Intel Arc 3 A380 vs. AMD Radeon RX 6400: 51 Game Benchmark

Unfortunately this seems like a common theme for those that test the Intel low end A380 card. The Intel Arc control is even worse than the drivers. And rumor has it there are also hardware issues that may need to be addressed in the next revision. Most want to see what the A750-A770 can do but they are being so tight lipped on cost, performance and release date its almost comical at this point.
 
Interesting and enlightening review, thanks. What gets me is the huge difference in size between both cards.

One comment: While I see why you don‘t include games that weren‘t playable in the final totals, there‘s also a point in including them with ‚-100%‘ as that‘s the actual performance you get, particularly if we‘re talking about several games and not just one odd outlier.

The saddest part is that neither of the two cards is an upgrade option for older OEM type PC.
 
If one assumes there's nothing vastly different in terms of architectural configuration, then on paper, the A380 should be significantly better than the RX 6400 - 41% higher peak FP32 throughput, 41% higher peak texel fillrate, 45% more peak local memory bandwidth, 4 times more L1 cache, 4 times more L2 cache. It also has separate units for matrix multiply units that supports FP16, just like Nvidia's tensor cores. It obviously lacks AMD's Infinity Cache, but still - the hardware should more than good enough.

But 11% slower at 900p? Yeah, the drivers are a mess. Would be interesting to see 1440p performance figures to see if it gets any average lead at all.

What gets me is the huge difference in size between both cards.
The DG2-128 chip is 48% larger than the Navi 24, with 33% more transistors. The A380 consumes 42% more power too, as well as 2 GB more RAM, which is why Intel have gone for a larger board and cooler design.
 
It s a good gpu when you dont need all that space and connections like amd nvidia. low power and fast enough for CAN it run crysis test too.
for using exel powerpoint 735 kb 1 mb are enough. and win 3.0-3.1 11
 
Pretty garbage attempt by Intel. I don't expect higher end models to be any different. Hype, once again, fails miserably. I don't know who dreamt first gen would be any good. If they're lucky, they're going to need at least a couple of years to get where Nvidia and AMD are now. And I'm not talking about packing a ton of cores, transistors and whatnot, but about expertise. Assuming they won't sell the division with a massive loss by then.
 
These results are really dissapointing, the a380 was my hopeful upgrade to move away from my aging LP 560 in my media PC, but these performance results and the fact that the cards come with 8 pin connectors destroys that idea. I guess the old polaris GPU will have to hold out another year or two, and pray the 7400/7500 doesnt suck magnum dong....
If one assumes there's nothing vastly different in terms of architectural configuration, then on paper, the A380 should be significantly better than the RX 6400 - 41% higher peak FP32 throughput, 41% higher peak texel fillrate, 45% more peak local memory bandwidth, 4 times more L1 cache, 4 times more L2 cache. It also has separate units for matrix multiply units that supports FP16, just like Nvidia's tensor cores. It obviously lacks AMD's Infinity Cache, but still - the hardware should more than good enough.

But 11% slower at 900p? Yeah, the drivers are a mess. Would be interesting to see 1440p performance figures to see if it gets any average lead at all.


The DG2-128 chip is 48% larger than the Navi 24, with 33% more transistors. The A380 consumes 42% more power too, as well as 2 GB more RAM, which is why Intel have gone for a larger board and cooler design.
The card's ability to game in linux, with open source drivers, would be very interesting to see. Open source intel drivers are actually quite good, maybe phoronix will test the 380 at some point?
 
Pretty garbage attempt by Intel. I don't expect higher end models to be any different. Hype, once again, fails miserably. I don't know who dreamt first gen would be any good. If they're lucky, they're going to need at least a couple of years to get where Nvidia and AMD are now. And I'm not talking about packing a ton of cores, transistors and whatnot, but about expertise. Assuming they won't sell the division with a massive loss by then.


Yeah, anyone who missed the TWO DECADES of crap drivers from Intel GMA Graphics division were the only ones even potentially hopeful about ARC!

its going to be at-least 5 more years before Intel will touch AMD's wide comparability and optimizations and maybe double that time to match NVIDIA
 
Last edited:
Well it's not a great start, but a journey of a thousand miles begins with the first step.

I think Intel should focus on making the perfect 1650 low-profile killer before moving on to more ambitious targets. AMD had their chance but blew it with the PCI-E 4.0 requirement and choked memory bus, not to mention higher power draw.
 
Well it's not a great start, but a journey of a thousand miles begins with the first step.

I think Intel should focus on making the perfect 1650 low-profile killer before moving on to more ambitious targets.
That would be the A380 but we see how that turned out. Plus, look at the size and power of that thing vs the 6400.

Personally, I had really hoped for the A380 to be just that, supporting the latest feature set and codecs - a worthy, reasonably priced modern entry level card.

With neither nVidia nor AMD offering this, Intel had a great chance but blew it.
 
The saddest part is that neither of the two cards is an upgrade option for older OEM type PC.

Actually the 6400 is a fine option to upgrade an older OEM type PC, especially the "desktop-size" ones with only space for half height and/or single slot. I've been using one in an Optiplex 9020 (PCIe 3.0) for over a month and it works well, noticeably faster than the 1050 Ti I have in another identical Optiplex 9020 (which I did not expect). The only thing holding it back for these use cases is the price is too high. The 6400 should be $99 or thereabouts.
 
Actually the 6400 is a fine option to upgrade an older OEM type PC, especially the "desktop-size" ones with only space for half height and/or single slot. I've been using one in an Optiplex 9020 (PCIe 3.0) for over a month and it works well, noticeably faster than the 1050 Ti I have in another identical Optiplex 9020 (which I did not expect). The only thing holding it back for these use cases is the price is too high. The 6400 should be $99 or thereabouts.
Sure, the 6400 can be used size and power wise in older PC but they are nowhere near as nice as the RX 550 was for that application back when it was released.

Due to its ‘mobile APU companion card‘ limitations / cuts it is not as nice as it could or should be. Had the 6400 and 6500 been specifically designed as an entry level gaming card like the 550 and 560 were that would have been great.
 
To me it looks like the hardware isn't that terrible. The results just scream driver immaturity. When you're facing companies that have drivers packed with a good 25 years of tweaks, tuning and enhancements you're always chasing to catch up.

You let the hardware side continually evolve and you hit the software side all day, everyday. 24/7 shifts with a big team. You know you have to work three times as hard as the opposition to catch up so you hire and hire until you're there.

Start with the most played games on a platform and work down the list. Rome wasn't built in a day, Nvidia's graphics drivers certainly weren't.
 
Actually the 6400 is a fine option to upgrade an older OEM type PC, especially the "desktop-size" ones with only space for half height and/or single slot. I've been using one in an Optiplex 9020 (PCIe 3.0) for over a month and it works well, noticeably faster than the 1050 Ti I have in another identical Optiplex 9020 (which I did not expect). The only thing holding it back for these use cases is the price is too high. The 6400 should be $99 or thereabouts.
Actually the 6400 is a terrible option for upgrading older OEM type PCs. It'd be great, that is if the 1050ti hadn't already been passed by the RX 560 and GTX 1650LP, both of which have been available for over 3 years at the same price. Oops.

Even at the appropriate sub $100 price, the lack of VRAM and the restricted x4 bus, and inconsistent performance on 3.0 systems lead to the card being a non starter. AMD should be ashamed for releasing such a horrible product, one that moved the market backwards.
To me it looks like the hardware isn't that terrible. The results just scream driver immaturity. When you're facing companies that have drivers packed with a good 25 years of tweaks, tuning and enhancements you're always chasing to catch up.

You let the hardware side continually evolve and you hit the software side all day, everyday. 24/7 shifts with a big team. You know you have to work three times as hard as the opposition to catch up so you hire and hire until you're there.

Start with the most played games on a platform and work down the list. Rome wasn't built in a day, Nvidia's graphics drivers certainly weren't.
Intel has been writing GPU drivers for their hardware for over 20 years at this point. Intel's GMA has been around for a *LONG* time.
 
Intel has been writing GPU drivers for their hardware for over 20 years at this point. Intel's GMA has been around for a *LONG* time.

He wrote very well "When you're facing companies that have drivers packed with a good 25 years of TWEAKS, TUNING and ENHANCEMENTS" > INTEL just updated drivers so that a new chip with the same old 4x00 or then (U) HD/Iris GPUs would be detected correctly. Fixed the drivers? Hummmm no.
 
Intel has been writing GPU drivers for their hardware for over 20 years at this point. Intel's GMA has been around for a *LONG* time.
They sure have. Problem being they were never any good and never updated rapidly enough for modern gaming. Nobody bought a GMA machine to run Crysis and Intel knew it. Fine if you wanted to play that ten year old game that finally runs above 20FPS on your integrated GPU or wait six months for them to fix a refresh incompatibility issue. Not so good if you need a new driver with specific fixes for the four top tier games released this month and the six next month and so on. You have to insert yourself into the gaming industry quite a bit for starters to get the jump.

It was only a couple of years ago they decided to move to faster driver iterations. In order to create a real competitor you have to produce the full goods. Intel have to create all the tweaks for known and existing modern game engines, massage performance for all the old ones, and get up to date with the very latest titles with specific fixes for them too.

It's a massive task when that's what Nvidia and AMD have been doing for decades with reams of historical hardware and software fixes for games. They're putting the icing on each month, Intel are still cracking the eggs and whisking right now.
 
Actually the 6400 is a terrible option for upgrading older OEM type PCs. It'd be great, that is if the 1050ti hadn't already been passed by the RX 560 and GTX 1650LP, both of which have been available for over 3 years at the same price. Oops.

Prices from 3 years ago are irrelevant when you want something now. The 1650 LP is $240 and up. The 6400 LP is $150 and that beats the hell out of the 1650LP for cost per frame at PCIe 3.0. And the 560 gets half the frames of the 6400 and is unavailable unless you go used.

Even at the appropriate sub $100 price, the lack of VRAM and the restricted x4 bus, and inconsistent performance on 3.0 systems lead to the card being a non starter. AMD should be ashamed for releasing such a horrible product, one that moved the market backwards.
How can a slot-powered card that delivers more frames at PCIe 3.0 at a "$100" price level than any other be a non-starter? What is your $150 alternative (using today's actual prices) for an affordable slot powered card at this price? The best thing I can find is used slot-powered 1650 (MSI Ventus) at $150 and that is a better card. But used, so there's a bit of taking your chances.
 
Sure, the 6400 can be used size and power wise in older PC but they are nowhere near as nice as the RX 550 was for that application back when it was released.

Due to its ‘mobile APU companion card‘ limitations / cuts it is not as nice as it could or should be. Had the 6400 and 6500 been specifically designed as an entry level gaming card like the 550 and 560 were that would have been great.

I agree that it's limitations are grating but you have to ask the question: Why is someone buying this card? Are they buying it to turn a decommissioned office PC into an HTPC or are they buying it to turn that office PC into a cheap gaming box?

My feeling is that most people who are considering repurposing an old office PC want to do gaming with it and the 6400 works pretty well for that, even at PCIe 3.0. Are some of these people wanting to stream? Maybe, but if this is going to be your best gaming PC option (why do this if you have something better already?), it's pretty likely they just want a decent card to game with.

The 6400 is a card with limitations but as a drop-in gaming card for an old office PC it's a good option, especially if you need/want to buy new. However if the 1650 was the same price, there would be zero reason to buy the 6400.
 
Last edited:
6400 is one of the worst cards ever, and A380 get's smashed in most benchmarks. I really would like to know who would even consider these POS when you could buy far more powerful second hand cards for same money. I wouldn't even take one if you gave it to me for free.
 
Last edited:
AMD and Intel battling for who will release the worst GPU of all time and it seems like Intel has AMDs number.
 
That AMD card it is what it is, but that Intel card...........what a waste of materials.700 grams of electronics waste.
 
Back