• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Video/CPU Bottlenecking? Expert Help Needed

X

xplicit

[H]ard|Gawd
Joined
Sep 21, 2003
Messages
1,333
Can someone explain to me bottlenecking in easy to understand CORRECT terminology?
I understand it currently as this:

Considering that the 4x AGP bus (266MHz) hasn't even been saturated yet to even begin to bring 8x AGP into play, how do people figure current CPUs are going to be a bottleneck?
Until the 8x bus becomes saturated, isn't the AGP always going to be the bottleneck?

Putting PCI Express into the mix: PCI Express will place (eventually) that bottleneck back onto the CPU, correct?

Am I correct, or way off base here? HELP!
 
bottlenecking in other places, by cpu/gpu bottlenecking we're all basicly saying the GPU finishes its cycles before the CPU does so it has to wait on the CPU to finish, and what your thinking of its AGP bottlenecking, which we have yet to come across
 
Xxplosiv said:
bottlenecking in other places, by cpu/gpu bottlenecking we're all basicly saying the GPU finishes its cycles before the CPU does so it has to wait on the CPU to finish, and what your thinking of its AGP bottlenecking, which we have yet to come across
Click to expand...

i agree. if you have a slow cpu with a high end graphics card then you wont get the same performance as you would with a faster cpu. even the RAM speeds can create a bottleneck by not being able to handle the flow of data.
 
Well if your running less than a Pentium 4 3.0GHz or less than an Athlon XP/64 3000+ I'd say you'll be horribly bottlenecked. The differences between an X800Pro/XT-PE/6800GT/6800U won't be that apparent in games for the most part.

Past 3.4ish or 3400+ things start to take shape and the faster cards start to show their muscle over their lower cost cousins. Still a CPU hasn't been released that will really allow these cards to fully be utilized.
 
Still confused (or horribly retarded).

If the 4x AGP bus (266MHz) isn't even being saturated by the current line of AGP cards, how can it be limited by a CPU that's running at bus speeds of 266 or higher?

As long as you're running a CPU at bus speeds of 266 or higher, isn't the unsaturated 266MHz AGP bus always going to be the bottleneck?
 
Don't be overly concerned about the bus traffic. Consider the GPU and CPU as two independent processors working on a percentage of the total load. Forget about what compromises each load (AI, physics, triangle setup, texturing, shaders, etc).

For there to be no "bottle neck" each processor completes whatever work is required in roughly the same amount of time. As soon as one side is waiting on the other, you become limited by the slower component. Take the extreme limit of each side. Say the CPU is supplying no data at all to the video card to process. The videocard is cable of doing infinately more work than the CPU since the CPU is doing nothing. The same goes the other way around and the videocard does nothing.

A well balanced system will have the load distributed with no single component racing far ahead, or falling significantly behind the other. Either case would provide a system limiting function to some (often exaggerated) extent.
 
xplicit said:
Still confused (or horribly retarded).

If the 4x AGP bus (266MHz) isn't even being saturated by the current line of AGP cards, how can it be limited by a CPU that's running at bus speeds of 266 or higher?

As long as you're running a CPU at bus speeds of 266 or higher, isn't the unsaturated 266MHz AGP bus always going to be the bottleneck?
Click to expand...

Ur way of thinking is making it seem like taking ur numbers and having a 10$ 8x agp card or 600$ agp card will yield the same speed in the end! =) Soo stop thinking of the bus as the only indicator of speed and what not.
 
Well physics and ai which are done on the cpu primarily right now has increased 10 fold in the last three years, which has caused a major switch as before graphics before was the bottleneck.

Physics is still increasing and with the attachment of ai animation blending, ai too is still increasing, but later on in the next couple generation of cards, I think you will see phsyics will be handled more on the gpu side or another chip that will be placed on the graphics card itself. This will help reduce the cpu cycles for other operations plus also decrease the neccessary bandwidth games need during animations as animated characters are sent through the agp/ pci-e port every single frame.
 
Here, I'll try to make it as simple as possible:

The CPU has work to do and the GPU has other work to do. The AGP (or PCIe) bus is simply the flow of data between the two of them (ie. independant of the actual processing of the data. It is just the flow of data).

If your GPU has to wait for the CPU to finish it's calculations, you are CPU limited. If the CPU finishes it's calculations first, and the GPU is not quite finished, you are not CPU limited (aka. you are limited by your video card).

You will only be limited by the bus between the two if both the CPU and GPU are waiting for the data to be transfered. This will only be the case when there is a lot of data to be transfered, and hence, why we are not limitted by AGP 4X or 8X. Once AI and physics become extremely complicated (not just in the calculation, but also in the amount of data needed to send to the video card), then we will need things such as PCIe.
 
Sir-Fragalot said:
Well if your running less than a Pentium 4 3.0GHz or less than an Athlon XP/64 3000+ I'd say you'll be horribly bottlenecked. The differences between an X800Pro/XT-PE/6800GT/6800U won't be that apparent in games for the most part.

Past 3.4ish or 3400+ things start to take shape and the faster cards start to show their muscle over their lower cost cousins. Still a CPU hasn't been released that will really allow these cards to fully be utilized.
Click to expand...

with this system :

P4 2,53
512 MB corsair memory platinum 2-2-2-5 dual channel 266 mhz (2100)
asus P4G8X with intel E7205 chipset supporting dual channel

due to the cpu bottleneck factor, adding anything more than an x800 pro would
be useless ,right ?

also adding one more GB of ram (1,5 GB total ) ,instead of 512 would have any
noticable effects in everyday use (Games etc) or i will see difference
only in benchmarks ?
 
intercollector said:
Here, I'll try to make it as simple as possible:

The CPU has work to do and the GPU has other work to do. The AGP (or PCIe) bus is simply the flow of data between the two of them (ie. independant of the actual processing of the data. It is just the flow of data).

If your GPU has to wait for the CPU to finish it's calculations, you are CPU limited. If the CPU finishes it's calculations first, and the GPU is not quite finished, you are not CPU limited (aka. you are limited by your video card).

You will only be limited by the bus between the two if both the CPU and GPU are waiting for the data to be transfered. This will only be the case when there is a lot of data to be transfered, and hence, why we are not limitted by AGP 4X or 8X. Once AI and physics become extremely complicated (not just in the calculation, but also in the amount of data needed to send to the video card), then we will need things such as PCIe.
Click to expand...

It's official: I'm braindead.
The way you explained it makes so much obvious sense, I'm humbled (and ashamed) at it's simplicity..and to think I call myself a computer geek.

Why couldn't I see that?
Easy: I was trying to work around the answer instead of working though it.
Thank you.

And thanks to everyone else for the replies. You are all appreciated.
 
I think maybe I can give you a good real world example.

I am the kinda guy that waits to upgrade till my computer is slow in a game I play, unlike some here that buy a new proc every 6 months or whatever. That way I get big jumps.

Now I just upgraded everything, but I got a new motherboard/proc and stuck with the same video card for 2 weeks till I got my 6800 GT.

So I went from an Athlon XP 1800+ @ 1.69 ghz with a Radeon 9700 Pro, to a Athlon 64 2.2 ghz @ 2.6 ghz with the same 9700 Pro

I ran UTbench and a hoard of other benchmarks before I upgraded and after.

Now in UT2k4 on the 1800+ I was getting 32.546 fps @ 1024, with the exact same settings on the A64 I got 66.1

FarCry PC games demo I went from 34.39 fps to 64.41fps.

So the Radeon 9700 Pro was seriously limited by the 1800+ @ 1024x768 no aa no af.

The biggest difference would have to be in FarCry and doom 3, soon as bullets flew, hit bad guys, and physics started going, my fps would dive to the teens or lower, but on the a64, it was smooth sailing.

So since the 9700 Pro was waiting on the 1800+, even if I had run out and got a 6800 Ultra Extreme Edition and stuck with the 1800+, I still would be getting low fps, I would see a bump yes, especially with AA/AF, but overall that would have been a pretty stupid upgrade when you consider over all system responsivness and other things that a faster proc and memory system do for you.

Hope that helps :)

P.S. Oh yeah, the jump from the 9700 Pro to 6800 GT when I finally got it was very nice :)
 
You must log in or register to reply here.
Back
Top