Kabini Rumors

[mikk]

It is probably rather easy to optimize the driver profile for maximizing 3DMark11 scores (or any canned demo for that matter).

Because I'm testing Intel drivers regularly I can say there never were really untypically improvements in 3dmark Vantage or 3dmark11. When Intel added driver improvements primarily games improved not 3dmark. From launch driver to the current 15.31 3dmark11 and Vantage improved only by 5% on Ivy Bridge. This isn't that special, there are lots of games which improved much more. Of course Intel suffered (or still suffer) from a couple of performance issues in games. The biggest issue was their annoying low clock bug with 15.26 and 15.28 Ivy bridge drivers. I'm glad they solved this in 15.31. In several games with this low clock issue I saw roughly a doubling in performance.

For 4600M is so because of the lower GPU clock and memory BW bottleneck. It's a moot point since Richland is what will be competing against HD4600 and not Trinity. Also price will play major role so if you are to have more fps in games and less in 3dmark(with Richland) with noticeably lower price for laptop, what will you choose?

Trinity isn't cheaper, I guess same applies to Richland.

 

[Exophase]

It is probably rather easy to optimize the driver profile for maximizing 3DMark11 scores (or any canned demo for that matter).

Another possibility (no idea how true it is, just throwing it out there) is that 3DMark has lower CPU demands than the average game and CPU + IGPs with any kind of load balancing that let the GPU run faster when the CPUs run slower can benefit. This is especially possible for the lowest TDP ULV Intel parts where the turbo range for the GPU is huge. Should not apply to Temash though..

Also less potential for being CPU limited in general.

 

[monstercameron]

another review...funny in the comments, some one pointed out

Surface Pro with the Core i5/HD4000 and 42 Wh battery life is 4-5 hours.
Surface RT ARM A9 and 32-watt-hour battery life is 9-10 hours.
This goods with a 30-watt-hour battery life is 3 hours? Also the main power consumption?

http://translate.google.com/translate?hl=en&sl=zh-CN&tl=it&u=http%3A%2F%2Fwww.expreview.com%2F25482.html

seems to be ripped from netbookcheck.com

 

[AtenRa]

3dmark11 Entry only. Here is a proper comparison:

http://www.legitreviews.com/article/2047/5/

=+102% for the A10-5800k

Average in games depending on test +60-80%. Similar for the A10-4600M. There is a 50% advantage in 3dmark11 if not more and in games it is 25-35%. AMD doing better in 3dmark11 than in games.

102% is for the GPU score not the overall 3D Mark 11 score. Overall score is 89% faster than 3770K, close to the actual gaming difference. Games use both the CPU and GPU.

 

[mikk]

Another possibility (no idea how true it is, just throwing it out there) is that 3DMark has lower CPU demands than the average game and CPU + IGPs with any kind of load balancing that let the GPU run faster when the CPUs run slower can benefit. This is especially possible for the lowest TDP ULV Intel parts where the turbo range for the GPU is huge. Should not apply to Temash though..

Also less potential for being CPU limited in general.

No this isn't the reason. GPU always clocks at max on a desktop model.

102% is for the GPU score not the overall 3D Mark 11 score. Overall score is 89% faster than 3770K, close to the actual gaming difference. Games use both the CPU and GPU.

GPU score is exactly what we need and want because the overall score is slightly influenced from the CPU and Intels Haswell graphics slides also were based on the GPU score and not overall score. 89% is much higher than gaming performance too.

 

[inf64]

89% is much higher than gaming performance too.

Nope it is not much higher, I gave you hard numbers from hardware canucks review on previous page.

Deus Ex - 47% faster than HD4000
Dirt3 - 63% faster than HD4000
Skyrim(DX9) - 84% faster than HD4000
StreetF.4 - 72% faster than HD4000
Torchlight2(DX9) - 333% faster than HD4000

With Torchlight average is 120% faster, without this outlier it's 66.5% and that's with 1600Mhz memory. Add 10% as their results for 1866Mhz memory show us and you end up at 83% (1.665x1.1=1.83) which is what you are disagreeing with.

 

[Exophase]

No this isn't the reason. GPU always clocks at max on a desktop model.

Why would Intel specify i7-3770K's GPU clock as 650MHz base and 1.15GHz max if it's always max?

Anyway I didn't know we were only talking about desktops, I thought this was referring to ULV i7s in particular :/

 

[Enigmoid]

Temash Review

http://www.notebookcheck.net/Review-AMD-A6-1450-APU-Temash.92264.0.html

 

[mikk]

Nope it is not much higher, I gave you hard numbers from hardware canucks review on previous page.

With Torchlight average is 120% faster, without this outlier it's 66.5% and that's with 1600Mhz memory. Add 10% as their results for 1866Mhz memory show us and you end up at 83% (1.665x1.1=1.83) which is what you are disagreeing with.

Here it is 62%: http://www.computerbase.de/artikel/grafikkarten/2012/alle-desktop-trinity-im-gpu-vergleichstest/4/

Why would Intel specify i7-3770K's GPU clock as 650MHz base and 1.15GHz max if it's always max?

Anyway I didn't know we were only talking about desktops, I thought this was referring to ULV i7s in particular :/

Ask Intel. Base clock is important for idle. Some Boards didn't enable the idle state.

 

[inf64]

Here it is 62%: http://www.computerbase.de/artikel/grafikkarten/2012/alle-desktop-trinity-im-gpu-vergleichstest/4/

Same link 3rd chart (with basic AA and AF). Put a strain on 7660D and HD4000 and Trinity's advantage raises to nearly 80%. Also check the page with memory scaling. Trinity gains +10% with 1866Mhz memory and scales further with more dram clock. This will increase the advantage to close to 100% if you pick up faster modules for your iGPU rig. HD4000 is completely outclassed I'm afraid. And we didn't touch the OCing part of the GPU yet (Richland can run iGPU in games at higher clock and during longer periods of time Vs Trinity).

 

[mikk]

Same link 3rd chart (with basic AA and AF).

3dmark in default doesn't use AA. If you want to compare games with AA, compare 3dmark11 with MSAA. I would expect an even bigger gap then. The gap in 3dmark11 is bigger than in games as you can see.

Trinity gains +10% with 1866Mhz memory and scales further with more dram clock.

It doesn't gain anything because Trinity has been tested with DDR3-1866.

 

[inf64]

It doesn't gain anything because Trinity has been tested with DDR3-1866.

Can you show me where does they state which modules they used for that average comparison? On the test page they list everything from 1333 to 2133MHz and on the average page they mention zero information on memory used.

 

[mikk]

Can you show me where does they state which modules they used for that average comparison? On the test page they list everything from 1333 to 2133MHz and on the average page they mention zero informationm on memory used.

Have a look to their DDR3 scaling comparison and you should notice that they used DDR3-1866. FPS exactly the same.

 

[monstercameron]

resurrection...pcper kabini over view video
https://www.youtube.com/watch?v=j5KKlxFml7Q

 

[Rvenger]

http://anandtech.com/show/6974/amd-kabini-review

 

[monstercameron]

it seems everyone is releasing at the same time...

 

[SPBHM]

any review using the 25w version?

 

[frozentundra123456]

Might be nice in a tablet, still sort of underpowered for a full size laptop. Dont any of the parts have turbo on either cpu or gpu? Seems like that would be the ideal way to allocate the power budget between cpu/gpu depending on workload.

 

[monstercameron]

they are distancing themselves from netbook...

 

[monstercameron]

the a6-1450 is 8W tdp 128GCN @300-400MHz, ddr3L-1066
the a4-1250 is 9W tdp[in both temash and kabini versions?], 128GCN @300, ddr3L-1333
the a4-1200 is 3.9W, 128GCN @225MHz, ddr3L-1066

 

[IntelUser2000]

http://hothardware.com/Reviews/AMD-2013-ASeries-Kabini-and-Temash-Mobile-APUs/?page=3

So the earlier speculations of the quad core Temash having a TDP of 8-9W was spot on.

A6-1450: 4 cores/1.0GHz/1.4GHz Boost

The A4-1200 is at 3.9W with 2 cores and 1GHz frequency.

 

[Remobz]

Can someone sum up this thread so far please?

 

[monstercameron]

Can someone sum up this thread so far please?

weeks of speculation and tidbits of info...the truth is released today and thats it.

the quanta tablet is using ddr3U and an msata drive in a fanless design.
http://www.slashgear.com/amd-temash-official-ipad-smoothness-and-x86-grunt-for-tablets-and-hybrids-22283195/

 

[raghu78]

http://hothardware.com/Reviews/AMD-2013-ASeries-Kabini-and-Temash-Mobile-APUs/?page=3

So the earlier speculations of the quad core Temash having a TDP of 8-9W was spot on.

A6-1450: 4 cores/1.0GHz/1.4GHz Boost

yeah. the single thread performance of Temash A6-1450 at 1.4 Ghz matches Brazos at 1.7 Ghz. multithread performance is much better. so effectively a much better user experience than brazos notebook in a Temash tablet. thats very good progress. AMD has a 6 month headstart over Intel Baytrail. They should make the most of it. AMD's Beema with Jaguar+ cores needs to be a strong update in early - mid 2014 to remain competitive with Baytrail. Hopefully by moving to GF- ST Micro 28nm FDSOI they can cut power consumption significantly.

 

[georgec84]

Give 'em hell, AMD!