News Intel GPUs - Battlemage officially announced, evidently not cancelled

[IntelUser2000]

Nvidia released the performance enhancing 522.25 driver shortly after ARC release. It benefits even older cards like Turing. This reminds me of when AMD released the competitor to Geforce 3 and Nvidia shortly released a driver that improved performance drastically and got ahead of AMD.

The gain is not huge but 5% there and 10% there adds up.

 

[PingSpike]

My A380 runs ~12W lower with the fix. Important to note that GPUz is not reliable, I measured it on the power socket with energy check 3000 from the entire system. While GPUz power usage is basically unchanged my entire system dropped by about 12W. Tomshardware seems to rely only on GPUz, they have to test it and not rely on GPUz. GPUz reports 15-16W GPU power on my A380 regardless of the idle profile being enabled or disabled, considering that I only gain about 7W without a dGPU installed this power number seems inaccurate.

I usually test with before and after and a kill-a-watt measurements. Not as fancy as real reviews like techpowerup but I've never thought any of the software measuring systems are accurate.

 

[IntelUser2000]

I usually test with before and after and a kill-a-watt measurements. Not as fancy as real reviews like techpowerup but I've never thought any of the software measuring systems are accurate.

It can be but it depends on what you are measuring. Some have inaccurate systems.

Microsoft study said the RAPL(Runtime Average Power Limit) algorithm, while being an approximation is very close to the actual measured value.

Addendum: For example the laptop battery life is measured by software. Well, you could argue that it's hardware since it has sensors to do so and it exposes it to software measurement. With computers you really can't separate the two(HW/SW).

 

[Cstops]

Some probing / (micro)benchmarking and commentary of the A770 by the folks at Chips and Cheese.

Microbenchmarking Intel’s Arc A770

Intel’s Arc GPUs represent the company’s third attempt at taking on the dedicated GPU market. The company’s first attempt, the i740, tried to use the then-new AGP interface to sto…

chipsandcheese.com

 

[igor_kavinski]

If anything, Xe-HPG is even more reliant on high occupancy than Vega is. This partly explains why the A770 scales so well to higher resolutions because higher resolutions will keep the GPU at higher occupancy levels. We saw this behavior with Vega except it is even more prevalent here with the A770.

If this GPU is able to beat the 6800 XT at 8K, that would be pretty cool.

 

[maddie]

If this GPU is able to beat the 6800 XT at 8K, that would be pretty cool.

Now, that's a clue.

 

[IntelUser2000]

Some probing / (micro)benchmarking and commentary of the A770 by the folks at Chips and Cheese.

Microbenchmarking Intel’s Arc A770

Intel’s Arc GPUs represent the company’s third attempt at taking on the dedicated GPU market. The company’s first attempt, the i740, tried to use the then-new AGP interface to sto…

chipsandcheese.com

Someone was saying the issue with performance in games like CS: Go seems like it's due to lack of parallelism. The micro-level benchmarks done by chips and cheese reflect that. In that particular instance the A770 was on the level of A380.

-Higher execution unit latency
-Low memory controller and cache performance, and bad at hiding them
-Difficulty scaling memory bandwidth with threads, yet it especially needs it to make up for above mentioned deficiencies.

So how much can be solved by software and how much by hardware? They also mention part of the problem is due to the "iGPU" mentality.

Some of the aspects are not just a generation behind, but two, three, or four generations behind. Some micro level tests put it on par with AMD's Terascale 2 GPU!

 

[mikk]

On DX11 and DX9 a big part is driver, DX12 is obviously closer to the real performance. It won't reach RTX 3070 with this flaws though. I think the upcoming iGPUs based on Xe HPG (or LPG they call) should work a lot better considering they don't have to deal with the GDDR6 memory controller and PCI Express issues (which are also memory controller related). 128EU iGPU vs 128 EU dGPU will be interesting.

 

[igor_kavinski]

128EU iGPU vs 128 EU dGPU will be interesting.

I don't see how it could be interesting. iGPU would be massively constrained by lower bandwidth.

 

[mikk]

I don't see how it could be interesting. iGPU would be massively constrained by lower bandwidth.

The advantage isn't huge and Xe HPG has IMC flaws.

 

[Glo.]

https://twitter.com/x/status/1584506374746435584

Intel SOC3 on the pipeline.

CES 2023, possibly?

2048 ALUs, 192 bit bus, roughly 3050 performance, it appears.

 

[Heartbreaker]

More DXVK testing. Big improvement in Oblivion ( DX9). Witcher 3 (DX11) saw slightly lower average FPS, but improved frametime consistency:

 

[IntelUser2000]

New Beta driver, and look at the Fixed part:
• Some Intel® Arc™ A700-series Desktop Graphics products may exhibit lower than expected VRAM frequency values

From Videocardz:

Their cards equipped with 16GB VRAM is supposed to run at 17.5 Gbps, whereas in many instances the clock speed was lower at 16 Gbps.

That's a potential 9.3% increase in memory bandwidth. I assume in applications that are affected in average there could be 3-4% extra performance just with this. Don't know how widespread the effect is. You can't overclock the memory anyway.

 

[DAPUNISHER]

That Random Gaming video really downplays how bad the quality of life is with these ARC cards. I say that despite the fact he shows the viewer the extra hoops they have to jump through to do something other GPUs do out of the gate. Steve has a Intel contact now, and they provide him with hardware. It is clearly influencing his content, and not in a good way.

Daniel-San is going to give you the straight dope. TLDW is big OOF! Here is the longer version -

1. The control overlay is broken in the game the card comes bundled with. That caused multiple issues until the problem was diagnosed.
2. You can't record at 60fps above 1080p on any codec. AV1 included.
3. VRR doesn't seem to work. This could be PINIC. I don't think he tried smart sync or some of the other options? I skipped around so perhaps I missed it? He says latency is an issue and the game does not play as well as he is used to when he has VRR active.
4. The game took an extra day to download because people redeeming their games DDOS'd the server.
5. XeSS didn't seem to work, while FSR 1.0 did.
6. The ARC app wouldn't find his games. Furthermore, it made fun of him for not having any. The things it said read as passive-aggressive cyber bullying.

You have to bleed blue to want one of these for a daily driver. As a collector's item? Cool. I keep reading wait for the drivers to mature. ARC owners be like -

 

[igor_kavinski]

They should just opensource the drivers. Better yet, identify the best opensource hackers and send them an ARC A770 16GB as a gift.

 

[psolord]

They should just opensource the drivers. Better yet, identify the best opensource hackers and send them an ARC A770 16GB as a gift.

From what I am reading, maybe this would be perceived as an act of aggression towards the hackers and they start hacking Intel? xD

 

[igor_kavinski]

From the ARM/Qualcomm lawsuit thread:

Intel paid big money to NV to do graphics.

Does Intel still have access to Nvidia patents? It could be the reason behind ARC's strong raytracing performance compared to AMD.

 

[DeathReborn]

From the ARM/Qualcomm lawsuit thread:


Does Intel still have access to Nvidia patents? It could be the reason behind ARC's strong raytracing performance compared to AMD.

They do but only up to Kepler I believe. Everything after that would be a new deal.

 

[KompuKare]

From the ARM/Qualcomm lawsuit thread:


Does Intel still have access to Nvidia patents? It could be the reason behind ARC's strong raytracing performance compared to AMD.

I suspect that their relative good RT performance is as "simple" as them dedicating enough transistors to it.

Nvidia have, while AMD do some of the work on their shaders AFAIK. Nothing wrong with re-using existing hardware but as RT becomes more important transistors dedicated to fixed function is bound to become more important. Still, until MS & Sony are willing to spend big on RT performance I cannot see AMD dedicating much die space to it.

 

[DAPUNISHER]

Phil had a better gaming experience with overall. And if you are familiar with his channel, he prefers older games. He notes all of the issues and limitations with an appreciated brevity. Talks about using a frame rate limiter to cut the crazy power draw.

 

[NTMBK]

They should just opensource the drivers. Better yet, identify the best opensource hackers and send them an ARC A770 16GB as a gift.

Why would open source community bother? Plenty of abandonware open source projects out there.

 

[igor_kavinski]

Why would open source community bother?

A lot of them like a good challenge. The possibility of being hired by Intel for a lucrative salary may also factor into their willingness to hack away at the open-sourced drivers.

 

[Hulk]

Consider the overall negativity of the ARC reviews why are they out of stock at many of the larger retailers like NewEgg?

Either they are selling or there isn't any supply?

 

[VirtualLarry]

Either they are selling or there isn't any supply?

^_^ /collectors...

 

[igor_kavinski]

Either they are selling or there isn't any supply?

I think they are trying to dump the cards in markets where the Nvidia/AMD cards are too expensive for the average gamer. Even a dollar less in those places can mean saving on the cost of a meal.