Trinity review

[piesquared]

AMD as in graphics, or AMD as in processors? APUs are still a ways away from powering "featured" gaming laptops without a dedicated GPU, and if you're getting a dedicated GPU might as well pair it with a superior Intel processor.

Many consumers would prefer the dedicated GPU for gaming and the superior APU for accelerating applications beyond what a standard, old fasioned, legacy CPU can offer.

 

[frozentundra123456]

Well here's a hint: how many AMD gaming laptops have been featured over the past several years?

Zero, unless you count AMD GPUs. I dont consider Llano a gaming laptop at all. It is a low to mid range laptop that has the capability to play some games at very limited details and resolutions.

 

[frozentundra123456]

And the $1000 price tag along with it.

If you want a true gaming laptop, you have to pay the price. However, Microcenter has an asus laptop with a quad Ivy bridge CPU and GT630M for 799.00. Still not really a gaming laptop as such, but a much better CPU than Llano or Trinity and better graphics performance as well.

 

[Red Hawk]

Many consumers would prefer the dedicated GPU for gaming and the superior APU for accelerating applications beyond what a standard, old fasioned, legacy CPU can offer.

What can you do with an APU that you can't do with a simple modern CPU + GPU combination? Aside from running at a lower TDP.

 

[AtenRa]

If you want a true gaming laptop, you have to pay the price. However, Microcenter has an asus laptop with a quad Ivy bridge CPU and GT630M for 799.00. Still not really a gaming laptop as such, but a much better CPU than Llano or Trinity and better graphics performance as well.

You know you can have a trinity + Discrete for higher performance.
I bet at the same price or lower than an Intel Laptop with Discrete GPU a Trinity + Discrete in CF will always be faster.

 

[Nemesis 1]

So now were talking battery power computers and AMDs big trump card for playable graphics on a battery operated APU is to XF for the 1 hour runtime and the win.

 

[AtenRa]

No, we are talking about how Trinity can be a winner in both Low Power performance and High Computing/Gaming.

 

[Red Hawk]

If you want a true gaming laptop, you have to pay the price. However, Microcenter has an asus laptop with a quad Ivy bridge CPU and GT630M for 799.00. Still not really a gaming laptop as such, but a much better CPU than Llano or Trinity and better graphics performance as well.

You can get a Llano A8-3500m laptop for under $500. It may not have the smoothest performance or be able to run every game at its best settings, but it will play just about anything out there. If you're a gamer on a budget, that's all you need. If you can pay more just get a desktop (or spend $500-$600 on a desktop and spend the remaining $200-$300 on a tablet or netbook for mobility).

 

[frozentundra123456]

You know you can have a trinity + Discrete for higher performance.
I bet at the same price or lower than an Intel Laptop with Discrete GPU a Trinity + Discrete in CF will always be faster.

Maybe, but at least for llano, crossfire was very hit and miss, rarely working properly. Besides if I were getting a laptop with a discrete card, I would not hamper myself with an amd CPU,

 

[FlanK3r]

No, we are talking about how Trinity can be a winner in both Low Power performance and High Computing/Gaming.

Its mainstream...good CPU with great iGPU, if Im talking about A10-5800K 🙂. Of course, i5-2500K/3570K is better in pure performance, but still is A10-5800K nice quadcore with good performance. or do u think, example i7-920 is crap today because in graphs is more lower than Ivy Bridges? Graphs are not all...

 

[boxleitnerb]

We do not want CF (or SLI for that matter) mainstream as long as they are based on AFR. The average user cannot and should not burdened with their shortcomings. Especially in notebooks the low fps scream for microstuttering which becomes even worse when two GPUs with different speeds are connected.

 

[FlanK3r]

5800K can be easy overclock over 5 GHz for benchmarking, so...There is point for Crossfire (with LN2 more) with mainstream GPUs.

 

[boxleitnerb]

5800K can be easy overclock over 5 GHz for benchmarking, so...There is point for Crossfire (with LN2 more) with mainstream GPUs.

What does that even mean? Clocking the CPU part doesn't make the GPUs magically faster. Also the dual graphics config will be mostly present in notebooks. No OC there. One fast GPU is almost always preferable over two slower ones. Do you expect the average user to deal with stuttering, graphics glitches, no scaling/negative scaling? Leave this tech to the more knowledgeable people who know how to deal with it.

 

[ShintaiDK]

5800K can be easy overclock over 5 GHz for benchmarking, so...There is point for Crossfire (with LN2 more) with mainstream GPUs.

Can you show this? Or is it just words you make up?

 

[blckgrffn]

5800K can be easy overclock over 5 GHz for benchmarking, so...There is point for Crossfire (with LN2 more) with mainstream GPUs.

The IGP portion will still be crippled by measily bandwidth no matter the speed of the CPU, so if you are doing performance runs you are going to be just using an add-in card with some balls anyway.

Crossfire is a box that gets checked to say "we can do it!" Not that it is useful.

 

[KompuKare]

We do not want CF (or SLI for that matter) mainstream as long as they are based on AFR. The average user cannot and should not burdened with their shortcomings. Especially in notebooks the low fps scream for microstuttering which becomes even worse when two GPUs with different speeds are connected.

Is there actually an alternative rendering approach instead of AFR? And if there is, would it require game engines and APIs (DirectX/OpenGL) to be re-written?

I still think that when used with a dGPU it makes more sense to use the iGPU for physics, AA or something. And if for instance i7-3770K could use its HD4000 for something when paired with a dGPU we wouldn't have these constant forum posts about the wasted die space in IB. That at least would be worth something...

While Alternate Frame Rendering might be a poor thing in PCs due their variability, in next-gen consoles - where the capabilities of CPU and GPU are fully known and predicable - having two GPUs makes a lot more sense. One problem might be that due to the closed nature of GPU drivers, getting 'to the metal' of both GPUs and keeping threads etc. in sync is harder than with multiple CPUs.

 

[pelov]

If you want a true gaming laptop, you have to pay the price. However, Microcenter has an asus laptop with a quad Ivy bridge CPU and GT630M for 799.00. Still not really a gaming laptop as such, but a much better CPU than Llano or Trinity and better graphics performance as well.

The GT630m is just a rebrand 540m and though it's good at x768 it just doesn't have enough horsepower for higher res at higher settings. The 540m isn't Kepler but rather a hot and power hungry Fermi. It's either a GF108 or GF106 and either 28nm or 40nm. The rebrands with horrible naming schemes continue... 🙁

There are some great deals on GT640m + Ivy i7's every now and then, though. HP has had a 30%+ coupon for those laptops that brought it under $1000. If you're going to buy an nVidia discrete mobile GPU I'd *much* rather buy a Kepler and that means at least a 640m as everything under that is Fermi and really not good for laptops.

 

[Martimus]

Is there actually an alternative rendering approach instead of AFR? And if there is, would it require game engines and APIs (DirectX/OpenGL) to be re-written?

I still think that when used with a dGPU it makes more sense to use the iGPU for physics, AA or something. And if for instance i7-3770K could use its HD4000 for something when paired with a dGPU we wouldn't have these constant forum posts about the wasted die space in IB. That at least would be worth something...

While Alternate Frame Rendering might be a poor thing in PCs due their variability, in next-gen consoles - where the capabilities of CPU and GPU are fully known and predicable - having two GPUs makes a lot more sense. One problem might be that due to the closed nature of GPU drivers, getting 'to the metal' of both GPUs and keeping threads etc. in sync is harder than with multiple CPUs.

There are a lot of different methods to use multiple GPUs. 3DFX used a tile based SLI where they broke the screen into multiple tiles and each card would render one of them. While this avoids microstutter, it does not give nearly as good increases in raw FPS, like AFR does.

 

[ShintaiDK]

There is currently no alternative to AFR, since all alternatives will impact scaling performance very heavily. And hence make SLI/Crossfire utterly useless.

 

[frozentundra123456]

The GT630m is just a rebrand 540m and though it's good at x768 it just doesn't have enough horsepower for higher res at higher settings. The 540m isn't Kepler but rather a hot and power hungry Fermi. It's either a GF108 or GF106 and either 28nm or 40nm. The rebrands with horrible naming schemes continue... 🙁

There are some great deals on GT640m + Ivy i7's every now and then, though. HP has had a 30%+ coupon for those laptops that brought it under $1000. If you're going to buy an nVidia discrete mobile GPU I'd *much* rather buy a Kepler and that means at least a 640m as everything under that is Fermi and really not good for laptops.

I know the GT630 is just a rebadged GT540, but according to Anand's review, if I recall correctly, it beat the A10 in every benchmark, sometimes by significant margins. I did even acknowledge that the laptop I was referencing was not a high end gaming machine. I am sure if one watched closely, better deals will become available. And I agree I would also much prefer the newer GT640m.

 

[boxleitnerb]

Is there actually an alternative rendering approach instead of AFR? And if there is, would it require game engines and APIs (DirectX/OpenGL) to be re-written?

I still think that when used with a dGPU it makes more sense to use the iGPU for physics, AA or something. And if for instance i7-3770K could use its HD4000 for something when paired with a dGPU we wouldn't have these constant forum posts about the wasted die space in IB. That at least would be worth something...

While Alternate Frame Rendering might be a poor thing in PCs due their variability, in next-gen consoles - where the capabilities of CPU and GPU are fully known and predicable - having two GPUs makes a lot more sense. One problem might be that due to the closed nature of GPU drivers, getting 'to the metal' of both GPUs and keeping threads etc. in sync is harder than with multiple CPUs.

I don't think so. At least no method that provides the same efficiency as AFR. In a couple of years with massive bandwidth due to inteprosers you will probably able to put several dies on one PCB near each other and they will act as one single GPU.

APU+dGPU could provide some benefits, though. You mentioned physics, which is a very good example. Or some kind of post processing.

 

[KompuKare]

APU+dGPU could provide some benefits, though. You mentioned physics, which is a very good example. Or some kind of post processing.

The real problem is getting a company to develop drivers for something like that. For physics OpenCL might eventually get there but post processing AA type stuff would require total co-operation between the APU and the dGPU so only AMD can really do it. And AMD and writing software is usually a slow combination. And AMD, Nvidia and Intel coming together to develop a post processing AA mode which could use Intel HD4000 is not very likely. OpenCL Physics seems the most likely thing.

 

[boxleitnerb]

Yup, that about sums it up. In the PS4 it might be different though - if the rumored PS4 specs (APU+dGPU) are in fact true.

 

[Martimus]

There is currently no alternative to AFR, since all alternatives will impact scaling performance very heavily. And hence make SLI/Crossfire utterly useless.

That is far from accurate. AFR is used because it has the best per framerate scaling. Usinf tiled or other methods might reduce scaling in certain situations (when one part of tbe screen is harder to render than another part) but it wont reduce scaling down to zero in any scenario. It actually scales better than AFR in games that don't support SLI/Crossfire since it doesn't require specific support to work.

 

[ShintaiDK]

Lets see...

Scissors? No.
Tiles? No.

Any suggestions?

The original SLI scissor method went out because the scene buildup changed. It simply doesnt work with advanced graphics. Same with tiles.

Thats also why we use the essentially worst possible method, AFR. So each GPU handles the entire frame instead of the splitup and essentially double workload.