[ToTTenTranz]

I shall patiently wait for the hat-eating to comence.

[Paran]

Over 50% to which GPU competition exactly? HD3000?

[nAo]

Originally Posted by Paran

Over 50% to which GPU competition exactly? HD3000?

50% over IVB in 3DMark Vantage, where they extrapolate IVB performance as 30% over SNB.

[CarstenS]

Interesting that they extrapolated from only the second fastest ULV ci5 (2537M). There's a 2557M, which interestingly enough, provides a maximum graphics turbo frequency exactly.... 30% higher than 2537M (900 vs. 1200 MHz). I find it hard to believe, that Intel would not increase graphics performance at all for IVB.

So, they are basically comparing this-gen Intel-CPU vs. next-gen AMD-APU.

Additionally, they are using an A6-Trinity for this 17-Watt ULV comparison. Isn't A6 a triple-core APU right now? In the footnotes, there's an A10 Low-Voltage APU in the 25 Watt TDP marked as reaching 3600 Vantage points.

[itsmydamnation]

Originally Posted by CarstenS

Interesting that they extrapolated from only the second fastest ULV ci5 (2537M). There's a 2557M, which interestingly enough, provides a maximum graphics turbo frequency exactly.... 30% higher than 2537M (900 vs. 1200 MHz). I find it hard to believe, that Intel would not increase graphics performance at all for IVB.

So, they are basically comparing this-gen Intel-CPU vs. next-gen AMD-APU.

Additionally, they are using an A6-Trinity for this 17-Watt ULV comparison. Isn't A6 a triple-core APU right now? In the footnotes, there's an A10 Low-Voltage APU in the 25 Watt TDP marked as reaching 3600 Vantage points.

it depends, just because its a turbo value doesn't mean it will actually ever hit it within the target TDP or for long enough to make any real difference. that said i cant find a good review that covers performance and power usage of the 2557.

[CarstenS]

IIRC the game tests for Vantage are not heavily threaded.

[itsmydamnation]

aren't we talking about GPU here?

[ToTTenTranz]

Originally Posted by CarstenS

Interesting that they extrapolated from only the second fastest ULV ci5 (2537M). There's a 2557M, which interestingly enough, provides a maximum graphics turbo frequency exactly.... 30% higher than 2537M (900 vs. 1200 MHz). I find it hard to believe, that Intel would not increase graphics performance at all for IVB.

So, they are basically comparing this-gen Intel-CPU vs. next-gen AMD-APU.

Additionally, they are using an A6-Trinity for this 17-Watt ULV comparison. Isn't A6 a triple-core APU right now? In the footnotes, there's an A10 Low-Voltage APU in the 25 Watt TDP marked as reaching 3600 Vantage points.

Where are you getting all those specs from?
Mobile Llano A6 are all quad-core. And I don't think there can be odd numbers of cores in the bulldozer/piledriver architecture.

Furthermore, has anyone seen socket info about desktop trinity?
Is it really a new socket for everyone, as rumoured?

[CarstenS]

Originally Posted by itsmydamnation

aren't we talking about GPU here?

Maybe I'm confusing things, but do GPU and CPU not share a chip-wide power budget?
edit: This presentation supports my recollection. http://www.hotchips.org/archives/hc2...otem-Intel.pdf

Originally Posted by ToTTenTranz

Where are you getting all those specs from?

From the footnotes your pictures refers to. And from Intels ark nee processorfinder website.

Originally Posted by ToTTenTranz

Mobile Llano A6 are all quad-core. And I don't think there can be odd numbers of cores in the bulldozer/piledriver architecture.

That's why I wrote "Isn't A6 a triple-core APU right now?", but you're probably right, uneven number of cores do seem unlikely.

[Lightman]

Small correction: current A6 APU's come in 3 and 4 core flavours.
A6 36xx are Quad-Core and A6 35xx are Tri-Cores.

Anyway as already noted by ToTTenTranz Trinity based APU's should only be Dual or Quad cores due to modular nature of Piledriver cores.

I'm eagerly awaiting new info and release of these into the market.
I'm both looking for laptop solution as well as unnecessary upgrade from A6-3650 to Trinity based APU for my HTPC .

[CarstenS]

Ok, then please disregard that part of my posting.

[itsmydamnation]

Originally Posted by CarstenS

Maybe I'm confusing things, but do GPU and CPU not share a chip-wide power budget?

yes, but you miss my point. my point is can the GPU hit max turbo in under 17watt TDP regardless of what the CPU is doing.

[ToTTenTranz]

Originally Posted by Lightman

I'm eagerly awaiting new info and release of these into the market.
I'm both looking for laptop solution as well as unnecessary upgrade from A6-3650 to Trinity based APU for my HTPC .

You do know there's a good chance you'll have to change motherboard as well, right?
I haven't seen any official confirmation, but word is that Trinity will use socket FM2 and won't be compatible with FM1 motherboards..

Originally Posted by itsmydamnation

yes, but you miss my point. my point is can the GPU hit max turbo in under 17watt TDP regardless of what the CPU is doing.

I don't get your question.
If the 17W Trinity has a max clock value for CPU turbo and GPU turbo, then of course this max clock can be reached on either the CPU or GPU.

Are you asking if these max turbo values are going to be the same between the 17W and the 35W parts? Of course not..

[CarstenS]

Originally Posted by itsmydamnation

yes, but you miss my point. my point is can the GPU hit max turbo in under 17watt TDP regardless of what the CPU is doing.

Let me put it this way. The TDP for the whole package is 17 watts, shared by CPU and GPU cores as outlined in the PDF I've linked. The 1.2 GHz GPU is specced in the 17 watt part I mentioned, so it should be able to reach higher than minimum values when - again, IIRC - 2-3 out of four threads in 3DMark Vantages Game Tests are idling, since there is a bigger portion of the TDP for GPU to consume.

The same obviously should be true of Trinity.

[fehu]

http://www.xbitlabs.com/news/other/d...s_for_AMD.html

[Lightman]

Originally Posted by ToTTenTranz

You do know there's a good chance you'll have to change motherboard as well, right?
I haven't seen any official confirmation, but word is that Trinity will use socket FM2 and won't be compatible with FM1 motherboards..

....

Yes, I do .
I'm fine with that. That's also the reason why I've picked cheap A55 FM1 board and not any higher end solution.

I like new technologies and so far I've had no experience of Bulldozer architecture because it was not good enough for my needs. So I want Trinity for both new CPU and iGPU tech.

[jimbo75]

Originally Posted by CarstenS

Interesting that they extrapolated from only the second fastest ULV ci5 (2537M). There's a 2557M, which interestingly enough, provides a maximum graphics turbo frequency exactly.... 30% higher than 2537M (900 vs. 1200 MHz). I find it hard to believe, that Intel would not increase graphics performance at all for IVB.

There is actually a 17W i7 2677M as well - http://ark.intel.com/products/54617

It scores 8.81 fps in cinebench opengl - http://laptopreviewshop.com/asus-zen...iew-video.html

IVB has been tested scoring 12.17 fps - http://ultrabooknews.com/2012/01/29/...lthy-ces-test/

So that's what, 38% faster than the old model in this test which I guess will be as close to best case as it will get?

[DavidC]

Originally Posted by itsmydamnation

it depends, just because its a turbo value doesn't mean it will actually ever hit it within the target TDP or for long enough to make any real difference. that said i cant find a good review that covers performance and power usage of the 2557.

http://www.notebookcheck.net/Intel-H...0.37948.0.html

The lowest i5-2557M score is 1414, which is 22% higher than the AMD footnote number. It's also interesting the 1158 score they quoted exists in Notebookcheck database.

Then there are much higher performing ones like the Macbook Air's 2557M getting 1712 points and UX21's 2677M getting 1661. That's over 40% higher than the footnote number.

So that's what, 38% faster than the old model in this test which I guess will be as close to best case as it will get?

http://www.notebookcheck.net/AMD-Rad...G.54675.0.html

The 35W Llano is getting 15 points in the same benchmark, meaning if 17W Trinity is equal to it, is only 25% faster than that Ivy Bridge.

[jimbo75]

Originally Posted by DavidC

http://www.notebookcheck.net/AMD-Rad...G.54675.0.html

The 35W Llano is getting 15 points in the same benchmark, meaning if 17W Trinity is equal to it, is only 25% faster than that Ivy Bridge.

According to this the A8-3500M scores 23.33 -



http://hothardware.com/Reviews/AMD-F...Review/?page=8

Even the 320 SP's A6 scores over 20 fps - http://www.youtube.com/watch?v=M88-YsVr8xQ

I think this 20 fps mark will be nearer what the 17W Trinity will score.

[Paran]

Originally Posted by jimbo75

There is actually a 17W i7 2677M as well - http://ark.intel.com/products/54617

It scores 8.81 fps in cinebench opengl - http://laptopreviewshop.com/asus-zen...iew-video.html

IVB has been tested scoring 12.17 fps - http://ultrabooknews.com/2012/01/29/...lthy-ces-test/

So that's what, 38% faster than the old model in this test which I guess will be as close to best case as it will get?

Maybe the OpenGL performance is flawed due to bad drivers or others reasons. The more important d3d performance should be much better. If it was an Ultrabook possibly it was an ES model with lower GPU clock. The Ultrabook CPUs are a little bit behind schedule wise.

[jimbo75]

Well that was just the 1.7 GHz chip and there will be a 2GHz one so it will rise by a little more. I think what is going on is intel has benched IVB vs older chips in order to get the 60% figure that was talked about. It'll be closer to 50% best case, and generally 30%-40% faster in my opinion.

Even if this 17W Ivy Bridge hits 13 fps, Trinity @17W will be a good 50% faster at ~20 fps. Again I believe this is also as good as it will get in a comparison and Trinity will be further ahead in actual gaming.

[Paran]

I meant GPU clocks. And again, OpenGL benchmarks could differ to d3d Benchmarks. Intel is weak in OpenGL. AMD supports OpenGL 4.2 whereas Intel just 3.1.

[ToTTenTranz]

Originally Posted by jimbo75

Well that was just the 1.7 GHz chip and there will be a 2GHz one so it will rise by a little more. I think what is going on is intel has benched IVB vs older chips in order to get the 60% figure that was talked about. It'll be closer to 50% best case, and generally 30%-40% faster in my opinion.

Even if this 17W Ivy Bridge hits 13 fps, Trinity @17W will be a good 50% faster at ~20 fps. Again I believe this is also as good as it will get in a comparison and Trinity will be further ahead in actual gaming.

Not to mention the fact that the top-performer 17W Trinity will probably cost like 1/3rd of the top-performer 18W Ivy Bridge.

[Paran]

17W Ivy Bridge.

[CarstenS]

Intel actually promises an almost 3x increase for "Vantage Performance" performance compared to a somehow strange Ci7 2600k with HD2000 graphics - at least for desktops:
http://www.xbitlabs.com/news/cpu/dis..._Partners.html

TBH, I don't know how much the clock targets for IVB-graphics differ from desktop to mobile to ULV, but this might give a notion for the gains we can expect from an architectural overhaul.