Will the Arc A770 owners please stand up?

[QuietBob]

Behold, a pile of screenshots from a supported system. At some point I'll find time to do actual game testing. I didn't run any of these at stock settings, though my card does not appear to OC very well in comparison to others I've seen, so these results are closer to stock than many others.

Thanks a lot for sharing, let me add some comparisons. With that overclock, your A770 is faster than my stock 6600XT by:

dx12
+ 48% Time Spy Extreme @ 4K
+ 35% Time Spy @ 1440p

dx11
+ 8% Fire Strike Ultra @ 4K
+ 16% Fire Strike Extreme @ 1440p
+ 13% Fire Strike @ 1080p

Your oc'd card is slower than a stock 6600XT by:

dx11
- 32% 3DMark11 @ 720p (performance preset)

dx10
- 16% 3DMark Vantage @ 1024p (performance preset)

dx9.0c
- 14% 3DMark06 @ 1024p (SM 3.0 score)

dx9.0
- 8% 3DMark06 @ 1024p (SM 2.0 score)
- 23% 3DMark06 @ 1024p (Pixel Shader)
- 66% 3DMark06 @ 1024p (Vertex Shader - simple)
- 13% 3DMark06 @ 1024p (Vertex Shader - complex)
- 54% 3DMark03 @ 768p (Mother Nature)
- 31% 3DMark03 @ 768p (Pixel Shader 2.0)

dx8.1
- 36% 3DMark03 @ 768p (Battle of Proxycon)
- 41% 3DMark03 @ 768p (Troll's Lair)
- 29% 3DMark03 @ 768p (Vertex Shader)
- 22% 3DMark03 @ 768p (Ragtroll)

dx7
- 47% 3DMark03 @ 768p (Wings of Fury)

I'm intrigued to see how the A770 performs in older APIs. But I think we would get more relevant results with these legacy 3DMarks @ 1080p. 1024p and lower have little relevance today, unless run on period correct hardware.

And if you get some time, could you run GravityMark at 1080p, 1440p and 4K and upload the results to their database? They maintain a very practical Efficiency Ranking

 

[Fouquin]

Thanks a lot for sharing, let me add some comparisons. With that overclock, your A770 is faster than my stock 6600XT by:

dx12
+ 48% Time Spy Extreme @ 4K
+ 35% Time Spy @ 1440p

dx11
+ 8% Fire Strike Ultra @ 4K
+ 16% Fire Strike Extreme @ 1440p
+ 13% Fire Strike @ 1080p

Your oc'd card is slower than a stock 6600XT by:

dx11
- 32% 3DMark11 @ 720p (performance preset)

dx10
- 16% 3DMark Vantage @ 1024p (performance preset)

dx9.0c
- 14% 3DMark06 @ 1024p (SM 3.0 score)

dx9.0
- 8% 3DMark06 @ 1024p (SM 2.0 score)
- 23% 3DMark06 @ 1024p (Pixel Shader)
- 66% 3DMark06 @ 1024p (Vertex Shader - simple)
- 13% 3DMark06 @ 1024p (Vertex Shader - complex)
- 54% 3DMark03 @ 768p (Mother Nature)
- 31% 3DMark03 @ 768p (Pixel Shader 2.0)

dx8.1
- 36% 3DMark03 @ 768p (Battle of Proxycon)
- 41% 3DMark03 @ 768p (Troll's Lair)
- 29% 3DMark03 @ 768p (Vertex Shader)
- 22% 3DMark03 @ 768p (Ragtroll)

dx7
- 47% 3DMark03 @ 768p (Wings of Fury)

I'm intrigued to see how the A770 performs in older APIs. But I think we would get more relevant results with these legacy 3DMarks @ 1080p. 1024p and lower have little relevance today, unless run on period correct hardware.

And if you get some time, could you run GravityMark at 1080p, 1440p and 4K and upload the results to their database? They maintain a very practical Efficiency Ranking

Yeah older APIs are pretty brutal. As another point of comparison, here's the R9 285X eating Arc's lunch in 3DM03 and 06:

Also Vega 56 (heavily tuned ES with V64 clocks and power limits) is right on its butt in Vantage:

When I get my testing situation sorted out a little better so I'm not constantly ripping apart my main rig to swap cards around I'll get more results. I'd like to get access to any other tuning utility for Arc, something that actually functions properly and gives the most basic options like fan control.

 

[lexluthermiester]

Behold, a pile of screenshots from a supported system. At some point I'll find time to do actual game testing. I didn't run any of these at stock settings, though my card does not appear to OC very well in comparison to others I've seen, so these results are closer to stock than many others.

View attachment 269781View attachment 269782View attachment 269783View attachment 269784View attachment 269785View attachment 269786View attachment 269787View attachment 269788View attachment 269789View attachment 269790

Night & Day difference.

 

[R-T-B]

Have any of you played with perhaps using dxvk in the dx9 scenarios to improve performance?

 

[Fouquin]

Have any of you played with perhaps using dxvk in the dx9 scenarios to improve performance?

What do you recommend for accomplishing this? It's my understanding that Arc is already using 9on12 translation to run DX9, I'm not sure what the difference would be with another layer of translation.

 

[lexluthermiester]

What do you recommend for accomplishing this? It's my understanding that Arc is already using 9on12 translation to run DX9, I'm not sure what the difference would be with another layer of translation.

My understanding is that there's almost no difference. Could be wrong though..

 

[Calmmo]

It wouldnt be another layer, you'd by running dxvk as far the card's aware, skipping intels "lacking" driver level container or w/e they use. (in theory)

 

[R-T-B]

What do you recommend for accomplishing this? It's my understanding that Arc is already using 9on12 translation to run DX9, I'm not sure what the difference would be with another layer of translation.

It would be translating to Vulkan instead of dx12. Guess it would come down to the quality of the dx12 vs vulkan driver, and the quality of the translators.

To use dxvk, you typically download a release, and unzip the dlls (x64 or x32 depending) to your windows app root folder.

My understanding is that there's almost no difference. Could be wrong though..

That's what I think too but people said that about dxvk on Windows with AMDs old horrid DX11 driver and DXVK was actually faster, so who knows?

 

[Dristun]

Our future Chinese overlords came through and now I'm an excited owner of an A770. Apologies for improper cabling! Will sort that out later (maybe ).

I'm thinking of starting a "gaming with arc" megathread where we can share all the problems and solutions. Sure, Intel's discord exists, but it's not indexed by google and considering the state of drivers I imagine the thread might come in handy for unlucky kids who will get these in prebuilds.

Also if someone wants anything particular tested, feel free to ask, I might check it out on the weekend.

 

[Fouquin]

Our future Chinese overlords came through and now I'm an excited owner of an A770. Apologies for improper cabling! Will sort that out later (maybe ).

I'm thinking of starting a "gaming with arc" megathread where we can share all the problems and solutions. Sure, Intel's discord exists, but it's not indexed by google and considering the state of drivers I imagine the thread might come in handy for unlucky kids who will get these in prebuilds.

Also if someone wants anything particular tested, feel free to ask, I might check it out on the weekend.

Actually yes, definitely get a thread going for Arc gaming. Documenting successes and failures, plus the effects of driver updates would be valuable. I am fully aware that we're unpaid beta testers, but if we can help others make more informed decisions that's a win.

 

[Solaris17]

Man I honestly love mine, unbias as possible the intel A770 looks fantastic. I mean like the factory one. Like damn. best ref design I have seen in a long time.

EDIT:: As for drivers, they have come a LONG way since I got my A380. I ran into a few oddities most performance inconsistency, but it much more stable. The control panel is absolute trash, but atleast the driver has gotten better.

As for external enclosures, thats a big no. None of my 3 cards like it in any fashion.

 

[sepheronx]

I cant find one here in Canada.

Memory Express had them briefly but were in limited quantity and that was it.

750's on other hand are available.

 

[Fouquin]

Behold, a pile of screenshots from a supported system. At some point I'll find time to do actual game testing. I didn't run any of these at stock settings, though my card does not appear to OC very well in comparison to others I've seen, so these results are closer to stock than many others.

View attachment 269781View attachment 269782View attachment 269783View attachment 269784View attachment 269785View attachment 269786View attachment 269787View attachment 269788View attachment 269789View attachment 269790

So I did some fiddling around with Arc now in my main system with my curve optimized settings, a new board BIOS and the new driver. Results are trending dramatically different now. I'm not entirely sure why things have so drastically improved, perhaps better support in general from ASRock for Arc on their boards? Either way, my new results are NOTHING like the old ones.

Result

www.3dmark.com

Thanks a lot for sharing, let me add some comparisons. With that overclock, your A770 is faster than my stock 6600XT by:

dx12
+ 48% Time Spy Extreme @ 4K
+ 35% Time Spy @ 1440p

dx11
+ 8% Fire Strike Ultra @ 4K
+ 16% Fire Strike Extreme @ 1440p
+ 13% Fire Strike @ 1080p

Your oc'd card is slower than a stock 6600XT by:

dx11
- 32% 3DMark11 @ 720p (performance preset)

dx10
- 16% 3DMark Vantage @ 1024p (performance preset)

dx9.0c
- 14% 3DMark06 @ 1024p (SM 3.0 score)

dx9.0
- 8% 3DMark06 @ 1024p (SM 2.0 score)
- 23% 3DMark06 @ 1024p (Pixel Shader)
- 66% 3DMark06 @ 1024p (Vertex Shader - simple)
- 13% 3DMark06 @ 1024p (Vertex Shader - complex)
- 54% 3DMark03 @ 768p (Mother Nature)
- 31% 3DMark03 @ 768p (Pixel Shader 2.0)

dx8.1
- 36% 3DMark03 @ 768p (Battle of Proxycon)
- 41% 3DMark03 @ 768p (Troll's Lair)
- 29% 3DMark03 @ 768p (Vertex Shader)
- 22% 3DMark03 @ 768p (Ragtroll)

dx7
- 47% 3DMark03 @ 768p (Wings of Fury)

I'm intrigued to see how the A770 performs in older APIs. But I think we would get more relevant results with these legacy 3DMarks @ 1080p. 1024p and lower have little relevance today, unless run on period correct hardware.

And if you get some time, could you run GravityMark at 1080p, 1440p and 4K and upload the results to their database? They maintain a very practical Efficiency Ranking

Might need to revisit this comparison soon. It looks like whatever combination of settings, hardware, and/or drivers I've stumbled upon is going to put up a much better showing. I'll look at getting 1080p results for the older stuff with default settings, as well as GravityMark at some point.

Also yes game testing SOON. I had a mishap with Elden Ring constantly crashing to desktop (now solved), and I went back to using a different card for pure gaming. Now though, I'm fully set up to stick with the A770 as my primary for at least a little while.

 

[Stephen.]

@Fouquin

@Solaris17

I will gladly daily drive the Intel ARC A770 16GB for you and for the community as a whole.

Not only am I a fan, but I am also an enthusiast and I love tinkering with hardware and software.

As you can tell I have an Intel based system and it's My daily driver.

I have a proposal if you are willing to agree to it, "Send the Intel ARC A770 16GB to me and I will use it in my rig and post all sorts of findings that I will get"

I have an MSI RTX 2070 Aero that I am willing to send your way if you agree.

I will only run the Intel ARC A770 16GB and report to the community, I don't mind teething issues and I love me some Intel tech #Intel fanboy

Man I honestly love mine, unbias as possible the intel A770 looks fantastic. I mean like the factory one. Like damn. best ref design I have seen in a long time.

EDIT:: As for drivers, they have come a LONG way since I got my A380. I ran into a few oddities most performance inconsistency, but it much more stable. The control panel is absolute trash, but atleast the driver has gotten better.

As for external enclosures, thats a big no. None of my 3 cards like it in any fashion.

So I did some fiddling around with Arc now in my main system with my curve optimized settings, a new board BIOS and the new driver. Results are trending dramatically different now. I'm not entirely sure why things have so drastically improved, perhaps better support in general from ASRock for Arc on their boards? Either way, my new results are NOTHING like the old ones.

View attachment 271074View attachment 271079View attachment 271073View attachment 271078

Result

www.3dmark.com

Might need to revisit this comparison soon. It looks like whatever combination of settings, hardware, and/or drivers I've stumbled upon is going to put up a much better showing. I'll look at getting 1080p results for the older stuff with default settings, as well as GravityMark at some point.

Also yes game testing SOON. I had a mishap with Elden Ring constantly crashing to desktop (now solved), and I went back to using a different card for pure gaming. Now though, I'm fully set up to stick with the A770 as my primary for at least a little while.

 

[lexluthermiester]

So I did some fiddling around with Arc now in my main system with my curve optimized settings, a new board BIOS and the new driver. Results are trending dramatically different now. I'm not entirely sure why things have so drastically improved, perhaps better support in general from ASRock for Arc on their boards? Either way, my new results are NOTHING like the old ones.

You're not alone. This is an experience many are having. I was expecting improvement's along the way as refinements to the driver and hardware comparability improves, but not by so much! Very interesting.

The score for the older benchmarks seem very telling indeed. Intel is delivering on their promise to improve performance for older titles as well. As you showed, the 3DMark 03 scores more than tripled. This directly implies that support for Old API's(OGL/DX10/DX9/DX8) is being worked on with serious intent on making ARC a viable card classic/legacy games.

Question, do you have time/desire to test that card on a non ReBar system? It would be interesting to see the deference with the current driver set.

 

[Fouquin]

Question, do you have time/desire to test that card on a non ReBar system? It would be interesting to see the deference with the current driver set.

I can disable ReBAR and give it a few runs through the various tests. I'm not keen to remove the card and throw it back onto the test bench right now using an older platform.

@QuietBob I will run a complete sweep through the rest of the resolutions/APIs/render schemes tomorrow, but here's some results tinkering with GravityMark at 1440p. Power testing with vulkan.

 

[lexluthermiester]

I can disable ReBAR and give it a few runs through the various tests.

That would work.

I'm not keen to remove the card and throw it back onto the test bench right now using an older platform.

Understandable. No worries.

 

[QuietBob]

I will run a complete sweep through the rest of the resolutions/APIs/render schemes tomorrow

Thanks a lot for these! I expected the A770 to be closer in efficiency to the A380. Maybe default settings will yield better results? In the meantime I ran legacy 3DMarks at 1080p on my stock 6600XT:



11 and Vantage use the Extreme preset, while 06 and 03 the default one.

 

[Fouquin]

Maybe default settings will yield better results?

Nope, because default is 2400MHz @ 190W PL, so it falls right between those two results shown.

DX12 doesn't run, OpenGL and DX11 barely run, Ray Tracing doesn't run. So perhaps GravityMark is not going to be a good way to draw comparisons except in the, "does it work," category.

 

[QuietBob]

DX12 doesn't run, OpenGL and DX11 barely run

In my testing Vulkan is the most efficient API in this benchmark. It'd be great if you could run it in the remaining resolutions. Those anonymous 6600XT results in the efficiency ranking are mine

Also, interesting to see how your A770 is pegged at 155w at all times, while the 6600XT's power consumption varies with a 145w PPT limit:

 

[Solaris17]

Just an FYI Newegg has the entire range in stock currently

 

[lexluthermiester]

Just an FYI Newegg has the entire range in stock currently

Ninja'd again. That's twice in one day...

 

[Stephen.]

To the owners I have a request, kindly test for me

Metal Gear Solid V (Ground Zeroes $ The Phantom Pain)

Tom Clancy's The Division

Death Stranding

 

[Fouquin]

To the owners I have a request, kindly test for me

Metal Gear Solid V (Ground Zeroes $ The Phantom Pain)

Tom Clancy's The Division

Death Stranding

Good choices. Sadly I own none of those.

 

[QuietBob]

Good choices. Sadly I own none of those.

The Division has a demo that includes a benchmark.