Video card for AoE 3

[imported_g33k]

Originally posted by: chinkgai
aoeIII = :thumbsup:

:thumbsup:

Originally posted by: coomar
i'm fairly certain a 6800gt will run a0e3 maxed out on the nf3/nf4 platform

I hope so, but I really doubt it. Besides HDR, 16x12 with all settings max would be great on a 6800gt. But I'm thinking a 7800 would be required to get there.

 

[VIAN]

I don't think you should make any decisions until the game comes out and some benches pop up.

 

[CU]

Originally posted by: Gamingphreek

I noticed "almost" but you said that (memory bandwidth) was the reason it has an edge on the on the 6800nu which is not true. The reason is because it has just alittle more fillrate. Also note you can not just look at number of pipes or gpu speed. A gpu with 48 pipes running at 200mhz can outrun a card with 8 pipes running at 1000mhz. Same goes for memory bandwidth and ram speed.

Ummm no it cant. How do you plan on keeping all those pipes filled. Why do you think that ATI and Nvidia are going to unified shaders?

You are correct. Raw fillrate, and clockspeed matter more in RTS. Bandwidth is more a FPS aspect.

-Kevin

Actually it can, I used number of pipes like I believe most people us it ie pixel pipes. I did not say how many vertex pipes were in each gpu. And given the same number of vertex pipes in each gpu the one with more pixel pipes will always do the same or better than the one with fewer pixel pipes. With today's games which are not vertex bound it will alway do better assuming the gpu has a reasonable about amount of vertex pipelines say 6. So keeping the pixel pipes full is just a matter of matching the other gpu components like vertex pipes up with the number of pixel pipes. The ratio between all the gpu parts will depend on the efficiency of each part. As for unified shader model. It is not the kill all design. While no pipeline will ever be empty with unified shader pipes they will be less efficiency. Specialized parts always run faster. The only reason I see to go to unified shaders is to not have to worry about balancing out the ratio of gpu parts and maybe to reduce cost of the gpu. I do not see speed as a reason. Unified shaders compared to the shaders we have now could be viewed like the cpu's that can do graphics, AI, sound, etc., vs video card, sound card, and soon to be physics card. Sure the cpu's can do everything, but they will have to run extremely fast to beat dedicated parts like the video card and physics card in games, but you will not be able to tell what video card you have while using MS Word. So end the end it is a trade off. Do you think it is better to go generic or add more parts that may or may not be used all the time. Myself, I don't really like the generic solution.

 

[Gamingphreek]

No. I could add 1000 pipes and ill bet you it is only marginally better.

Its like a check out line at a grocery store. I could open 100 registers, but if there is only 50 people in the store you cannot keep the other 50 registers doing something. Obviously in a GPU's case it can attempt to balance the load and keep the pipelines full at all times, but there is only so much you can do.

Just out of curiosity do you even know exactly what a unified shader model is? It doesn't seem like you do.

-Kevin

 

[HendrixFan]

AoE3 has been optimized for Nvidia cards, you will see a boost over a comparable ATI card. Optimizations for ATI may come later, but it isnt certain.

The game will be a beast at full quality settings, and a fast CPU will also make a big difference. Pathing for the Age series games has always been a big CPU burden.

If I were you, I would wait until the game comes out to buy your card, if you are indeed choosing the card based on this game. The prices will be cheaper in the month and a half between now and expected release.

 

[CU]

Originally posted by: Gamingphreek
No. I could add 1000 pipes and ill bet you it is only marginally better.

Its like a check out line at a grocery store. I could open 100 registers, but if there is only 50 people in the store you cannot keep the other 50 registers doing something. Obviously in a GPU's case it can attempt to balance the load and keep the pipelines full at all times, but there is only so much you can do.

Just out of curiosity do you even know exactly what a unified shader model is? It doesn't seem like you do.

-Kevin

In a game you can up the resolution or turn on stuff like AA to increase the number of things the gpu needs to do. If you try to compare it to a grocery store and people then if you only have 50 people and have 100 registers then you increase the number of custumers by having a sale. Now you can use all your registers. I have read about the unified shader model, but I have only wrote programs for our current hardware that is not based on the unified shader hardware. So, I don't have first hand experience with it. My understanding of them is they are pipes that can adapt to either do vertex or pixel data. But they will not be as fast as dedicated shaders. They should also be a little easier to program for. But if you want speed you use dedicate hardware and program it the most efficent way using all the tricks no matter the complications. You don't use generic hardware and program it the easiest way. But it all comes down to what ATI and Nvidia decide to do. Right now ATI want to go unified but Nvidia doesn't, atleast not yet.

 

[Gamingphreek]

You can argue this all you want. There is no game out there that will take advantage at even 100 Pixel Pipelines at Max everything. Your analogy is, however, correct.

-Kevin

 

[Marsumane]

Originally posted by: VIAN
I don't think you should make any decisions until the game comes out and some benches pop up.

:thumbsup: