[no-X]

Nice performance increase in synthetic benchmarks, but the only gaming test shows exactly opposite results. IB's GPUs should be at least 50% faster than SB's to be competitive with Llano.

[DavidC]

Originally Posted by no-X

Nice performance increase in synthetic benchmarks, but the only gaming test shows exactly opposite results. IB's GPUs should be at least 50% faster than SB's to be competitive with Llano.

Totally opposite from TKK's response. Still, performs almost 15% better than HD 3000.

The extra ~50% clock would do what's required. Besides, even the CPU is seriously under 2600K's clocks.

[iwod]

Not to mention Intel has never been good at GPU drivers, and therefore optimizing for ANY games related paths, I just hope they put more resources into it.

But this is finally some good news for IB's GPU. I was disappointed when i heard its performance were only 30% faster then SB. Now IB, a VERY good CPU, with 22nm Tri Gate, Low Power, and finally ( or hopefully ) a decent enough GPU, should significantly rise the bar of the lowest GPU power on PC.

And i could finally get a Macbook Air

[no-X]

Originally Posted by DavidC

The extra ~50% clock would do what's required. Besides, even the CPU is seriously under 2600K's clocks.

Even +50% GPU clock and +50% CPU clock doesn't translate to +50% gaming performance, because BW stays at the same level.

[ninelven]

Well, that depends on if you are primarily bandwidth limited or not.

[Psycho]

50% is far from enough to reach a8-3800

[TKK]

Originally Posted by DavidC

Totally opposite from TKK's response. Still, performs almost 15% better than HD 3000.

Well, I based my opinion on the fact that it beats SB despite less than half the base clock. However, if it ran at full turbo most of the time, the results are not that impressive, I must admit. I forgot that with such a low CPU clock, GPU turbo might have kicked in all the time due to more TDP headroom, so maybe things don't look as grim for AMD as I originally thought.

[DavidC]

50% is far from enough to reach a8-3800

That's enough to do it in the Street Fighter benchmark linked.

That said, I don't think it'll be much faster than Llano if it is. The aim seems to be "just Llano", nothing more.

Apparently there are two chips out there, 2.0GHz and 2.2GHz both with 400-900MHz GPU speeds.

[mczak]

Originally Posted by TKK

Well, I based my opinion on the fact that it beats SB despite less than half the base clock. However, if it ran at full turbo most of the time, the results are not that impressive, I must admit. I forgot that with such a low CPU clock, GPU turbo might have kicked in all the time due to more TDP headroom, so maybe things don't look as grim for AMD as I originally thought.

I believe all intel igps pretty much always use max turbo igp clock when running 3d apps. You simply don't need 4 3.5Ghz cores to keep that IGP busy.
I don't think IVB will be able to catch (fully-enabled) Llano - intel apparently saying IGP clock should be a bit lower than on SNB, and a 50% increase is not enough to really threaten Llano except maybe in few select titles.
BUT I don't think Trinity will improve that much over Llano (not least because it seems it won't be more tightly integrated with the cpu with the cpu having no L3 cache hence it will likely be even more bandwidth limited than Llano), which means intel will get a lot closer.

[Paran]

Originally Posted by DavidC

According to this site, Ivy Bridge's GPU is at 400-900MHz(latter speed is Turbo). Compared to 850-1350MHz on Sandy Bridge.

http://translate.google.ca/translate...26prmd%3Dimvns

I know it. You have realized this is an Engineering Model with low clock speed? Or do you think a CPU base clock of 2,2 Ghz means final performance?

[fellix]

AMD Trinity Detailed Further, Compatible with A75 Chipset

The integrated memory controller will get an overhaul, too. Unlike with K10-based processors that have two independent 64-bit wide memory interfaces that can be configured to work ganged or unganged, Trinity will have a single 128-bit memory interface, the controller will support dual-channel DDR3-2133 MHz memory standard, with DRAM voltages of under 1.5V. Trinity will include a 24-lane PCI-Express root complex, it supports 2-way multi-GPU configurations.

Moving on to the integrated GPU component, AMD promises a 30% performance improvement over Llano's iGPU. The GPU component is DirectX 11 compliant, and features UVD 3 hardware HD video acceleration, with SAMU and native VCE. Featuring AMD Eyefinity technology, this integrated GPU will support up to three displays without needing a discrete graphics card. Eyefinity can be used to step up productivity.

[Alexko]

Originally Posted by fellix

AMD Trinity Detailed Further, Compatible with A75 Chipset

The Eyefinity bit is great news, it means you can have an Eyefinity setup driven by a powerful discrete card (e.g. HD 7800/7900) but completely turn it off while not gaming, and still retain your three displays for productivity.

That's actually a pretty big selling point for me, enough to steer me towards Trinity rather than Ivy Bridge, provided its CPU proves to be decent.

[fehu]

only 20% over llano that wasn't that fast on his own, and a less advanced memory controller?
we must keep waiting?

[mczak]

Originally Posted by fehu

a less advanced memory controller?

I wouldn't call that less advanced. K10 could do ganged and unganged, but it only had chip select interleaving (aside from node interleaving like all others but I'm not looking at that as it will only apply to multichip configurations). Ganged was AFAIR also limited because it needed the dimms to be of the same size.
BD, Trinity (I just assume it's the same), and even Llano got rid of unganged, but they are more flexible otherwise - not only can they do chip select interleaving, but they can do channel interleaving, hence no need for unganged really, and they don't require dimms to be of the same size in different channels in this mode neither.
Unganged wasn't really much of a win over ganged in K10 anyway, some apps were slightly faster some slightly slower, unganged was more of a bandaid because ganged was not flexible enough.
So getting rid of ganged vs. unganged should be a good thing - I think the interleaving now used is probably also much closer to what intel is doing since a decade or so...

[fehu]

for sure you are right, but the unganged mode was promoted as a so big advancement that i can't believe that they putted it apart

[rpg.314]

Will it have PCIe3?

[fehu]

Originally Posted by rpg.314

Will it have PCIe3?

apparently no

[fehu]

semiaccurate says that GF has problems with 28nm and amd will skip new generation bobcat :S

[AnarchX]

Performance slides inside the video: http://www.donanimhaber.com/islemci/...da-her-sey.htm

Desktop APUs:

Trinity A8: 715GFLOPs
Llano A8: 415GFLOPs

Trinity A8: 4500 3DMark Vantage
Llano A8: 3335 3DMark Vantage
- probably GPU scores, both @ DDR3-1866

[mczak]

Originally Posted by AnarchX

Performance slides inside the video: http://www.donanimhaber.com/islemci/...da-her-sey.htm

Desktop APUs:

Trinity A8: 715GFLOPs
Llano A8: 415GFLOPs

Trinity A8: 4500 3DMark Vantage
Llano A8: 3335 3DMark Vantage
- probably GPU scores, both @ DDR3-1866

17 min video for 2 slides (in case you're wondering the performance slide is at 10 minutes)?
If that slide is real AMD can't even count the gflops on Llano... as the gpu is good for 480 gflops in the a8 on the desktop.
In any case, seems to confirm earlier predictions. A solid increase but nothing earth-shattering (and gflops increasing more than performance - no matter the gpu arch this is almost certainly due to memory bandwidth not going up).

[Lightman]

Originally Posted by mczak

17 min video for 2 slides (in case you're wondering the performance slide is at 10 minutes)?
If that slide is real AMD can't even count the gflops on Llano... as the gpu is good for 480 gflops in the a8 on the desktop.
In any case, seems to confirm earlier predictions. A solid increase but nothing earth-shattering (and gflops increasing more than performance - no matter the gpu arch this is almost certainly due to memory bandwidth not going up).

But with Trinity you can get more memory bandwidth. IMC will officially support DDR3-2133, which should be enough to keep these extra FLOPS somewhat happy.

[AlexV]

Originally Posted by Lightman

But with Trinity you can get more memory bandwidth. IMC will officially support DDR3-2133, which should be enough to keep these extra FLOPS somewhat happy.

Not nearly enough even if one creatively interprets Amdhal's opinion (yeah, perhaps not the best rule to take into account these days, but it holds merit!). They should get a better MC though (this is one of the areas where BD did improve siginifcantly vs K8L), which means reasonable increases on that front even without ludicrously priced DRAM (2133 is quite expensive when considered in the context of its slower clocked siblings).

[denev2004]

Actually I was wondering why AMD do not think of triple channel. I habour its gonna be useful.

[AlexV]

Originally Posted by denev2004

Actually I was wondering why AMD do not think of triple channel. I habour its gonna be useful.

It's also expensive, and unrealistic for their targets (laptops, cheapie desktops etc.).

[ToTTenTranz]

Originally Posted by AlexV

It's also expensive, and unrealistic for their targets (laptops, cheapie desktops etc.).

It's none of that if the purpose is to make an APU that really discards a mid-high end discrete GPU.

Unless you think adding a third memory channel is more expensive than an entire MXM card...