Intel Arc A380

[catulitechup]

It is for Linux users But seriously, it will be amazing on Linux. Intel drivers for Linux are very good. Furthermore, it does not matter that the (Windows) driver has poor support for DX11 and below because you will just use DXVK and Zink on Linux. GPGPU (OpenCL) support will probably be good on Linux too unlike with AMD Radeon. Aside from that it has AV1 decoding support and Intel GPUs are well-supported with VAAPI hardware video acceleration on Linux and now that hardware video acceleration works in browsers too finally on Linux, you will be able to watch YouTube in AV1 format without wasting electricity and hearing the fan in your laptop/desktop.

@W1zzard thanks for testing but as other said some linux tests will be very interesting

 

[Vario]

I think its hard to pull off a new release in a product category that is novel to them, so I will give them that. Maybe some of the technology will also filter into CPU integrated graphics.

 

[Lew Zealand]

Frankly the performance is staggeringly bad. $100 fine, but not at MSRP. We should be easily beating the 1660 Super in 2022.... everything slower is terrible, I'd buy used.

Right now the 1660 Super and RX 6600 are about the same price new, $253 vs $260, and the 6600 is at least 38% faster at 1080p, so that's a decent performance uplift. The real problem here is that nobody is making a less expensive video card with actually good performance. The bottom is $250. Even if the 3050 came down to MSRP, that'd still be $250.

Apparently that's the price floor for decent performance and *profits* nowadays. Maybe Intel's midrange GPUs can do something decent in the $200 range, we will see...

 

[beautyless]

"it only works on Intel and only works with resizable BAR". So when the card came in, I kept an open mind, plopped it into my Zen 3 Ryzen test system right away and, surprise, it worked perfectly fine. Well, not perfectly fine, because there are numerous layers of driver bugs to deal with, but the card performed as expected with Resizable BAR enabled on the AMD platform. No special tweaks, BIOS updates, or settings were necessary as I simply inserted the card, powered up the machine, and its BAR size was 6 GB—the exact same mechanics as with AMD and NVIDIA—it just works.

Resizable BAR is a must but this function on Intel GPUs is work on both AMD and Intel. Nice!

A380 looks under perform for it price. But it can be changed after a while.

 

[mechtech]

Well better than Larrabee

Not bad considering first kick at the can.

One more company using TSMC nodes....................

@W1zzard you should have tested this against the best Iris integrated

 

[Dirt Chip]

Nice start, good luck to the (very) early adopters out there...
The right thing to do is let intel do it`s beta testing with "Alchemist" and wait with conclusion about the ARC endeavor to the next gen, "battlemage".
If one is 'disappointed' about A380 or ARC because of this (excellent) review than please check your expectation meter and parameters as they are not align with realty.
This card is a beta GPU of a beta gen of the very first attempt in the big boys territory (let put Larrabee aside, please).

It is very encouraging to see that the beta driver solved all of the bugs.
Driver support and it`s maturation process is the important thing that Alchemist should be measured by, power and FPS are the side dish so to much debating about them is pretty null at this (very) early point in time.

 

[HD64G]

Exactly as expected. Bad drivers. And their top GPU (A770) will not even reach average performance of 6700XT/3060Ti me thinks. Next gen cannot arrive soon enough for Intel and they need (much?) time to fix their drivers.

 

[Dirt Chip]

Exactly as expected. Bad drivers. And their top GPU (A770) will not even reach average performance of 6700XT/3060Ti me thinks. Next gen cannot arrive soon enough for Intel and they need (much?) time to fix their drivers.

From the conclusion:
"I then tried Intel's newest 101.3220 Beta driver and it magically solved all serious issues. The card would now work and be 100% stable in all games, including Witcher, DoS 2 and RDR2—looks like the driver team is actually fixing bugs, good job! Just to emphasize, as soon as I switched to the beta drivers there was not a single crash during all my testing—a huge improvement."

So not so bad and obviously getting better quite fast.

also from the cons:

• "Immature drivers and software"

Immature is not so bad, it is just expected at this point in time.
So if we`ll see improvement at this (driver) front on a monthly manner I can call it better than expected.

 

[Fatalfury]

Being an Early Adopter and BETA TESTER for intel is not worth it ATM no matter how you see it.

will think on buying Intel GPU maybe in 2025 or atleast 3Gen or even when 5th Gen of Intel Cards comeout..

Until till..Good to see a new player..Let the new ship keep sailing further into deep waters.

 

[AusWolf]

Being an Early Adopter and BETA TESTER for intel is not worth it ATM no matter how you see it.

will think on buying Intel GPU maybe in 2025 or atleast 3Gen or even when 5th Gen of Intel Cards comeout..

Until till..Good to see a new player..Let the new ship keep sailing further into deep waters.

But we need some people to "beta test" it, right?

If everyone held out with buying an Arc GPU until 2025, then there would be no Arc GPU in 2025. It's not economically feasible to spend money on developing a product line that no one buys.

If you personally prefer a well tested product and don't want to risk having issues with something new, that's fine. But you can't expect everyone to be on the same opinion, unless you want Arc to fail miserably - which I don't see why you would, as you said it yourself that it's good to have a third player in the game.

So yes, "beta testing" is worth it.

As for me, I'll buy an A770 if it comes at a fairly decent price because I'm curious. I don't even care if it doesn't reach 3060 Ti heights.

 

[Dirt Chip]

But we need some people to "beta test" it, right?

If everyone held out with buying an Arc GPU until 2025, then there would be no Arc GPU in 2025. It's not economically feasible to spend money on developing a product line that no one buys.

If you personally prefer a well tested product and don't want to risk having issues with something new, that's fine. But you can't expect everyone to be on the same opinion, unless you want Arc to fail miserably - which I don't see why you would, as you said it yourself that it's good to have a third player in the game.

So yes, "beta testing" is worth it.

As for me, I'll buy an A770 if it comes at a fairly decent price because I'm curious. I don't even care if it doesn't reach 3060 Ti heights.

You shouldn't beta test just to help intel. You should beta test if you love the process of dealing with childbirth problems.

I advise everyone on this form NOT to buy A380 of A770 just because they "want to help the competition". It is futile argument as it can be.

Buy it if you are curious, want to "play" with it and love the feeling of holding something at its crud, un-refined form. Of course, be very well aware that you WILL held back by immature driver and need much patients to deal with it but- that is the fun and beauty of being an early adopter of totally new tech.

Arc is an uncharted territory at its best and you can see from the very (A380) beginning. It's DNA is different from GeForce and Radeon ,and that's a very good thing.

 

[ZoneDymo]

I'm still hoping RT dies some unnatural death & we get something much less taxing/efficient maybe with a hybrid RT+raster approach completely "revolutionizing" gaming.

Dont really get that mentality, RT has been the end game goal for like...3 decades now, we always been able to do it, just not in real time.
Now we are approaching it and you would rather have an in-between step?

thats like saying you hate electric cars and would rather have hybrids...

 

[AusWolf]

You shouldn't beta test just to help intel. You should beta test if you love the process of dealing with childbirth problems.

I advise everyone on this form NOT to buy A380 of A770 just because thay "want to help the competition". It is futile argument as it can be.

Buy it if you are curious, want to "play" with it and love the feeling of holding something at its crud, un-refind form. Afcures, be very vell aware that you WILL held back by immature driver and need much patients to deal with it but that is the fun and beauty of being an early adopter of totally new tech.

Arc is an uncharted territory at its best and you can see from the very (A380) begining. It's DNA is different from NV or AMD, and that's a very good thing.

Of course.

All I'm saying is that we need early adopters to let Intel iron out its issues for future generations, and make enough cash so that developing those future generations won't seem like a pointless idea.

Would I recommend everyone to buy an Arc GPU? Definitely not! You need a tech-savvy and adventurous mind to do venture out, a hundred percent. But you can't say that it isn't worth it for anyone.

Dont really get that mentality, RT has been the end game goal for like...3 decades now, we always been able to do it, just not in real time.
Now we are approaching it and you would rather have an in-between step?

thats like saying you hate electric cars and would rather have hybrids...

I hate electric cars and would rather have hybrids petrol/gasoline.

As for RT, I agree.

 

[W1zzard]

@W1zzard on the second page you mention typical A380 should have 2 DP + 2 HDMI ports but this one has 3 DP + 1 HDMI, typo maybe? (from what i've seen at least only gigabyte sometimes does the 2+2 instead of 3+1). Also regarding the price, Ryan Shrout and Tom Petersen mentioned on gamers nexus a price of 130$ (129$ to 139$ which Steve nuked with something like "130$ without marketing lingo").

Haven't seen it offered for $130 anywhere. What would you say if you were on camera as Intel Marketing and people are like "uh, but it's listed for $190"

Not likely as no ReBAR on PCIe3 and the frametimes will be awful to deal with. Which is too bad as one target market for low end GPUs is older computers. The 6400 is at least still competent on PCIe3 as it's less bottlenecked than the 6500XT by simply being a lower performance part on the same bus. The 1650 is a better option if you can find it for a decent price.

Great point, while technically it is easy to force Gen 3 on a modern board and get test results, actual platforms with Gen 3 limitation do not have the BIOS updates for ReBAR.

it is just expected at this point in time.

For us tech nerds certainly, but a general consumer audience should not spend their money on a product in this state. But unfortunately it seems to have become the norm to accept sub-par software .. imagine the drama if they sold houses or cars that don't work

But we need some people to "beta test" it, right?

Well .. some of those bugs make it seem like Intel has no Q&A and their programmers don't execute their own code to see if it works. (I know both those statements are false)

 

[GreiverBlade]

I think it's better than a 6pin, most PSU come with 6+2 connectors (leaving you with a 2pin connector dangling if it was a 6pin connector on the gpu) and i'd wager the gpu might even work with a regular 6pin connector anyway since it's such a low power card

actually it would work without any connector and limited to a max powerdraw of 75w given its V-Sync 60hz consumption if kept at 1080p (my PSU offer 6pin and 6+2pin tho but i usually use sleeved extensions)

well it does have a max of 94w with spikes at 101w, which make, in the end given the small difference in performance, it a poor concurrent for the 6400 4gb even tho it has 2gb more and PCIex 4.0 X8
although if they had a 75w limited LP model at a lower price than the 6400, i would take one for giving "the new player on the market" a go (and if there was no RE-BAR "mandatory" )

 

[Dirt Chip]

Of course. - Fix, thanks.

All I'm saying is that we need early adopters to let Intel iron out its issues for future generations, and make enough cash so that developing those future generations won't seem like a pointless idea.

Would I recommend everyone to buy an Arc GPU? Definitely not! You need a tech-savvy and adventurous mind to do venture out, a hundred percent. But you can't say that it isn't worth it for anyone.


I hate electric cars and would rather have hybrids petrol/gasoline.

As for RT, I agree.

Well, the nature of truly curious early adopters is that they don't need any encouragement to do it
They are out there and will do their "thing" anyway.
And there are always those who will use A380 without even knowing it (they just need a working PC, spec doesn't matter).

Intel has tremendous amount of cash to spill on ARC. Please don't buy because you are worry for them income. It will actually just hurt the maturation process- if Intel is really serious about GPU`s and ARC as they say the best thing one can do is to buy their produce IF and only IF it fits his need and in his budget, no matter if ARC will fail completely in the process. If intel mean business with ARC (and I think they are) they need to prove that they can take a blow of two, get up and better themselves with new relevant product. If not, "not helping" them will just do good to us all, Intel included.

Buying a produce that isn't right for you in the excuse that it "help" and/or "support" do the exact opposite- you will just see more products that aren't what you need and all you are left with is the "hope" that sometime something will change. That's not how capitalisms works...

 

[Solid State Brain]

Idle power consumption is unusually high at 17W; hopefully it can be sorted out through driver updates.

By the way, an obvious advantage compared to the RX6400/6500 is that it has 4 display outputs. I think 3 is the bare minimum for power users, using for example 2 for standard displays and 1 for head-mounted display/VR. And before anybody asks: if the RX470/480 were "VR-Ready", so should be these GPUs.

 

[AusWolf]

Well, the nature of truly curious early adopters is that they don't need any encouragement to do it
They are out there and will do their "thing" anyway.
And there are always those who will use A380 without even knowing it (they just need a working PC, spec doesn't matter).

Intel has tremendous amount of cash to spill on arc. Please don't buy because you are worry for them income. It will actually just hurt the maturation process- if Intel is really serious about GPU`s and ARC as they say the best thing one can do is to buy there produce IF and only IF it fits his need and in his budget.

Buying a produce that isn't right for you in the excuse that it "help" and/or "support" do the exact opposite- you will just see more products that aren't what you need and all you are left with is the "hope" that sometime something will change. That's not how capitalisms works...

I absolutely agree again.

Buy an Arc if you're up for some adventure with something new. Don't buy one if you need a 100% reliable product with decades of history and testing. Everybody needs to decide which category they fall into. We can't say that one is more right than the other.

I didn't mean to correct you on the "of course". It was a genuine "of course" to show that I agree with you.

Well .. some of those bugs make it seem like Intel has no Q&A and their programmers don't execute their own code to see if it works. (I know both those statements are false)

That would be very sad. At least there are no issues with the latest beta driver, which is a positive sign.

 

[Solid State Brain]

@W1zzard thanks for testing but as other said some linux tests will be very interesting

I'm very curious too since I use Linux as well, but it appears that Kernel 5.20, which isn't out yet, is the one where complete support will be expected. I think it should be ready by roughly the end of September/early October.

Linux 5.20 Looks To Be Primed For Usable DG2/Alchemist Desktop Graphics & ATS-M Support - Phoronix

www.phoronix.com

After many months and a lot of work by the open-source Intel Linux graphics driver developers, Linux 5.20 looks like it will be the base kernel where the DG2/Alchemist Arc Graphics desktop GPUs and Arctic Sound M (ATS-M) server graphics card support will be ready in usable shape. [...]

 

[Dirt Chip]

For us tech nerds certainly, but a general consumer audience should not spend their money on a product in this state. But unfortunately it seems to have become the norm to accept sub-par software .. imagine the drama if they sold houses or cars that don't work

Much agreed, it should bet dealt with as toxic waste- don't touch it unless you have the tools and knowledge to deal with it.

W1zzard said:
Well .. some of those bugs make it seem like Intel has no Q&A and their programmers don't execute their own code to see if it works. (I know both those statements are false)

I tend to believe they have priority list and that bugs and glitches' are known thing that they choose not to deal with because of more important ones. Time will tell, but for now I give them the 100 days "early grace" and mark it as "understandable".

 

[docnorth]

For me most promising is the stability improvement with the new beta driver. If this trend continues (and it should), after 1 or 2 months we could talk only about a few driver bugs, well known already from the competition, cough, cough.
Or I am too optimistic...

 

[rutra80]

Doesn't look bad, looking forward for better cards and drivers.

 

[ZoneDymo]

I hate electric cars and would rather have hybrids petrol/gasoline.

As for RT, I agree.

not to go too off topic, but tell me with a bit of detail and reason, why do you hate electric cars and would rather have hybrids petrol/gasoline?

 

[chrcoluk]

Not likely as no ReBAR on PCIe3 and the frametimes will be awful to deal with. Which is too bad as one target market for low end GPUs is older computers. The 6400 is at least still competent on PCIe3 as it's less bottlenecked than the 6500XT by simply being a lower performance part on the same bus. The 1650 is a better option if you can find it for a decent price.

I checked the review again and it doesnt say rebar only works on pcie4, so where did you get that from?

for reference my pcie3 board has rebar. so is it a arc limitation?

 

[kapone32]

Now I understand why AMD is releasing a 8GB Version of the 6500XT. The review was in-depth as always and I would not personally get one of these cards. Unless the price is much more attractive once the novelty wears off, the issues will make you never buy another one for some people.