Even with these improvements, Starfield on Arc still seems to underperform, sometimes catastrophically!

It seems to be a standout thorn in Intel's side and I'm wondering if the issue isn't actually Intel, rather Bethesda's useless, outdated engine that's DX12 in name only and is still largely using DX9 calls which Arc doesn't natively support?

Chrispy_Even with these improvements, Starfield on Arc still seems to underperform, sometimes catastrophically!

It seems to be a standout thorn in Intel's side and I'm wondering if the issue isn't actually Intel, rather Bethesda's useless, outdated engine that's DX12 in name only and is still largely using DX9 calls which Arc doesn't natively support?

What?

ARC went native DX9 for more than 6+ months across all titles now, and started way back early last year.

Don't discount the hardware for lack of performance either. This is their first time really scaling their iGPUs up and bringing it to be a dGPU, which requires different thermal, memory, and interface requirements. Chips and Cheese tests show many imbalances in the hardware itself.

It doesn't matter if there's 400mm2 worth of hardware if the resources aren't all working in harmony. Many previous Intel iGPUs had their blame placed on the driver for lack of performance, when later was found out it was due to the hardware.

The most famous is that of the GMA X3000, which was their first programmable shader architecture GPU, the one their current GPUs are still an offspring of. They released a driver with hardware T&L support only for people to wonder about it's wildly varying performance, with some worse with HW T&L on than off! It's because their geometry engine and culling capabilities were very weak, things that were drastically improved and thus mostly addressed in the successors.

DavidC1What?

ARC went native DX9 for more than 6+ months across all titles now, and started way back early last year.

Ah, I didn't realise that.
I'm just speculating. Arc performance is a bit all-over-the-place but I figured they'd have managed to iron out the massive pauses and half-second stutters in Starfield by now, it's been out almost 9 months!

Good on Intel pushing forward with their driver updates, basically the only thing that reminds me that Intel and DGPUs are still a thing.

Battlemage cannot come soon enough (presuming the rumors that it was cancelled are not true/only partially true).

The pace and improvements Intel ARC is showing in their drivers, has me half-hoping Intel drops their GPU line. -Just so AMD can pick up their Driver Dev Team :laugh:

LabRat 891The pace and improvements Intel ARC is showing in their drivers, has me half-hoping Intel drops their GPU line. -Just so AMD can pick up their Driver Dev Team :laugh:

Not going to happen. Intel is making money from the ARC lineup.

lexluthermiesterNot going to happen. Intel is making money from the ARC lineup.

Not only that, but all the improvements to their GPU driver division is causing Intel's woeful IGPs to advance faster than AMD's, and laptops outsell desktops so that's an important market segment to win/improve in....

Don't get me wrong, the best intel IGP is still inferior to anything AMD offers but AMD continually fails to impress with its APUs. We were stuck with Vega IGPs for 3 architectures, and it's only in the last year that you've been able to buy a mainstream APU/laptop part and be sure that it's at least RDNA2. They launched laptop and desktop parts with Vega IGPs in January 2023 FFS, a 6-year old architecture that's on the cutting block for software support at the moment.

2022 was the first launch of a non-Vega IGP, four years after Vega was superseded and 18 months later than desktop RDNA2 parts were out in the wild. RDNA2/3 in Rembrandt/Pheonix Point APUs are mighty fine when you get them in the Ryzen7/Ryzen9 but realistically a lot of mainstream models end up with the castrated 6CU variants that aren't really any better than the basic intel Arc IGPs, outside of better drivers that cause fewer glitches in games.

Chrispy_Not only that, but all the improvements to their GPU driver division is causing Intel's woeful IGPs to advance faster than AMD's

Exactly. The Intel IGPs are not the embarrassing silliness they have been for decades. They are now reasonable options for budget and lower mid tier gaming.

Chrispy_Don't get me wrong, the best intel IGP is still inferior to anything AMD offers but AMD continually fails to impress with its APUs.

I don't think I can agree with you there. Both companies have been making strides and both have very good offerings these days for IGPs.

lexluthermiesterI don't think I can agree with you there. Both companies have been making strides and both have very good offerings these days for IGPs.

That's fair enough. The last couple of years AMD have upped their IGP game again, possibly in response to Intel IGPs no longer being a total joke.

I'm mostly disappointed in how little progress AMD made with their IGPs in the 15 years between buying ATI in 2006 and the point when Intel finally started caring about IGPs. At best, earlier IGPs in the K10 era (llano, trinity, richmond etc) seemed to be intentionally underpowered and bad, worse than the equivalent bottom-tier DDR3 variants of their lowest-end dGPUs. It almost seemed like they were intentionally holding back their APU progress to avoid cannibalising their dGPU division, likewise with Zen mobile - Raven ridge was a good start with a relatively new architecture (Vega/GCN5) but they just sat on it for about 5 years with no progress until Intel started providing some competition. For half a decade all we got were clockspeed improvements that were a result of the manufacturing process node improving, and faster RAM speeds as Micron/Samsung/Hynix incremented DDR4 speeds up over time.

Everyone likes to paint AMD as the underdog, myself included, but those 15 years of minimum-viable improvements from AMD's APU division is proof to me that without competition, all companies will just stagnate and do the bare minimum.