, continue that damn it.

ZoneDymoI mean if they ever really want the compete they kinda have to....my main problem is that they are so quite now, I want a roadmap regarding drivers, I want them to set targets per month or bi-monthly or so and then update with vids that they hit those targets or if they did not, why not etc.
They were reasonably transparant so far, (apart from only measuring themselves against Nvidia and not also AMD) so yeah, continue that damn it.

It's fine to measure themselves against Nvidia, because Nvidia is the market leader.

AnotherReaderIt's fine to measure themselves against Nvidia, because Nvidia is the market leader.

im not saying its not fine to measure themselves against Nvidia, im saying its sketchy to ONLY measure themselves against Nvidia, because as every reviewer as pointed out already, if you include AMD, Intel's product's look less favorable.

ZoneDymoim not saying its not fine to measure themselves against Nvidia, im saying its sketchy to ONLY measure themselves against Nvidia, because as every reviewer as pointed out already, if you include AMD, Intel's product's look less favorable.

You're right about that. The A770 has the potential to be much better than the 6600 XT, but right now, they are too close.

Too bad none of the arc cards only required a 6 pin connector, instead of both 6 and 8, i would have bought one already.

u2konlineToo bad none of the arc cards only required a 6 pin connector, instead of both 6 and 8, i would have bought one already.

I would have bought one if it weren't for the terrible idle power consumption and performance in old games.

ZoneDymoim not saying its not fine to measure themselves against Nvidia, im saying its sketchy to ONLY measure themselves against Nvidia, because as every reviewer as pointed out already, if you include AMD, Intel's product's look less favorable.

Maybe it's just to hard for their marketing team to create performance comparison slides with more than two different brands.

AnotherReaderI would have bought one if it weren't for the terrible idle power consumption and performance in old games.

Performance in old games is still a bit of a questionmark, luckily some more videos are coming out on youtube of people who got their hands on an Arc card that test some variety of games (instead of the same roster of games that every damn reviewer tests).
One showed how GTA4 ran poorly by default but if you used DXVK to basically run GTA4 through Vulkan, it ran really well.

The power consumption also recently got an update but TPU has not released a test of it yet, I found on Tom's Hardware that it reduced the idle powerdraw of the A750 from about 35 watt to 15 watt so thats great, but weirdly it did nothing for the A770 so maybe that needs a driver update or so.

My main problem is that I cant even buy it anywhere seemingly and if I can, the price is just too high for what it is.

ZoneDymoPerformance in old games is still a bit of a questionmark, luckily some more videos are coming out on youtube of people who got their hands on an Arc card that test some variety of games (instead of the same roster of games that every damn reviewer tests).
One showed how GTA4 ran poorly by default but if you used DXVK to basically run GTA4 through Vulkan, it ran really well.

The power consumption also recently got an update but TPU has not released a vid about it yet, I found on Tom's Hardware that it reduced the idle powerdraw of the A750 from about 35 watt to 15 watt so thats great, but weirdly it did nothing for the A770 so maybe that needs a driver update or so.

My main problem is that I cant even buy it anywhere seemingly and if I can, the price is just too high for what it is.

That's encouraging. Here, it was available at a local brick and mortar store, but there's no stock now.

ZoneDymoPerformance in old games is still a bit of a questionmark, luckily some more videos are coming out on youtube of people who got their hands on an Arc card that test some variety of games (instead of the same roster of games that every damn reviewer tests).
One showed how GTA4 ran poorly by default but if you used DXVK to basically run GTA4 through Vulkan, it ran really well.

The power consumption also recently got an update but TPU has not released a test of it yet, I found on Tom's Hardware that it reduced the idle powerdraw of the A750 from about 35 watt to 15 watt so thats great, but weirdly it did nothing for the A770 so maybe that needs a driver update or so.

My main problem is that I cant even buy it anywhere seemingly and if I can, the price is just too high for what it is.

I want arc for same reason, but in my case use linux only and dxvk allow many games without troubles

:)

Legacy-ZAOh, this is a good start, Intel seems to be active in ironing out issues. I hope they continue with this effort, they will definitely appeal to me more with future products.

Is it? Considering the amount of issues listed as fixed in this driver vs issues that still exist one would hope that they start picking up the pace. Rather they need to pickup up the pace if they want to get stable drivers by the time next gen Intel cards are out.

evernessinceIs it? Considering the amount of issues listed as fixed in this driver vs issues that still exist one would hope that they start picking up the pace. Rather they need to pickup up the pace if they want to get stable drivers by the time next gen Intel cards are out.

Yeah, they are much much more stable then they were even a few weeks ago. This release isnt a representation (and seems to focus on games) of the amount of progress they have made.

Even thoiugh Intel may not be a new name to all of us, the new graphics dept is new to them. Good for them to get this far so quickly.

Solaris17Yeah, they are much much more stable then they were even a few weeks ago. This release isnt a representation (and seems to focus on games) of the amount of progress they have made.

Thank you for informing me, I was not aware. That's extremely encouraging for the GPU market.

Huh.


I'm just having a moment now that we have player 3 posting driver updates and changelogs. Times have finally changed.

JrRacinFanEven thoiugh Intel may not be a new name to all of us, the new graphics dept is new to them. Good for them to get this far so quickly.

iGPU and discrete GPU drivers aren't that different

The red spiritiGPU and discrete GPU drivers aren't that different

If I have to believe....I think it was Linus, they really are different, in that one tries to put as much work as possible on the cpu while the dedicated gpu has to pull everything towards itself

The red spiritiGPU and discrete GPU drivers aren't that different

they really are, since this is an entirely different segment and the onboard graphics had almost none of the features of these new cards

ZoneDymoPerformance in old games is still a bit of a questionmark, luckily some more videos are coming out on youtube of people who got their hands on an Arc card that test some variety of games (instead of the same roster of games that every damn reviewer tests).

Most DirectX 9 games will probably be light enough that it would theoretically be easy enough to emulate, but that shouldn't be the main concern. The big problem with emulating an API is bugs, glitches and consistency (lag, stutter, etc.). A reasonable comparison would be Wine/Proton on Linux, which may be "good enough" for some games, but it's really a lottery, with games crashing randomly, having random glitching and stutter, etc. Graphics APIs are very complex state machines, and eliminating all bugs when implementing one API on top of another is magnitudes more difficult, if not completely impossible.

The only good solution is to pressure Intel into implementing DirectX 9 support in their drivers. They have plenty of resources and it's the only way to ensure proper support.

And while Intel have been claiming that "more and more" games are being ported to DirectX 12, that's only going to be a very tiny selection of games. What most of you don't know is that there is very big difference in DirectX 10+ vs. 9 and earlier. It's actually very easy to convert a DirectX 1-8 game to DirectX 9 (probably a few hours of work in most cases), but requires a rewrite to convert it to DirectX 12, which is why we see so many games from the late 90s and early 2000s being released as updated DirectX 9 games, even in recent years. And most of us probably have 50+ DirectX 9 games in our Steam or GOG collections, so DirectX 9 isn't going away anytime soon.

The red spiritiGPU and discrete GPU drivers aren't that different

Correct, the differences are on the hardware level, at least the significant ones. This is why AMD have the same driver for APUs and dedicated graphics cards.
The job of the graphics driver is to translate the graphics APIs (DirectX/Vulkan/OpenGL) into the GPU's native API. And contrary to popular belief, the GPUs themselves doesn't not execute DirectX or Vulkan API calls directly.

ZoneDymoIf I have to believe....I think it was Linus, they really are different, in that one tries to put as much work as possible on the cpu while the dedicated gpu has to pull everything towards itself

That's a complete misunderstanding. And it wouldn't surprise me if Linus S. said that, as he have made similar incorrect claims about dedicated network and audio cards.

Firstly, whether something is dedicated or integrated just refers to the processing hardware being on an add-in card or integrated into the motherboard or the CPU itself, it does not refer to how it's operating. Take the two big ones in the PC space; Intel and AMD. Their GPUs work pretty much the same as their dedicated ones, except for the memory controller which is shared with the CPU. It is also possible to have some shared resources or extra interconnects. Some server GPS like the Aspeed AST2500 have it's own embedded memory and isn't tied to the CPU at all (it even have an embedded ARM CPU to do its other tasks), but it's still regarded as integrated graphics because it's integrated into the motherboard.

As mentioned, the relevant integrated graphics from Intel and AMD work pretty much like their dedicated counterparts. Their architectures (except for memory) are the same, and they don't execute some of their instructions on the CPU.
But there is one significant difference between GPUs with dedicated or shared memory though; sharing memory can lead to both inefficiency and latency. CPUs and GPUs accesses memory in very different patterns; CPUs do small accesses which are very latency sensitive, and even tiny latencies will cause large slowdowns to program execution. Meanwhile, GPUs do accesses in large batches where throughput(bandwidth) is much more important than latency (which is the reason why most larger GPUs use GDDR which are bandwidth optimized). This means however, that an integrated GPU/CPU with shared memory can cause slowdowns for each other, even when the GPU load isn't that big, which is the reason why such integrated GPUs are known for causing stutter, even outside gaming. This issue has become less noticeable in the past ~10 years with faster CPUs, but is still noticeable to those of us who are very sensitive to stutter and latency.

The red spiritiGPU and discrete GPU drivers aren't that different

That was what Intel though and found out not to be true at all

ZoneDymoIf I have to believe....I think it was Linus, they really are different, in that one tries to put as much work as possible on the cpu while the dedicated gpu has to pull everything towards itself

Still, various Intel HDs were the most popular "gaming" GPUs on Steam. I'm not convinced that Intel is new to gaming and didn't have tons of experience beforehand. Despite that, even before Xes and Arcs, Intel botched their drivers often. Also Intel had tons of "performance" iGPUs too like whole Iris line-up. Not to mention that they somehow suck the most in older games, where they already spent so much time writing drivers for. Smells like Arc perhaps wasn't meant for gaming, instead to work like compute accelerator or something similar or Intel doesn't give too much damn about them.

The red spiritStill, various Intel HDs were the most popular "gaming" GPUs on Steam. I'm not convinced that Intel is new to gaming and didn't have tons of experience beforehand. Despite that, even before Xes and Arcs, Intel botched their drivers often. Also Intel had tons of "performance" iGPUs too like whole Iris line-up. Not to mention that they somehow suck the most in older games, where they already spent so much time writing drivers for. Smells like Arc perhaps wasn't meant for gaming, instead to work like compute accelerator or something similar or Intel doesn't give too much damn about them.

You are approaching to the core problem that most people don't get in conversations about Intel Arc;
The fact that Arc performs pretty well in synthetics and compute workloads but comparatively poor in gaming workloads is an indication of a hardware issue of resource balancing (namely scheduling). This is very analogous to AMD's issues with Polaris and Vega, but just a bit worse. And in both cases people keep blaming the drivers and thinking this will work itself out, but it wouldn't. (Keep in mind that we should have expected A770 to perform in the ~RTX 3080 range instead of between RTX 3060 and RTX 3060 Ti)

Secondly, it's important to understand that the bugs are unrelated to the GPUs underperforming. Most of the bugs are related to added features and gimmicks such as new AA modes, upscaling and other auxiliary gaming related features. The bugs are results from sloppy work and can be fixed, but the performance issues will remain.

Thirdly, as you mentioned, Intel have long experience with graphics drivers. And those thinking we should cut them some slack because they're "new" to dedicated graphics is just wrong. The drivers work pretty much the same regardless of the GPU being a tiny iGPU or a massive beast. And keep in mind, if driver overhead was the real issue, we should see a gradually increasing overhead with the higher GPUs. Also, most people have forgotten Intel's "prototype" DG1 graphics card from 2020, so they are not new to this. Intel also have a long record of better API compliance than AMD. So we should stop making excuses for Intel and regard the products as they truly are; flawed by design.

efikkanYou are approaching to the core problem that most people don't get in conversations about Intel Arc;
The fact that Arc performs pretty well in synthetics and compute workloads but comparatively poor in gaming workloads is an indication of a hardware issue of resource balancing (namely scheduling). This is very analogous to AMD's issues with Polaris and Vega, but just a bit worse. And in both cases people keep blaming the drivers and thinking this will work itself out, but it wouldn't. (Keep in mind that we should have expected A770 to perform in the ~RTX 3080 range instead of between RTX 3060 and RTX 3060 Ti)

Secondly, it's important to understand that the bugs are unrelated to the GPUs underperforming. Most of the bugs are related to added features and gimmicks such as new AA modes, upscaling and other auxiliary gaming related features. The bugs are results from sloppy work and can be fixed, but the performance issues will remain.

Thirdly, as you mentioned, Intel have long experience with graphics drivers. And those thinking we should cut them some slack because they're "new" to dedicated graphics is just wrong. The drivers work pretty much the same regardless of the GPU being a tiny iGPU or a massive beast. And keep in mind, if driver overhead was the real issue, we should see a gradually increasing overhead with the higher GPUs. Also, most people have forgotten Intel's "prototype" DG1 graphics card from 2020, so they are not new to this. Intel also have a long record of better API compliance than AMD. So we should stop making excuses for Intel and regard the products as they truly are; flawed by design.

Coming from the horse's mouth directly (relevant time stamp 2:31 and also 4:50 ?t=151s ?t=290 )


And from the reviews here on the A750/A770 and looking at the performance graphs seems pretty in line with what Raja Koduri was describing, once you move the bottlenect to the gpu it's able to stretch it's legs more

Once you increase the gaming resolution to 1440p, the Arc A750 gains on its competitors. It's now 13% faster than RX 6600, 4% faster than RTX 3060, and almost matching the RX 6600 XT. It seems that Intel A700 Series GPUs scale better with resolution than the other cards in our test group.

efikkanThis is very analogous to AMD's issues with Polaris and Vega

And the common problem has a common origin - Raja Koduri ;)