AMD Benchmarks Zacate APU, 2x Faster GPU Performance than Core i5

[informal]

The only botched result is Direct2D hw acceleration in IE9.The game is Direct3D and probably ran as it should.We have a user here who will run it on his i3 laptop for comparison soon.

[jimbo75]

Attachment 107723

http://www.youtube.com/watch?v=-lLNHTu4oLk

What is interesting here is Zacate appears to be 1.5x faster in that, whereas in the other browser/city of heroes/amazon video it's more often at least 2x faster. The 520M has 50% higher clocks, so that might make even more sense. (note these are totally different benchmarks being discussed - there is nothing to link the city of heroes/amazon bookshelf demo to this AMD spacewhatever one).

[Manicdan]

ok i got home and tested it myself

i have 4850xfire, and those never went up a hair durring any test
my 1055T however saw about one core being near maxed out to do it, so 4ghz of a modern cpu can be used. i did not see any real multi core scaling at all.
even at full 1080p i still get 1900-2000 on the regular, and 600ish (660 tops) on the hallucinogenic version. shrinking the screen DID NOT reduce cpu usage, or affect my score much at all. i also noticed that some times task manager dosnt even report the same cpu usage (in one run i get 30% of a single core being used, the next 100%), but with the same scores. somethings really just weird.

looks like a mobile chips with no turbo will not enjoy that benchmark much
anyone have a task manager cpu usage screenshot for a low speed laptop quad or anything?

more testing
amazon uses about 5% of my cpu, none of my gpu and im 60fps locked. no idea how stressful that site really is.

i think ill have to downclock my cpu to 800mhz and do this all over again and see what happens

[The Coolest]

Definitely a limitation of some sort, here are the scores i got on my h/w and FireFox 4 Beta 5
And the bookshelf demo is pretty much at a solid 60fps.
e6850, 8800 ultra and 8gb of ram - all stock

[Particle]

The results could all be the same because the 2D acceleration of most cards is roughly on par. Unless Direct2D uses shaders instead of the traditional 2D engines in cards (no idea), there wouldn't be a lot of differentiation between cards. Low end cards and high end cards are all the same.

As for CPU utilization, I saw someone way it pegged a core, but that isn't what I'm seeing at 4GHz. I tied IE9 to core 0 for this test. It just hits very mild utilization. GPU utilization never exceeds 22%, however that is calculated. It may not be accurate for what we're trying to measure since this is Direct2D.

Test 1 & 2

[jimbo75]

I'm starting to wonder at these 1900+ intel scores in the Psychedelic test btw. Is there any way to change the name of it to see if the scores stay the same or plummet?

[The Coolest]

Disappointing, I just looked at the AMD video.
You can clearly see that on the AMD, there's FRAPS in the top-left corner of the window - meaning Direct2D, but on the Intel demo FRAPS shows up in the game, but not in the psychadelic demo, meaning it wasn't GPU accelerated.
Shame on AMD.

[jimbo75]

Disappointing, I just looked at the AMD video.
You can clearly see that on the AMD, there's FRAPS in the top-left corner of the window - meaning Direct2D, but on the Intel demo FRAPS shows up in the game, but not in the psychadelic demo, meaning it wasn't GPU accelerated.
Shame on AMD.

I'd be interested to know exactly how does this prove "shame" on AMD? It couldn't be anything else, ie the intel system doesn't know when it should turbo in 2d etc?

Have you run this psychedelic test without gpu acceleration btw? It's a lot worse than what this intel system is showing.

[Deluded]

I can test it on my desktop i3 when I get home. I can test it both on stock and OC'ed.

[gallag]

something mighty strange about all this.

[jimbo75]

AMD hasn't "disabled" gpu acceleration on the intel system - try running the psychedelic browser test in your normal browser and you'll see that the performance is in single digits when there is no gpu acceleration.

http://ie.microsoft.com/testdrive/

There you go, see how Psychedelic Browsing runs on your current non-accelerated browser. Single digits?

[The Coolest]

I'd be interested to know exactly how does this prove "shame" on AMD? It couldn't be anything else, ie the intel system doesn't know when it should turbo in 2d etc?

Have you run this psychedelic test without gpu acceleration btw? It's a lot worse than what this intel system is showing.

Did you read what I said? FRAPS showed up on the AMD demo and it didn't on Intel's.
This is enough to tell you that it was not rendered on the IGP.

[Manicdan]

AMD hasn't "disabled" gpu acceleration on the intel system - try running the psychedelic browser test in your normal browser and you'll see that the performance is in single digits when there is no gpu acceleration.

http://ie.microsoft.com/testdrive/

There you go, see how Psychedelic Browsing runs on your current non-accelerated browser. Single digits?

a score of 5 dosnt seem to be what intel was getting. and many of us dont see the gpu working at all and still get great scores

[jimbo75]

a score of 5 dosnt seem to be what intel was getting. and many of us dont see the gpu working at all and still get great scores

Yes because the intel is using gpu acceleration, it's just being accelerated by a very weak gpu.

[The Coolest]

Just look this page. GPU vs CPU.
http://www.itwriting.com/blog/3003-f...d-enabled.html

jimbo75, have you seen the testing done by users here in this thread?
People with highend graphics cards and IGPs get very close numbers when using D2D acceleration.

[Manicdan]

ok i was lucky enough to just drop cpu speed using window power profiles
so at 960mhz with no turboing, i still get 1700+ scores, and cpu was never above 40% for more than one core.

i really have no idea how anyone can get a bad score on these benchmarks

[god_43]

i think microsoft rigged the tst to showcase ie9 in a better light. and amd marketing is just as sleazy as intels...its a sad day.


dont flame me lol!

[savantu]

3 pages later still no conclusion : Is the Intel result valid from AMDs test ? If not can we conclude the current GMA HD is as good if not better in 2D than the Ontario GPU ?

[kl0012]

3 pages later still no conclusion : Is the Intel result valid from AMDs test ? If not can we conclude the current GMA HD is as good if not better in 2D than the Ontario GPU ?

Definitely theirs test is not valid I dont know what they did, but I can notice they have a different version of IE9 (which, of cause is not an excuse for them).
I still want to test CoH to validate theirs 3D test, but downloading is paintfully slow so it will take another day.

BTW, tried to reduce GPU clock to 200MHz:

[-Boris-]

Since people don't get any results that makes any sense I think it's safe to say that this bench don't show anything. It can only be used to show that there is a difference between software and hardware acceleration.

[slaveondope]

They should just run Left 4 Dead which uses multiply threads and is high/moderately gpu intensive.
I could care less about web browsing apps.

Benches here = Borked I retract my earlier statement regarding a win for AMD.

[jimbo75]

The ie 9 ones are definitely just weird, the psychedelic one especially because it doesn't seem to matter what gpu you have or what the screen resolution is.

I think we can pretty safely assume that Zacate is twice as fast as an i5's graphics though (unless you believe 80 sp's are going to be slower on Zacate), so there isn't any reason why it wouldn't be performing twice as fast in benchmarks that actually show up gpu prowess.

[STEvil]

need some quake 3 benches and 3d01. lol

[jimbo75]

Apparently AMD are using a different version of ie 9 preview that is available from the public, (presumably the developer one). That pretty much renders all the browser tests invalid.

Couple of weird things i noticed about the Kitguru comparison video, they are running a different version of the Platform Preview than is available to the public. I just downloaded the latest one from M$ which is 1.9.7916.6000, the version they are using is 1.9.7930.16394.

- Fry on S|A

[duploxxx]

3 pages later still no conclusion : Is the Intel result valid from AMDs test ? If not can we conclude the current GMA HD is as good if not better in 2D than the Ontario GPU ?

oh thx one can start no better a day with a good lauch.

fyi, ontario is a max 18W TDP product so it goes against i3 330UM, i5 520UM etc with a 500mhz gpu, i don't think we need more info

some examples according to anandtech: an i5 540m with 766mhz GPU is on par with a AMD 790GX chipset which is an ati3300 which has 40SP @ 700mhz.
Ontario has 80SP on an unknown clock but you have Mobility Radeon HD 5430 80@550MHz with a TDP lower then 8W so you can start guessing (which you like to do all the time) and will provide you about 50% more performance then 790GX. so keep dreaming with your on par performance. It will be double, even SB will have a hard time in the 18W parts to counter that mid next year.

you can argue as much as you want on the benchmarks provided by AMD and INtel all the time, but thinking that GPU performance is on par with clarckdale makes you an intel fanboy daydreamer.