Intel Integrated Graphics Performance From Gen9 To Meteor Lake Arc Graphics

[Jimzz]

Any power test testing from the outlet/brick and not software numbers?

[smitty3268]

The results I showed in the above post reflect a 7.5% average benefit. The difference between AMD's 680M and 780M iGPUs is only about 15%. So, Intel beat AMD by half as much as AMD's own generational improvement. Call what you want, as long as you don't say it's insignificant.

Presumably that includes things like this: https://phoronix.com/benchmark/resul...16-scalar.svgz

where it's 200% performance on a (for me) meaningless synthetic benchmark.

If you want to calculate what the actual average benefit is while stripping out all synthetic benchmarks, that'd be a lot more interesting to me. I would assume that 7.5% benefit wouldn't still be there, but if it is then that would be much more meaningful to me. I'm not bothered enough to do it myself.

If you don't like Michael's benchmark suite, then why not suggest to him which things to drop or add?

Eh, I could come up with a much more meaningful set of tests if I wanted to, but it's not going to change what Michael does. He's going to base stuff off of having easy to reproduce and automate benchmark modes, rather than what is actually a meaningful test. That's always how Phoronix has done things, and I'm sure it always will be. I understand why.

Besides, there is some value in knowing those numbers too. I just think it's important to know the context of stats in order to understand what they mean. There's a reason behind the old saying about lies, damn lies, and statistics. Interpretation matters a lot more than just reading off 1 final magic number you came up with.

[coder]

This is a made up "metric" by you. Non-argument whatsoever.

It wasn't meant as a standard metric, or anything like that. It was just to put their achievement in some perspective.

It's also interesting that AMD's 780M has roughly the same specs as their 680M, but shrank from N6 to N4. The other big change seems to be that it's RDNA3, instead of RDNA2.

being able to surpass it even by 7.5% is also impressive, and ability to achieve this is a significant achievement.

Yes, and one thing surprising about that is that it's basically the same architecture as Intel's Alchemist dGPUs, which lagged so badly behind competing products with similar amounts of compute and memory bandwidth. I think that's a promising sign for Battlemage.

[coder]

Eh, I could come up with a much more meaningful set of tests if I wanted to, but it's not going to change what Michael does. He's going to base stuff off of having easy to reproduce and automate benchmark modes, rather than what is actually a meaningful test. That's always how Phoronix has done things, and I'm sure it always will be. I understand why.

The entire suite is pretty huge, though. If there are issues with some benchmarks, he can drop them or reduce the number of test cases to reduce their impact on the overall scores.

I just think it's important to know the context of stats in order to understand what they mean. There's a reason behind the old saying about lies, damn lies, and statistics. Interpretation matters a lot more than just reading off 1 final magic number you came up with.

True, there's rarely any transparency into his benchmark selections. These selections can be used to skew the numbers one way or another. When it's a product he buys with his own money, I think that's less likely. For a product that he's given, I have more suspicions about the pressure he's under to show it in a positive light.

[smitty3268]

True, there's rarely any transparency into his benchmark selections. These selections can be used to skew the numbers one way or another. When it's a product he buys with his own money, I think that's less likely. For a product that he's given, I have more suspicions about the pressure he's under to show it in a positive light.

I have zero concern that Michael is intentionally skewing his reviews one way or another for money from the manufacturers. My concern would only be that he's skewing things unintentionally by using easy to run tests rather than good tests, and not worrying about how that skews things at all (if they even do).

[coder]

I have zero concern that Michael is intentionally skewing his reviews one way or another for money from the manufacturers.

Pay attention to how he tests new products. I believe he's always transparent about when the hardware was given to him vs. purchased himself.

My concern would only be that he's skewing things unintentionally by using easy to run tests rather than good tests, and not worrying about how that skews things at all (if they even do).

"Easy to run"? You know PTS is automated, right? I think the only practical concerns are how long certain tests take to run, and then each hardware configuration he wants to include in the comparison needs manual intervention (i.e. swapping out a hardware component).

[smitty3268]

"Easy to run"? You know PTS is automated, right?

Yes, exactly. Which means he never runs any tests he can't easily automate.

Even something that has a benchmark mode won't be included if he can't run it from the command line and see the results there - needing to click a button and view the output in a window means it can't be included.