The most advanced integrated GPU series today?

[passmarkrocks]

Is the AMD trinity series the leading integrated GPU of today?

I was told that they are a tier better than the Intel HD series.

Apparently integrated GPU are getting more and more powerful, can we say that in the foreseeable future, integrated GPU will be good enough for new release games? How long do we have to wait for?

[godfather]

Yeah... the fastest integrated GPU is the Radoen HD 7660G. It's significantly faster than Intel's best (while consuming way more power).

It's good enough to play most recent games on LOW settings.

Integrated GPUs continue to get better and better, but they are still far enough from being sufficient for serious gamers.

And in a couple of years, the next-gen consoles will be released. So chances are, they're going to fall behind again.

[passmarkrocks]

thanks for the reply!

3DMark 11 - Performance 1280x720 scores:

1143 for 7660G
704 for 7640G
665 for HD Graphics 4000

so is it safe to say that HD Graphics 4000 is on par with 7640G but a tier lower than 7660G?

what is the difference between all of these benchmarks? which benchmark should I use if i only care about gaming performance?

edit: however, on videocardbenchmark's list, apparently the difference between 7660G and HD 4000 is very small. what gives?

http://i.imgur.com/B7Ahx.jpg

[godfather]

Well, you have to take benchmarks with a grain of salt.
From what I've seen, the 7640G should be slightly faster than the 7660g, which is slightly faster than the the Intel HD Graphics 4000.

Yeah, the difference isn't all that big. By significant, I'm talking probably 10% or so, with the 7660g being 20% faster.

[passmarkrocks]

Well, you have to take benchmarks with a grain of salt.
From what I've seen, the 7640G should be slightly faster than the 7660g, which is slightly faster than the the Intel HD Graphics 4000.

Yeah, the difference isn't all that big. By significant, I'm talking probably 10% or so, with the 7660g being 20% faster.

thanks for the quick reply!

[David (PassMark)]

There is also an issue in PerformanceTest V7 when it deals with machines that have multiple active video cards and the video cards are not homogeneous. A few years back this didn't happen. It use to be that if you had multiple video cards they were all the same model (e.g. SLI). Or if you had integrated graphics and a discrete card then the integrated graphics would be automatically turned off when you added the discrete card.

Now there a number of situations where you might have integrated graphics and a discrete card active at the same time. PT7 assumes the first active video card detected is the one used for the 3D tests. Some of the time this is not the case however. This has slightly distorted some of the results for mobile / integrated video cards.

In PerformanceTest V8 we detect all the available cards and also detect the card that is used during the 3D tests. This is then recorded along with the benchmark results. So the PT8 graphs should be a bit more accurate. They should be up on the site in a couple of weeks. Plus PT8 also does DX11, Direct compute and runs at higher resolution. Which should also help the accuracy of the rankings.

[passmarkrocks]

There is also an issue in PerformanceTest V7 when it deals with machines that have multiple active video cards and the video cards are not homogeneous. A few years back this didn't happen. It use to be that if you had multiple video cards they were all the same model (e.g. SLI). Or if you had integrated graphics and a discrete card then the integrated graphics would be automatically turned off when you added the discrete card.

Now there a number of situations where you might have integrated graphics and a discrete card active at the same time. PT7 assumes the first active video card detected is the one used for the 3D tests. Some of the time this is not the case however. This has slightly distorted some of the results for mobile / integrated video cards.

In PerformanceTest V8 we detect all the available cards and also detect the card that is used during the 3D tests. This is then recorded along with the benchmark results. So the PT8 graphs should be a bit more accurate. They should be up on the site in a couple of weeks. Plus PT8 also does DX11, Direct compute and runs at higher resolution. Which should also help the accuracy of the rankings.

thanks for the detailed explanation!

another question while we are on this topic:

assuming i want something whose 3DMark 11 - Performance 1280x720 score is in the 600~1000 range, should i just go for an highend integrated GPU, like intel HD or AMD trinity, over a medium level dedicated GPU like AMD HD 6650M? i was told by a friend that integrated GPUs last longer, because dedicated GPUs overheat much more often and has "solder joints" that crack much more easily.

[David (PassMark)]

I can't tell you what cards are in that range. 3DMark11 isn't our product and I don't monitor their scores.

I would think the opposite about video cards. Integrated cards are part of the CPU. In many cases these CPUs are in a socket (without any solder).

Dedicated cards often have fans which get full of dust.

Really depends on the design of the individual computer I think. I don't think there is any generalization that can be made.

[passmarkrocks]

I can't tell you what cards are in that range. 3DMark11 isn't our product and I don't monitor their scores.

I would think the opposite about video cards. Integrated cards are part of the CPU. In many cases these CPUs are in a socket (without any solder).

Dedicated cards often have fans which get full of dust.

Really depends on the design of the individual computer I think. I don't think there is any generalization that can be made.

ok thanks.