• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

PPU more then 100% faster then CPU in UT3 benchmark here.

metalmackey said:
Can someone please expain how taking the normal physics from a quad core and move to a PPU increases the fps? I mean, there are hardly "cpu intensive" are they.
Click to expand...

If you are CPU limited any load you can move from the CPU to another component should result in an increase in fps as the CPU can then send data to the GPU faster to render. The GPU can only render as fast as the CPU can send data to it.

Can you do a test and see what your CPU usage is like on a quad core on a nomal UT3 map?
 
Potty why are you still posting? Its quiet clear these people have seen the evidence, read the benchmarks and are all old enough to come to their own conclusion on how good/bad the PPU is. Why do you insist on trying to argue their oppinion? They dont like it, end of!

I also can't see any reason at this moment why anyone would want to buy a PPU. As far as i'm aware, we are all getting 60+ fps in UT without the PPU on our mid range 8800gt's anyway :p
 
Looking at the benchmarks and resource usuage I don't think it is possible to conclude that the PPU is doing a bad job at calculating the additional physics without a bit more research. The charts show that the game is still CPU limited on a dual core machine when all physics prossesing is handed to the PPU. This indicates that it is still the CPU holding back performance as your fps will only be as fast as the cpu can send data to the gpu.

My understanding was that if you have a PPU installed all physics calculations handle by the Ageia API get handled by the card. In UT3 this would be all of the basic physics calculations in the game.

The fact that the game is still CPU limited means there is a pretty heavy load on the CPU outside of the physics calculations. It would be interesting to see if this is also the case on a standard map or when a quad core CPU is used.

The fact that the game is still CPU limited on the PPU maps with a PPU coud be down to a number of reasons. If there are a lot of additional objects being added to the scene these all need to be sent to the GPU to render. One of the main problems with DX9 is the CPU overhead in making draw calls. If rendering the extra objects are creating a lot of additional draw calls this could be a reason for the CPU limitation.

Another possible reason is an increase in the time taken to process the AI routines. You would need an AI routine to control a computer charaters movements around the level. If you add lots of additional objects these will need to be taken into account when determining the characters movements in order for them not to collide with them. You may also have the scenario where a previously accessible route may now be blocked by debris. In this case all of the pathfinding routines would need to be updated a lot more frequently than a level which didn't change.

I am not saying that ultimately the ppu is not to blame for the poor performance however I think a bit more research needs to be done. Especially given the fact that the game is still cpu limited with a PPU.
 
Marc Fraser said:
Looking at the benchmarks and resource usuage I don't think it is possible to conclude that the PPU is doing a bad job at calculating the additional physics without a bit more research.
Click to expand...
We know that the game runs at ~60fps with no PPU physics and no PPU, but runs at half the framerates with PPU physics and a PPU. It's not unfair to speculate that the physics effects generated by the PhysX API are even far too much for the Ageia PPU to cope with and maintain good framerates.
 
In the long term, PhysX will lose out against graphics cards with physics chips on them. Pointless paying for PhysX now, whereas you can buy a graphics card with a physics chip Q3/4 next year.

Anyway, who wants to play at bloody 30fps for just physics effects. When I play FPS games, anything under 50fps really gets on my man ****! :D
 
Ulfhedjinn said:
We know that the game runs at ~60fps with no PPU physics and no PPU, but runs at half the framerates with PPU physics and a PPU. It's not unfair to speculate that the physics effects generated by the PhysX API are even far too much for the Ageia PPU to cope with and maintain good framerates.
Click to expand...

But the game is still CPU limited. If it was the PPU that was the bottleneck you would expect the CPU to be waiting on the PPU and not have 100% usage?

I am not saying the decrease in performance is not down to the additional effects, what I am saying is that by looking at the evidence it looks like it is the CPU where the bottleneck is as it is running at 100% load.
 
Marc Fraser said:
But the game is still CPU limited. If it was the PPU that was the bottleneck you would expect the CPU to be waiting on the PPU and not have 100% usage?

I am not saying the decrease in performance is not down to the additional effects, what I am saying is that by looking at the evidence it looks like it is the CPU where the bottleneck is as it is running at 100% load.
Click to expand...
But the CPU doesn't play a part in the PhysX effects, that's what a PPU is for. If the framerate drops on a PPU-optimised map with PhysX effects then it's obvious where the problem lies if you ask me.
 
Paying £88 to have a few extra sparksand what-not and your framerate halved seems idiotic to me, id rather put that money into something that will be used by more than a few games (only one of which i will play anyway)

I just dont understand how buying one of these can be at all justified... I, and im sure 99% of people on this forum, when playing games, would have a MUCH more enjoyable experience WITHOUT the extra PPU effects at 60FPS+ than WITH the extra effects at less than 30FPS..
 
Ulfhedjinn said:
But the CPU doesn't play a part in the PhysX effects, that's what a PPU is for. If the framerate drops on a PPU-optimised map with PhysX effects then it's obvious where the problem lies if you ask me.
Click to expand...

Whilst the CPU does not play a part in calculating the physics when you have a ppu and add more effects, as I explained above by adding more objects to the game you will be increasing the none physics work that the cpu has to do.

The fact that the CPU is still running at 100% with a ppu doing all of the physics indicates that the cpu is still the bottleneck. Whilst the additional work the cpu is having to do is probably down to all the additional objects that have been added that doesn't meant that the PPU is struggling to calculate the physics. It would be interesting to see a graph showing the PPU usage.
 
Marc Fraser said:
Was that one of the PPU maps?
Click to expand...
What does it matter? On a PPU map with a PPU card, the CPU shouldn't have ANY additional load, that's the whole freaking point of the damn card, it shifts the extra load caused by the Ageia PhysX effects to the card.

The problem here is either the card can't handle the extra physics effects the PPU maps are throwing at it, or the technology in general is just bunk. I'm going to go with the latter.
 
MystaEB said:
What does it matter? On a PPU map with a PPU card, the CPU shouldn't have ANY additional load, that's the whole freaking point of the damn card, it shifts the extra load caused by the Ageia PhysX effects to the card.

The problem here is either the card can't handle the extra physics effects the PPU maps are throwing at it, or the technology in general is just bunk. I'm going to go with the latter.
Click to expand...
I'm going to go with both. :D
 
*not read whats above because i cant be bothered to read the same arguement over and over again... but i dont mind joining in :)*

Now as stated in reviews the physics isn't worth it in UT3. There isn't much added unless you play the physics only maps, which are poorly optimized.

Why do you constantly try and argue this pottsey? Sorry but you have lost this argument about the PPU being worth the buy now. In the other one you constantly said ooo look at these effects you cant do them on a quad core. You showed no proof its not possible. You claim crysis has bad physics because the quad cant handle it. Where is the evidence/ The quad doesn;'t even use half of its power it doesn't even max 2 cores. The reason it cant handle it is because the graphics cards cant handle producing anything more on screen then what we currently have.

Let it die pottsey. You have yet to come out with any decent facts about why we should have it. You said oo we can get liquid effects we cant make on the cpu where is your proof? The effects looked crap. You said its not avbout how they look its about the physics behind them. Well if developers using these physcs dont make them look good we dont want to know therefore we dont care.

You claim that it helps take load off the cpu bla bla bla and you say ooo well unles you have a quad core you need it. Sorry but why spend 80 quid on a ppu when you could put that into a quad core and the quad qwould benefit all games, well a lot more then the ppu does.

Your arguments are just badly flawed.

Bit-tech didn't like the physics in it either:
http://www.bit-tech.net/gaming/2007/11/23/unreal_tournament_3/4
 
Ulfhedjinn said:
If it wasn't, which I suspect from those results, then it just shows that the game isn't CPU limited. The game is obviously GPU-limited on normal maps and PPU-limited on PhysX "optimised" maps.

What don't you understand about this?
Click to expand...

Did you look at the original website Pottsey posted?

It was showing very high CPU usage (often hitting 100%) on a dual core cpu on a normal map which indicates that when you just have a dual core cpu it is CPU limited.

It has now just been tested with a quad core which shows lower cpu usage accross the four cores but still more than could be done on a dual core without becoming cpu limited.

With a quad core on a normal map it does not appear to be CPU limited so adding a PPU may not give much benefit. That is unless the cores that are handling the physics are consistantly hitting 100%.

From the other information on the website it is not yet possible to say that it is PPU limited on the PPU maps because even with a PPU handling the physics the CPU is still consistantly running at 100% indicating a cpu limitation. If in this scenario the cpu was not at 100% you could clearly say it was the PPU causing the problem however at present you cannot be sure.
 
MystaEB said:
What does it matter? On a PPU map with a PPU card, the CPU shouldn't have ANY additional load, that's the whole freaking point of the damn card, it shifts the extra load caused by the Ageia PhysX effects to the card.

The problem here is either the card can't handle the extra physics effects the PPU maps are throwing at it, or the technology in general is just bunk. I'm going to go with the latter.
Click to expand...

Did you read my post earlier when I explained how an additional load can be placed on the CPU by having more objects in the game? The kind of objects I am talking about are debris from destructable terrain that stays in the environment.

The only load that is shifted to the PPU is the physics calculations not the AI routines and draw call which are all increased if you have more objects in the scene.
 
You must log in or register to reply here.
Back
Top Bottom