• 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.

CPU bottlenecking graphics card?

I

igottapae

Weaksauce
Joined
Apr 17, 2008
Messages
118
well i was about to get a 5870 but i have a E7400 @3.5ghz and was wondering if the processor would bottleneck the card?
 
it could if your playing 1680 x 1050 or lower, you would be more CPU bound then vid card bound, if your cranking it upto 1920+ you should be be fine.

what games?
 
MrGuvernment said:
it could if your playing 1680 x 1050 or lower, you would be more CPU bound then vid card bound, if your cranking it upto 1920+ you should be be fine.

what games?
Click to expand...

pretty much just any game but recently it would be battlefield bad company 2.
 
I wouldnt think it would bottleneck but saving for a upgrade in the future might not be a bad idea. It would let that 5870 run a little more free.
 
Gym Rat said:
I wouldnt think it would bottleneck but saving for a upgrade in the future might not be a bad idea. It would let that 5870 run a little more free.
Click to expand...

yea i was planning on getting a Q9550 later on but it was ether the Q9550 now or a 5870.
 
It'll hold it back a bit in most games, but nothing so serious that you shouldn't buy the card. It's not like you're going to LOSE performance, you just won't get as much as you could have. If you're that concerned about it, you can always get a 5850.
 
Run some benchmarks, lower your OC, run some benchmarks again.

If the benchmarks went down, its possible its bottlenecking. If your FPS stays the same even after you lower your OC, then you know that it's already GPU bottlenecked
 
Easy way to tell.
Get the card and monitor CPU use.
If it regularly stays between 90 to 100% occupancy an ANY core, the CPU is slowing the gfx card down.
There is another caveat though, while a component can drag down the framerate, it doesnt matter if the framerate is high enough.
So if you dont notice any slow down then it doesnt matter if there is a bottleneck.

Test and ye shall find :)
 
just to clear up a couple things i dont have the card yet so i cant really test anything im getting it in a couple days along with some other stuff.

4gb of ddr2 800
750w psu
5870
antec 1200

also i was thinking about getting a 5850, but im real bad on wanting a lifetime warranty. the XFX 5850 is 350 and the XFX5870 is 399 so why not get the 5870 :).
 
Last edited:
Most people upgrade one piece at a time, at least those of us on a budget, so get the 5870, save up and get 2 more GB or RAM then save up some more then get you a nice E8500 or better if budget allows enjoy.

Thats how Im gonna do it. I know my Athlon X2 7750 could possibly be a bottleneck for a 5770 but itll do until I can scrounge up enough cash for either a Phenom II 550 or Athlon II 435.
 
igottapae said:
just to clear up a couple things i dont have the card yet so i cant really test anything im getting it in a couple days along with some other stuff.

4gb of ddr2 800
750w psu
5870
antec 1200

also i was thinking about getting a 5850, but im real bad on wanting a lifetime warranty. the XFX 5850 is 350 and the XFX5870 is 399 so why not get the 5870 :).
Click to expand...

You will be fine, save up for a little while and buy the Q9550 used or something (for cheap), or save up even longer and pick up an i5 750 setup.
 
the 2 things i would change right now are

600w OCZ psu (Either: corsair 650TX or 750TX)
2gb corsair ddr2 800

OCZ PSU's are not the best and 2GB of ram just wont cut it. i think you will be fine once these are changed.
 
I have a Pentium D 820 running at the stock 2.8ghz. I just recently bought a PCI-E PNY XLR8 Nvidia 250. I'm not sure exactly where I hit the ceiling but I know I do. A friend of mine has a Q6600 stock speed and using my card and me using his Nvidia 260 he gets much better framerates then I do.
 
mikegoes101 said:
I have a Pentium D 820 running at the stock 2.8ghz. I just recently bought a PCI-E PNY XLR8 Nvidia 250. I'm not sure exactly where I hit the ceiling but I know I do. A friend of mine has a Q6600 stock speed and using my card and me using his Nvidia 260 he gets much better framerates then I do.
Click to expand...

You are severely bottlenecked. My Athlon X2 bottlenecked my 8800GT and it shit stomps Pentium Ds. You should upgrade your CPU as soon as possible as you probably aren't even seeing half of that card's full potential.
 
Nenu said:
Easy way to tell.
Get the card and monitor CPU use.
If it regularly stays between 90 to 100% occupancy an ANY core, the CPU is slowing the gfx card down.
There is another caveat though, while a component can drag down the framerate, it doesnt matter if the framerate is high enough.
So if you dont notice any slow down then it doesnt matter if there is a bottleneck.

Test and ye shall find :)
Click to expand...

Your pseudo-test really won't tell you anything other than the CPU is being loaded. It won't be an indication *at all* of whether the CPU or GPU is the bottleneck.

igottapae said:
yea i was planning on getting a Q9550 later on but it was ether the Q9550 now or a 5870.
Click to expand...

Games like more clockspeed a hell of a lot more than more cores. You probably won't see much of a difference upgrading to a quad.
 
The only game that REQUIRES a quad core to be playable is GTAIV, most other games love high clock speeds.
 
kllrnohj said:
Your pseudo-test really won't tell you anything other than the CPU is being loaded. It won't be an indication *at all* of whether the CPU or GPU is the bottleneck.
Click to expand...

How so?
Thats a standard test to see if the CPU is bottlenecking and it does indeed show if the CPU is bottlenecking.
If so, higher frequency will help all games and more cores will help quite a few.
A reduction in frequency to gain more cores (ie moving to a 3GHz quad) will hurt those games that dont use more than 2 cores.

Its not a full test suite but at least it gets the troubleshooting process started!
 
Nenu said:
How so?
Thats a standard test to see if the CPU is bottlenecking and it does indeed show if the CPU is bottlenecking.
If so, higher frequency will help all games and more cores will help quite a few.
A reduction in frequency to gain more cores (ie moving to a 3GHz quad) will hurt those games that dont use more than 2 cores.

Its not a full test suite but at least it gets the troubleshooting process started!
Click to expand...

If we were dealing entirely with single threaded games, then it would be a *slight* indication of whether the CPU or GPU is bottlenecking. Even when the GPU is the bottleneck, you will still see almost full CPU usage of a single thread. Remember, the CPU has to feed the GPU so you can't know just by looking at the task manager if the CPU is busy waiting for the GPU to finish doing its thing or if the CPU is busy running game code. As soon as you throw multiple threads into the mix (which includes both ATI and Nvidia's drivers nowadays), CPU usage stops telling you squat. For example, I might have a physics thread that is going to constantly run at full tilt because it is never blocking on other calls. Of course, task manager also doesn't break things down into threads, so if you have a quad core and the CPU usage is sitting at 25% running one app, is that one thread using 100% of one core? Or two threads using 50%? Or 4 threads? etc, etc, etc...

And that is all before you even get into the mess of threads blocking on threads leading to situations where a single thread using 50% of a single core could still be CPU bound, and all sorts of issues such as a thread with 100% usage might just be waiting for some events and more CPU speed wouldn't help it in the slightest. The test is sound in theory in that if you know where the CPU is spending its time you can make a reasonable guess as to where the bottleneck is. The problem is that you don't have anywhere near enough information to be able to realistically determine squat, and the measurement tools are more of an approximation in the first place.
 
Last edited:
As you said, the CPU feeds the GPU, so if the CPU is bottlenecking, it will lower framerate.
I agree that there will be delays if the GPU is busy doing something but thats the nature of 2 devices communicating. It doesnt mean that it cant do a lot more work after it has finished if it isnt the bottleneck.
Theres little chance of the CPU and GPU both being 100% occupied, maybe during benchmarking or stress testing.
You'd be unlucky (fortunate?) when troubleshooting to come across this while gaming.

For example, playing Dragon Age or Dirt 2, my CPU is running 100% while my GPUs are running at 35% in Dragon Age and 70% in Dirt 2.
The primary reason is that I am running a E8400@4.1GHz with crossfire 5770.
Both games benefit greatly from more than 2 cores, crossfire uses a little extra for itself.
I achieve good enough framerates in both games but it does demonstrate how easy it is to check for bad framerate bottlenecks on slower hardware or more demanding games.
 
I agree with kllrnohj... Looking at task manager isn't a good indication.
 
May be in some scenarios where there's lot of geometry, a console port or simple graphics, you might be CPU bottlenecked, but you should be able to enjoy your HD 5870 without major issues. A Core 2 8600 performs very close to the i7 in most games and sometimes it even beat it, and cosidering that your CPU is 20% slower than the 8600, you will be fine.
 
Nenu said:
Bizarre, it works very well.
Click to expand...

Does it? So far you've only cited two anecdotes, neither of which actually support your theory as you didn't follow up to determine if your hypothesis was correct.
 
kllrnohj said:
Does it? So far you've only cited two anecdotes, neither of which actually support your theory as you didn't follow up to determine if your hypothesis was correct.
Click to expand...

Anecdotes?
I already explained that its a great way to see if the CPU needs upgrading, its not hard to follow.
Works very well in practice too, hasnt failed me :)

Slow your CPU down until it shows 100% CPU use while gaming and you will start to see your framerate drop.
Now speed up the CPU, amazingly, the framerate increases until the CPU isnt the bottleneck any more!
 
Nenu said:
Anecdotes?
Click to expand...
Nenu said:
For example, playing Dragon Age or Dirt 2, my CPU is running 100% while my GPUs are running at 35% in Dragon Age and 70% in Dirt 2.
Click to expand...

Yes, anecdotes.

I already explained that its a great way to see if the CPU needs upgrading, its not hard to follow.
Works very well in practice too, hasnt failed me :)

Slow your CPU down until it shows 100% CPU use while gaming and you will start to see your framerate drop.
Now speed up the CPU, amazingly, the framerate increases until the CPU isnt the bottleneck any more!
Click to expand...

And my point is that your "test" is fundamentally flawed. You are spreading something that you *believe* to be true but simply isn't, which is bad. Not only that, but games can be CPU limited *long* before they hit 100% CPU usage (hell, on my i7 I can be CPU limited at ~13% CPU usage)

Given that there are much, *much* better ways of telling if you are CPU limited quite easily, spreading a flawed method is a waste of everyone's time.
 
God so much bickering and no one offering a reliabale test, just arguing for the sake of it.

A better test for bottlenecking is to turn down video settings and see how it affects performance.

If you've got say 30fps and you turn down a bunch of video settings to increase performance and you still have 30fps then you're likely CPU bottlenecked. There are very few settings that significantly effect CPU load, physics is maybe the most common.
 
mikegoes101 said:
I have a Pentium D 820 running at the stock 2.8ghz. I just recently bought a PCI-E PNY XLR8 Nvidia 250. I'm not sure exactly where I hit the ceiling but I know I do. A friend of mine has a Q6600 stock speed and using my card and me using his Nvidia 260 he gets much better framerates then I do.
Click to expand...
holy cow what a bottleneck. even the lower end core 2 duos are twice as fast clock for clock as a Pentium D. that means my E8500 would still be much faster than your 2.8 Pentium D even if I lowered it to 1.4. your are literally pissing away about half of what even a decent mid range card is capable of.
 
kllrnohj said:
Yes, anecdotes.



And my point is that your "test" is fundamentally flawed. You are spreading something that you *believe* to be true but simply isn't, which is bad. Not only that, but games can be CPU limited *long* before they hit 100% CPU usage (hell, on my i7 I can be CPU limited at ~13% CPU usage)

Given that there are much, *much* better ways of telling if you are CPU limited quite easily, spreading a flawed method is a waste of everyone's time.
Click to expand...

As stated it works well for games, so if testing while gaming, the results are true.
It has been my assumption that the op cares about this, not wild uses of the processor that give it a 13% max load :)
 
an E7300 at 3.5 would most certainly limit the full potential of a 5870 and their is zero doubt about that. in games like Red Faction Guerrilla, Ghostbusters and GTA 4 somebody with a slower card but fast quad would get a better overall experience than a 5870 and 3.5 E7300 could deliver. now in most other games it wont matter much since the framerate will be high enough anyway. if it was my my money though I would go with the quad now and 5850 and overclock both. a 5850 is only about 5% slower than a 5870 when both are overclocked to their max anyway. that would be a very well balanced pc in my opinion.
 
Nenu said:
As stated it works well for games, so if testing while gaming, the results are true.
It has been my assumption that the op cares about this, not wild uses of the processor that give it a 13% max load :)
Click to expand...

There are more effective methods already posted in this thread. The OP has the info he needs to perform which ever tests (if any) he wants. He can choose the test that you think is good, or he can do one that several other posters thinks is better.
 
I have no problem with the op doing what he wants, I am not going to doubt what works though.
 
Your method would be "ok" for a sinlge threaded game on a single core cpu, and even then, not the best way. Things get much more complicated when me move to multiple threads and multiple cores. Lets assume 2 threads for a game on a dual core cpu, it is entirly possible that both threads are not utilized equally. You might be cpu limited on one and not the other which means a CPU utilization of under 100% but you're still cpu bound. That method becomes even less usefull with 4 cores and virtually useless if you factor in a cpu that has hyperthreading.
 
I agree in principal but in practice, it isnt what I have found.
I test with many games (and benchmarks to verify it is performing as expected) to make sure that the results are not isolated to just one game as theres not a lot of point in upgrading for just one game.

Try it for yourself, pick a game.
Reduce your CPU speed until at least one core is running at 100% and monitor framerate.
You will find that framerate starts to reduce the slower you make the CPU run.
If your case is true, there will be games that do not reduce framerate when a core is at 100% and you reduce CPU speed further.
Report the results and which game you used please.
 
lol thanks for the suggestions guys ill try these test when i get the card in a couple days.:cool:
 
Nenu said:
I agree in principal but in practice, it isnt what I have found.
I test with many games (and benchmarks to verify it is performing as expected) to make sure that the results are not isolated to just one game as theres not a lot of point in upgrading for just one game.

Try it for yourself, pick a game.
Reduce your CPU speed until at least one core is running at 100% and monitor framerate.
You will find that framerate starts to reduce the slower you make the CPU run.
If your case is true, there will be games that do not reduce framerate when a core is at 100% and you reduce CPU speed further.
Report the results and which game you used please.
Click to expand...

You don't get it. On my i7 I *can't* lower CPU speed until CPU usage hits 100%. It would require equal loading of 8 threads, which afaik no game does. It is *impossible* to tell if one thread is running at 100% with task manager. The test you are describing is simply not possible with the method you are describing. I don't care if it works for you on your C2D, that does *not* make it a valid test. I'm telling you this as a developer who has done threading and OpenGL work - the reality of the situation is vastly more complicated than you realize.

There are far better and easier ways to test. For example, change the amount of AA - if FPS changes appropriately then you aren't CPU limited.
 
You must log in or register to reply here.
Back
Top