Intel Arc A750 Limited Edition GPU Review & Benchmarks vs. AMD RX 6600, A770, & More
video description
Date: 2022-10-09
Related videos
Comments and reviews: 15
Dustin
I am quite unsurprised by Intels flubbing of the drivers. It's like you can see their corporate mismanagement made manifest in their software. Dozens of different departments cutting each others throats and competing against each other, which ends up costing the company as a whole and leading to an inferior product. It's like Microsoft with Steve Ballmer at the helm. He thought a company would do well if the employees were competing against each other and the departments competing against each other... and he almost drove them bankrupt. I don't think Intel is that explicitly and intentionally wrongheaded, they're more in the position where they are a huge company used to being able to ignore product development and just let it crank along at an easy pace without any competitor really pushing them. They'd set boring 5-year roadmaps of incremental improvements and just pop them out pretty reliably. Then when AMD came out with Ryzen, they paniced. When their management forced engineering to rush, the one thing you should never, under any circumstances, rush, the very predictable (except to management types who don't understand engineering at all) happened - it failed. 10nm couldn't get the yields needed, and they wasted time they didn't have. Then management and marketing and all the rest of the company was breathing down the necks of engineering again demanding they rush some other shit to market like junkies who would go into withdrawal and that gave us the endless parade of 14nm jokes. When Gelsinger took over, there seemed to be hope, but he's proven the job is too much for him. I really don't think it's his fault, he's very smart, but he's a very smart guy trying to work in an environment of management wonks who see intellectual ability as a handicap rather than an asset. To them, the engineers are the plebian workers and the least important part of the company. So he really has a den of vipers surrounding him at the C-level who have zero faith in him. Lisa Su becoming CEO of AMD was a last-ditch move of desperation, at best, as far as management was concerned. MBA management types had run AMD into the dirt for over a decade with their incompetence, undercutting their engineers at every turn (which is just what MBAs are trained to do on purpose but which is lethal for companies that live or die based solely on the performance of their engineers, something they were unwilling to admit), and Dr. Su was only able to get the job after the MBAs had essentially lost hope. They left AMD for dead and an engineer was finally given the opportunity to direct the company in the right way. She took a weekend and mastered business, like any PhD engineer could (it is a fantastically simple field) and didn't have other execs in her way. Gelsinger isn't so lucky. Intel will need another 7 or 8 years of losing probably before they shed enough dead weight from management to create a culture willing to value their front-line engineers, even if it means doing the unspeakable - increasing the price of labor. It's either that or bankruptcy. That's a bitter pill for MBAs to swallow who have been taught that a company losing money is preferable to one that pays a cent more than market rate or empowers their workers in any way.
reply
I am quite unsurprised by Intels flubbing of the drivers. It's like you can see their corporate mismanagement made manifest in their software. Dozens of different departments cutting each others throats and competing against each other, which ends up costing the company as a whole and leading to an inferior product. It's like Microsoft with Steve Ballmer at the helm. He thought a company would do well if the employees were competing against each other and the departments competing against each other... and he almost drove them bankrupt. I don't think Intel is that explicitly and intentionally wrongheaded, they're more in the position where they are a huge company used to being able to ignore product development and just let it crank along at an easy pace without any competitor really pushing them. They'd set boring 5-year roadmaps of incremental improvements and just pop them out pretty reliably. Then when AMD came out with Ryzen, they paniced. When their management forced engineering to rush, the one thing you should never, under any circumstances, rush, the very predictable (except to management types who don't understand engineering at all) happened - it failed. 10nm couldn't get the yields needed, and they wasted time they didn't have. Then management and marketing and all the rest of the company was breathing down the necks of engineering again demanding they rush some other shit to market like junkies who would go into withdrawal and that gave us the endless parade of 14nm jokes. When Gelsinger took over, there seemed to be hope, but he's proven the job is too much for him. I really don't think it's his fault, he's very smart, but he's a very smart guy trying to work in an environment of management wonks who see intellectual ability as a handicap rather than an asset. To them, the engineers are the plebian workers and the least important part of the company. So he really has a den of vipers surrounding him at the C-level who have zero faith in him. Lisa Su becoming CEO of AMD was a last-ditch move of desperation, at best, as far as management was concerned. MBA management types had run AMD into the dirt for over a decade with their incompetence, undercutting their engineers at every turn (which is just what MBAs are trained to do on purpose but which is lethal for companies that live or die based solely on the performance of their engineers, something they were unwilling to admit), and Dr. Su was only able to get the job after the MBAs had essentially lost hope. They left AMD for dead and an engineer was finally given the opportunity to direct the company in the right way. She took a weekend and mastered business, like any PhD engineer could (it is a fantastically simple field) and didn't have other execs in her way. Gelsinger isn't so lucky. Intel will need another 7 or 8 years of losing probably before they shed enough dead weight from management to create a culture willing to value their front-line engineers, even if it means doing the unspeakable - increasing the price of labor. It's either that or bankruptcy. That's a bitter pill for MBAs to swallow who have been taught that a company losing money is preferable to one that pays a cent more than market rate or empowers their workers in any way.
reply
Powernod
In NONE of your ARC reviews (A770 or A750) you have highlighted the major flaw that ARC has , which is ... half the performance for people with CPUs which don't support Re.siz. BAR. or simply , for people who haven't enabled that feature through bios.
I know you've mentioned that it in some older videos prior to the 2 reviews ,but when a customer searches for a video in order to buy something , guess what ... they will go and search for the review video which supposedly mentions all the positives and negatives .
It's somewhat comical for me to watch all those review numbers knowing that for a great percentage of users these numbers won't be even relevant.
--Once again , i would expect these kind of things from other reviewers , not from you .
( but unfortunately , for quite some time now i've noticed that certain things have changed . Thanks Steve !! )
reply
In NONE of your ARC reviews (A770 or A750) you have highlighted the major flaw that ARC has , which is ... half the performance for people with CPUs which don't support Re.siz. BAR. or simply , for people who haven't enabled that feature through bios.
I know you've mentioned that it in some older videos prior to the 2 reviews ,but when a customer searches for a video in order to buy something , guess what ... they will go and search for the review video which supposedly mentions all the positives and negatives .
It's somewhat comical for me to watch all those review numbers knowing that for a great percentage of users these numbers won't be even relevant.
--Once again , i would expect these kind of things from other reviewers , not from you .
( but unfortunately , for quite some time now i've noticed that certain things have changed . Thanks Steve !! )
reply
Radu
When it works it works really well, so that tells me it s all about drivers. You guys have to remember when NVIDIA hit the market or ATI(AMD) their backwards compatibility was terrible. We are talking like 20 years ago, when the first nvidia GPU came out it supported only the current dx version which was 7.0. You cannot expect Intel to do 20 years of backwards development to appease 1% of the population. We have to benchmark these cards in dx12 only to get a real idea of the performance of the silicon. I think they are on the right path and should continue, are they at nvidia levels, no amd is not even there yet, but hell at 290 they are putting on a good show. If they get a console deal with Microsoft for the Xbox series XI then we can really see some improvements
reply
When it works it works really well, so that tells me it s all about drivers. You guys have to remember when NVIDIA hit the market or ATI(AMD) their backwards compatibility was terrible. We are talking like 20 years ago, when the first nvidia GPU came out it supported only the current dx version which was 7.0. You cannot expect Intel to do 20 years of backwards development to appease 1% of the population. We have to benchmark these cards in dx12 only to get a real idea of the performance of the silicon. I think they are on the right path and should continue, are they at nvidia levels, no amd is not even there yet, but hell at 290 they are putting on a good show. If they get a console deal with Microsoft for the Xbox series XI then we can really see some improvements
reply
Penfolduk001
I wonder whether the overall better 1440p and 4K performance is down to the ARChitecture or because they've concentrated driver development at the higher resolutions?
Also, whilst I see what Steve is saying about Intel and AMD competing in the first instance, there's a danger they could cancel each other out and both exit the GPU market. We need someone(s) able to compete with nVidia to put their feet to the flames too.
When I got my gaming PC in 2020 I went with a RTX2060 Super. As nVidia was a no-brainer at that point.
If I'd waited 6 months it would have been a tougher choice.
When I next upgrade I hope the choice will be much, much harder.
reply
I wonder whether the overall better 1440p and 4K performance is down to the ARChitecture or because they've concentrated driver development at the higher resolutions?
Also, whilst I see what Steve is saying about Intel and AMD competing in the first instance, there's a danger they could cancel each other out and both exit the GPU market. We need someone(s) able to compete with nVidia to put their feet to the flames too.
When I got my gaming PC in 2020 I went with a RTX2060 Super. As nVidia was a no-brainer at that point.
If I'd waited 6 months it would have been a tougher choice.
When I next upgrade I hope the choice will be much, much harder.
reply
TheQuentincc
I think you should also bench a game for power consumption because furmark behave very differently from card to card, specially since every card (nowadays) on furmark is limited by it's power limit, for instance if you shunt mod a 3070 you could look at 300+W power consumption under furmark.
I guess a pretty game reprensentative benchmark to measure power consumption is Unigine Superposition, it's pretty recent and you can lock it on a specific scene, you can then measure the performance and power consumption on this specific scene to have a more accurante efficiency/power consumption graph.
reply
I think you should also bench a game for power consumption because furmark behave very differently from card to card, specially since every card (nowadays) on furmark is limited by it's power limit, for instance if you shunt mod a 3070 you could look at 300+W power consumption under furmark.
I guess a pretty game reprensentative benchmark to measure power consumption is Unigine Superposition, it's pretty recent and you can lock it on a specific scene, you can then measure the performance and power consumption on this specific scene to have a more accurante efficiency/power consumption graph.
reply
Crushonius
all you guys in the comments saying its a good start
get a grip none of you will even consider spending a dime on this garbled mess and you know it
it is a joke and it will fail as it should , intel has been in this business long enough to know
that you do not release something this bad and expect a win . as much as i would like there to be a third competitor
in the gpu space THIS IS NOT IT intel blew it and they know it too
this release is a hail mary attempt to hopefully keep their gpu division alive but the odds are horrendous
reply
all you guys in the comments saying its a good start
get a grip none of you will even consider spending a dime on this garbled mess and you know it
it is a joke and it will fail as it should , intel has been in this business long enough to know
that you do not release something this bad and expect a win . as much as i would like there to be a third competitor
in the gpu space THIS IS NOT IT intel blew it and they know it too
this release is a hail mary attempt to hopefully keep their gpu division alive but the odds are horrendous
reply
John
Considering the gains towards higher resolutions, it seems other reviewers were right about some issues with the scheduler. Intel cannot fully utilize and saturate its units under comparatively low loads, whereas with higher resolutions this problem solves itself as there's simply more data to spread amongst the units.
Igor's Lab claimed that Intel has a huge problem or a bug in the hardware scheduler. If he is right, newer driver versions may be able to improve the situation, but overall it cannot mitigate hardware bugs or design flaws.
reply
Considering the gains towards higher resolutions, it seems other reviewers were right about some issues with the scheduler. Intel cannot fully utilize and saturate its units under comparatively low loads, whereas with higher resolutions this problem solves itself as there's simply more data to spread amongst the units.
Igor's Lab claimed that Intel has a huge problem or a bug in the hardware scheduler. If he is right, newer driver versions may be able to improve the situation, but overall it cannot mitigate hardware bugs or design flaws.
reply
smartbutt10
I question the seemingly common insistence on using DX11 to test R6:Siege. To my knowledge, every card on the test suite right now supports Vulkan and would be a much more realistic representation of what people who play the game actually use (For reference, I'm currently on a collegiate siege team. All but one of our players use Vulkan exclusively, and the one that doesn't only does so because his older card doesn't play nice with it)
reply
I question the seemingly common insistence on using DX11 to test R6:Siege. To my knowledge, every card on the test suite right now supports Vulkan and would be a much more realistic representation of what people who play the game actually use (For reference, I'm currently on a collegiate siege team. All but one of our players use Vulkan exclusively, and the one that doesn't only does so because his older card doesn't play nice with it)
reply
Blart
For someone who plays middle-age games (not new, not old) on a GTX 1660, the Arc A750 is actually looking like a decent upgrade for a gamer who's just barely scraping 60fps on most 1080p titles. Even with intel's admittedly poor performance on older DX9/10/11 titles, all of them are still well above 60fps at even the highest resolutions. I'm not too opposed to the Arc and i'm pretty happy to see extra competition in the markets
reply
For someone who plays middle-age games (not new, not old) on a GTX 1660, the Arc A750 is actually looking like a decent upgrade for a gamer who's just barely scraping 60fps on most 1080p titles. Even with intel's admittedly poor performance on older DX9/10/11 titles, all of them are still well above 60fps at even the highest resolutions. I'm not too opposed to the Arc and i'm pretty happy to see extra competition in the markets
reply
TheGaluMoD
I said it on your A770 review and I'll say it again: I get that you guys are super busy right now, but I think releasing these videos as final reviews is not the right move, considering you haven't looked at ray-tracing performance at all. It's clearly an important feature that's gaining more traction each day, so ignoring it completely just isn't right. (Not shilling, defo not gonna get any of Intel's current jank cards)
reply
I said it on your A770 review and I'll say it again: I get that you guys are super busy right now, but I think releasing these videos as final reviews is not the right move, considering you haven't looked at ray-tracing performance at all. It's clearly an important feature that's gaining more traction each day, so ignoring it completely just isn't right. (Not shilling, defo not gonna get any of Intel's current jank cards)
reply
Never
I'm still excited and I have a 3090, 1440p is still the sweet spot (even though I have both 4k and 1440p monitors, gaming is much nicer near your face on 1440p and I just use 4k for timeline and production), and let's be honest the 750 could be brilliant for 1440p to price, with maturity and I'm willing to get an Intel gpu and watch it mature. Just am I willing to pay the extra for the A770 is more the question now
reply
I'm still excited and I have a 3090, 1440p is still the sweet spot (even though I have both 4k and 1440p monitors, gaming is much nicer near your face on 1440p and I just use 4k for timeline and production), and let's be honest the 750 could be brilliant for 1440p to price, with maturity and I'm willing to get an Intel gpu and watch it mature. Just am I willing to pay the extra for the A770 is more the question now
reply
Mr.
It's been said before, but this really reminds me of early Ryzen: still lots of jank, but it has promise. Games with Vulkan and ray-tracing are to Arc what 7-Zip was to Ryzen (no ray-tracing benchmarking in this review, but e.g. Digital Foundry did test it and it's about as good as Amperes, leaving AMD in the dust.) And Intel has already started throwing in free games to make the value proposition better.
reply
It's been said before, but this really reminds me of early Ryzen: still lots of jank, but it has promise. Games with Vulkan and ray-tracing are to Arc what 7-Zip was to Ryzen (no ray-tracing benchmarking in this review, but e.g. Digital Foundry did test it and it's about as good as Amperes, leaving AMD in the dust.) And Intel has already started throwing in free games to make the value proposition better.
reply
kiri101
I'd really like to see some benchmarking and comment on the hardware encoding performance - this has seemed like a potential differentiating feature and draw to the product for a while, now that the embargo is lifted it'd be great to get some hard numbers. How does it compare to software based encoding (CPU and/or GPU accelerated)? What is the performance per watt of energy? Performance per dollar?
reply
I'd really like to see some benchmarking and comment on the hardware encoding performance - this has seemed like a potential differentiating feature and draw to the product for a while, now that the embargo is lifted it'd be great to get some hard numbers. How does it compare to software based encoding (CPU and/or GPU accelerated)? What is the performance per watt of energy? Performance per dollar?
reply
nope
Eh, yeah. I guess I can deal with that for a little bit and hope that they keep pushing driver updates. isn't that the exact strategy that AMD has been using for their GPUs though? Why wouldn't it work for intel too?
Sounds like they made some hardware that's actually competitive for a change. Now it needs software. I'm sure they'll get there eventually. Unless they decide to ditch the whole thing.
reply
Eh, yeah. I guess I can deal with that for a little bit and hope that they keep pushing driver updates. isn't that the exact strategy that AMD has been using for their GPUs though? Why wouldn't it work for intel too?
Sounds like they made some hardware that's actually competitive for a change. Now it needs software. I'm sure they'll get there eventually. Unless they decide to ditch the whole thing.
reply
Ignorance
More and more looking like an investment. The hardware can perform. It really looks to be limited by the non existent history of Intel producing graphics drivers. If Intel can workout a quick turnover for optimizing lots and lots of games they have a winner. 1080p gaming performance works as a minimum but 1440p is quite noticeable for most PC gaming setups(4k much less so for most people).
reply
More and more looking like an investment. The hardware can perform. It really looks to be limited by the non existent history of Intel producing graphics drivers. If Intel can workout a quick turnover for optimizing lots and lots of games they have a winner. 1080p gaming performance works as a minimum but 1440p is quite noticeable for most PC gaming setups(4k much less so for most people).
reply
Add a review, comment
Other channel videos
