Sign in with
Sign up | Sign in
Your question
Solved

Which is bottlenecking which in this scenario?

Tags:
  • GPUs
  • Bottleneck
  • CPUs
Last response: in Graphics & Displays
Share
September 15, 2014 1:42:36 PM

Example:
GPU usage: 100%
CPU usage: 65%


My assumption here is that the GPU can't keep up with the CPU. Is this correct or incorrect?

Or can a GPU be bottlenecked by a CPU load that isn't at 100%? (Also RAM isn't maxed)

More about : bottlenecking scenario

a c 418 à CPUs
September 15, 2014 1:47:42 PM

Nothing is bottlenecking. Bottleneck is a term used to describe the CPU not being able to keep up with the gfx card. In your case your are getting 100% of your money out of your card. You don't want your CPU to run at 100% if you can help it. It has other tasks to perform.
m
0
l
a c 112 à CPUs
September 15, 2014 1:49:30 PM

The GPU should always be 100%, as it's always going to go faster to push out more FPS. Unless you have a very fast card and lock your refresh rate.

EG, your card can pump out 120FPS at 100% usage, but you refresh rate is 60fps and you lock your card to that, then your GPU probably won't be 100%, maybe not half, but not a full 100%.

Otherwise, your card can do 120FPS at 100%, so it puts out 120FPS. It won't go any less.

The only time there is a "bottleneck" really, is if the GPU was 50% and the CPU was 100%. Then the CPU isn't capable of keeping up with the game or GPU demands.
m
0
l
Related resources
September 15, 2014 1:52:32 PM

clutchc said:
Nothing is bottlenecking. Bottleneck is a term used to describe the CPU not being able to keep up with the gfx card. In your case your are getting 100% of your money out of your card. You don't want your CPU to run at 100% if you can help it. It has other tasks to perform.

That makes sense. The reason I ask this is because I'm using an old CPU with a decent GPU, but still am not getting good FPS. My GPU is able to reach 100% load, so that means my CPU is sufficient?

CPU: Phenom 9550
GPU: Gigabyte R9 270

Why do people purchase new CPU's if old one's suffice? (Not hardcore gaming setups)
m
0
l
a c 96 à CPUs
September 15, 2014 1:52:59 PM

If ur asking what to upgrade in a round about way yes, the GPU is your answer but I suspect if its loading your CPU up to 60% much of a GPU upgrade won't get the full effect.
m
0
l
a c 96 à CPUs
September 15, 2014 1:54:45 PM

Sometimes older boards don't allow a GPU to reach its true potential but that CPU will hold you back if you went to a 280 or better, or basically any "current" gen game
m
0
l
September 15, 2014 1:55:11 PM

getochkn said:
The GPU should always be 100%, as it's always going to go faster to push out more FPS. Unless you have a very fast card and lock your refresh rate.

EG, your card can pump out 120FPS at 100% usage, but you refresh rate is 60fps and you lock your card to that, then your GPU probably won't be 100%, maybe not half, but not a full 100%.

Otherwise, your card can do 120FPS at 100%, so it puts out 120FPS. It won't go any less.

The only time there is a "bottleneck" really, is if the GPU was 50% and the CPU was 100%. Then the CPU isn't capable of keeping up with the game or GPU demands.

If the GPU isn't always at 100%, does that mean the CPU isn't able to supply information fast enough to keep up?
m
0
l
September 15, 2014 1:55:11 PM

getochkn said:
The GPU should always be 100%, as it's always going to go faster to push out more FPS. Unless you have a very fast card and lock your refresh rate.

EG, your card can pump out 120FPS at 100% usage, but you refresh rate is 60fps and you lock your card to that, then your GPU probably won't be 100%, maybe not half, but not a full 100%.

Otherwise, your card can do 120FPS at 100%, so it puts out 120FPS. It won't go any less.

The only time there is a "bottleneck" really, is if the GPU was 50% and the CPU was 100%. Then the CPU isn't capable of keeping up with the game or GPU demands.

If the GPU isn't always at 100%, does that mean the CPU isn't able to supply information fast enough to keep up?
m
0
l
September 15, 2014 1:57:10 PM

Supahos said:
Sometimes older boards don't allow a GPU to reach its true potential but that CPU will hold you back if you went to a 280 or better, or basically any "current" gen game

You're referring to my 9550? That makes perfect sense. Do you think that applies to GTX 750ti and R9 270 as well?
m
0
l
a c 96 à CPUs
September 15, 2014 2:05:51 PM

Seems to be feeding it on whatever game you're playing, but ram speed, and other factors could be holding you back as well. 750ti and 270 are pretty even cards used to have one until I got a good deal on a 760 and gave my 750ti to my nephew
m
0
l
September 15, 2014 2:12:25 PM

Supahos said:
Seems to be feeding it on whatever game you're playing, but ram speed, and other factors could be holding you back as well. 750ti and 270 are pretty even cards used to have one until I got a good deal on a 760 and gave my 750ti to my nephew

I see, that's something I was thinking as well. Even though my RAM isn't maxed at 100%, it's 400mhz DDR2 as opposed to 1866mhz DDR3.

Which setup do you think would yield better gaming performance?

Phenom 9550 + Gigabyte R9 780x or FX 6300 + GTX 750ti?
m
0
l
a c 418 à CPUs
September 15, 2014 2:19:20 PM

A Phenom I X4 9550 is able to push that card to 100% while only running 65% usage itself? Must be GPU limited games. Did you test that while playing single player? Still, that's impressive for that old 1st gen K10 processor.

Although, I guess it should be too much of a surprise. My Phenom II X4 @ 4.0 GHz pushes a R9-280 to 100% usage.
m
0
l
September 15, 2014 2:30:47 PM

clutchc said:
A Phenom I X4 9550 is able to push that card to 100% while only running 65% usage itself? Must be GPU limited games. Did you test that while playing single player? Still, that's impressive for that old 1st gen K10 processor.

Although, I guess it should be too much of a surprise. My Phenom II X4 @ 4.0 GHz pushes a R9-280 to 100% usage.

Yes! It's very surprising to me. I've tested it on AC4, NBA 2K14, and Sleeping Dogs (With high res pack) and all settings maxed out. I'm guessing these are GPU limited games?

Does this mean the CPU is fine, and could probably even go to a r9 280-280x myself?
m
0
l
a c 418 à CPUs
September 15, 2014 2:39:24 PM

McDohl said:
clutchc said:
A Phenom I X4 9550 is able to push that card to 100% while only running 65% usage itself? Must be GPU limited games. Did you test that while playing single player? Still, that's impressive for that old 1st gen K10 processor.

Although, I guess it should be too much of a surprise. My Phenom II X4 @ 4.0 GHz pushes a R9-280 to 100% usage.

Yes! It's very surprising to me. I've tested it on AC4, NBA 2K14, and Sleeping Dogs (With high res pack) and all settings maxed out. I'm guessing these are GPU limited games?

Does this mean the CPU is fine, and could probably even go to a r9 280-280x myself?


I'm not sure whether those games are GPU or CPU limited. But if you are playing single player, the CPU won't have as much work to do as it would on a busy server with 64 other players to keep track of.

If you are talking strictly those games, you apparently have sufficient overhead for a faster card, yes. What were you using to record CPU and GPU usage at the same time?

I don't suppose that is a Black Edition processor that you can easily OC?

m
0
l
September 15, 2014 2:54:39 PM

clutchc said:
McDohl said:
clutchc said:
A Phenom I X4 9550 is able to push that card to 100% while only running 65% usage itself? Must be GPU limited games. Did you test that while playing single player? Still, that's impressive for that old 1st gen K10 processor.

Although, I guess it should be too much of a surprise. My Phenom II X4 @ 4.0 GHz pushes a R9-280 to 100% usage.

Yes! It's very surprising to me. I've tested it on AC4, NBA 2K14, and Sleeping Dogs (With high res pack) and all settings maxed out. I'm guessing these are GPU limited games?

Does this mean the CPU is fine, and could probably even go to a r9 280-280x myself?


I'm not sure whether those games are GPU or CPU limited. But if you are playing single player, the CPU won't have as much work to do as it would on a busy server with 64 other players to keep track of.

If you are talking strictly those games, you apparently have sufficient overhead for a faster card, yes. What were you using to record CPU and GPU usage at the same time?

I don't suppose that is a Black Edition processor that you can easily OC?


I was using MSI Afterburner. Is that sufficient? I rarely play multi, too many people are too good haha. But that makes sense in regards to multi. But it's great to know I have the overhead, I'd have never thought this old chip could still keep up. I guess I'll pair it with a 280x instead of going the fx 6300+750ti route.

Unfortunately it's not BE. :(  Here's a screen shot I just took: http://oi59.tinypic.com/23r9cz.jpg
m
0
l
a c 418 à CPUs
September 15, 2014 3:39:11 PM

In that screen shot, one core was hitting 70%. Getting up there. But that is still good. I don't know if I'd push my luck with a jump to a R9-280X. Then you would definitely have some bottle neck. In between what you have now and the 280X are two other cards; the R9-270X, and 280. I think you may be at the sweet spot for CPU/GPU/MB now... or maybe with an R9-270X. I imagine you still have an older AM2 or AM2+ MB with DDR2 1066MHz (or lower) system RAM.

Are you getting stutter or lag when you game? Other parts of the system may be the cause if you are. The memory may be bottle-necking. Or the Hyper Transport speed.
m
0
l
September 15, 2014 3:56:25 PM

clutchc said:
In that screen shot, one core was hitting 70%. Getting up there. But that is still good. I don't know if I'd push my luck with a jump to a R9-280X. Then you would definitely have some bottle neck. In between what you have now and the 280X are two other cards; the R9-270X, and 280. I think you may be at the sweet spot for CPU/GPU/MB now... or maybe with an R9-270X. I imagine you still have an older AM2 or AM2+ MB with DDR2 1066MHz (or lower) system RAM.

Are you getting stutter or lag when you game? Other parts of the system may be the cause if you are. The memory may be bottle-necking. Or the Hyper Transport speed.

Do you think I could push towards 280x if I were to overclock? I haven't touched that yet, currently at 2.2ghz and I think I could get to 2.7-3ghz.

Great points. I don't on Sleeping Dogs, but I DO have quite a bit of stuttering on AC4, then again it's AC4. I may get BF4 or Metro 2033 to test that out. Thanks for the input so far, you've been very helpful.
m
0
l
a c 418 à CPUs
September 15, 2014 5:25:00 PM

It's hard to get much more than a few-megahertz bump in BCLK frequency w/o an unlocked CPU clock multiplier. It would help, yes. But not enough to warrant an R9-280X... in my opinion. I doubt my Phenom II X4 965BE at 4.0GHz could keep up with an R9-280X. And it is on a late AM3 MB with DDR3 1600MHz RAM.
If you have money to burn and want to try it out, go for it. But I doubt you'd get any better gaming performance than with the R9-270X. Btw, there's a nice jump in performance between the 270 and 270X. http://www.anandtech.com/bench/product/1080?vs=1043
m
0
l
September 15, 2014 7:42:31 PM

clutchc said:
It's hard to get much more than a few-megahertz bump in BCLK frequency w/o an unlocked CPU clock multiplier. It would help, yes. But not enough to warrant an R9-280X... in my opinion. I doubt my Phenom II X4 965BE at 4.0GHz could keep up with an R9-280X. And it is on a late AM3 MB with DDR3 1600MHz RAM.
If you have money to burn and want to try it out, go for it. But I doubt you'd get any better gaming performance than with the R9-270X. Btw, there's a nice jump in performance between the 270 and 270X. http://www.anandtech.com/bench/product/1080?vs=1043

You're absolutely right, I forgot about that. Do you think an HD 7950 would be pushing it? I think it's slightly better than a 270x, basically a normal r9 280. Thanks for providing that link! I didn't know the difference between the 270 and 270x would be that considerable. If you don't think a 7950 would work I'll likely go the 270x route.
m
0
l

Best solution

a c 418 à CPUs
September 15, 2014 8:02:15 PM

The HD 7950 is indeed the R9-280. The "R9-280" is just re-badged*. In fact, my above-mentioned R9-280 shows up as an HD 7950 in most apps. Again, more than the older Phenom/MB/DDR2 could keep up with. I would expect CPU usage to be 100% and the card at less than that. And you don't want your CPU cranking away at 100%. It has other work to do... and the stock cooler (?) will sound like a turbine.

You could also consider the GTX 760. It falls in the same category as the R9-270/270X. But keep in mind, most of these close matches between cards are almost impossible to tell the difference in actual gaming. Here is another chart for you. It is a gaming gfx card hierarchy chart that places the cards in tiers. As the article states, you would have to move at least 3 tiers to see enough improvement to make the price difference worth while.
http://www.tomshardware.com/reviews/gaming-graphics-car...

* The reference HD 7950 is a bit lower performance than the R9-280 due to slightly lower clock speed. But most manufacturers have them at the same speed as the 280.
Share
September 15, 2014 9:13:27 PM

clutchc said:
The HD 7950 is indeed the R9-280. The "R9-280" is just re-badged*. In fact, my above-mentioned R9-280 shows up as an HD 7950 in most apps. Again, more than the older Phenom/MB/DDR2 could keep up with. I would expect CPU usage to be 100% and the card at less than that. And you don't want your CPU cranking away at 100%. It has other work to do... and the stock cooler (?) will sound like a turbine.

You could also consider the GTX 760. It falls in the same category as the R9-270/270X. But keep in mind, most of these close matches between cards are almost impossible to tell the difference in actual gaming. Here is another chart for you. It is a gaming gfx card hierarchy chart that places the cards in tiers. As the article states, you would have to move at least 3 tiers to see enough improvement to make the price difference worth while.
http://www.tomshardware.com/reviews/gaming-graphics-car...

* The reference HD 7950 is a bit lower performance than the R9-280 due to slightly lower clock speed. But most manufacturers have them at the same speed as the 280.

That's an extremely helpful chart! Damn, that would've saved me a *lot* of time. I'm relatively new to the PC gaming domain, so. But that makes sense, I think you're absolutely right about it not being able to handle the 7950.

I swapped the 270 out for the 750ti, ran it through AC4/SD/NBA2k and it was quite revealing.

http://oi60.tinypic.com/316q79h.jpg

As you can see, CPU is maxed. Three cores hit 100% at once, actually. This is very strange, though. With the 750ti it used only about 50% of the GPU resources, but *far* more CPU resources. In the end the FPS was far more stable, and far superior to the 270. With the 270 it was reversed: around 50% CPU resources, and 98% GPU resources. I don't understand that...

I'm starting to think upgrading the CPU may be the best bet. And that these combined scores are too similar for both cards not to be bottlenecked (Maybe mobo, ram, cpu):

r9 270:
http://www.3dmark.com/3dm/4054907

750ti
http://www.3dmark.com/3dm/4058934
m
0
l
a c 418 à CPUs
September 16, 2014 12:42:25 PM

How the heck did you get a GTX750 Ti to be recognized on that old Asus M3A78-EM MB? The new Maxwell GPU is usually a no-go for old MBs. Even with their most recent BIOS version.

As the synthetic benchmark points out, the R9-270 should be and usually is a considerably faster card than the GTX 750 TI. I have both cards on my shelf right now and have tested with both. The 270 always beats the 750 Ti.
But I can't explain the difference you report between CPU/GPU usage in that game you linked to. The CPU should have been at a higher usage with the R9-270 than the 750 Ti. (??) How did the frame rates compare? Possibly that game was optimized for Nvidia drivers? Seems like a big difference, but that is all I can think of.

Are you positive all settings were the same with both cards? Nivdia may have had PhysX enabled, where AMD didn't because it has no PhysX. As a result, it had to do the CPU physics with no added help.

If you do decide to go for a CPU upgrade, I really suggest you save for a new build. The old, slow DDR2 and the way obsolete AM2/AM2+ socket is going to always hold you back when it comes to CPU selection. And the selection will be limited to very used CPUs.
m
0
l
September 16, 2014 3:57:30 PM

clutchc said:
How the heck did you get a GTX750 Ti to be recognized on that old Asus M3A78-EM MB? The new Maxwell GPU is usually a no-go for old MBs. Even with their most recent BIOS version.

As the synthetic benchmark points out, the R9-270 should be and usually is a considerably faster card than the GTX 750 TI. I have both cards on my shelf right now and have tested with both. The 270 always beats the 750 Ti.
But I can't explain the difference you report between CPU/GPU usage in that game you linked to. The CPU should have been at a higher usage with the R9-270 than the 750 Ti. (??) How did the frame rates compare? Possibly that game was optimized for Nvidia drivers? Seems like a big difference, but that is all I can think of.

Are you positive all settings were the same with both cards? Nivdia may have had PhysX enabled, where AMD didn't because it has no PhysX. As a result, it had to do the CPU physics with no added help.

If you do decide to go for a CPU upgrade, I really suggest you save for a new build. The old, slow DDR2 and the way obsolete AM2/AM2+ socket is going to always hold you back when it comes to CPU selection. And the selection will be limited to very used CPUs.

Lmao, I have no idea man. I just plugged it in and it worked. I was surprised to find that the port was PCIE 2.0 and not 1.0.

It didn't make sense to me either, the R9 270 is obviously better than the 750 ti. It doesn't say whether it's optimized with Nvidia or AMD, but I think it has to be Nvidia, that's the only way it makes sense.

I went ahead and purchased the FX 6300, installed it, and ran the 6300+750 ti to these results: http://oi58.tinypic.com/2ce6fcw.jpg A much more stable FPS, and significantly higher.

In Sleeping Dogs it doesn't reach framerates as high as the Phenom 9550+270, but it's *far* more stable and playable: http://oi59.tinypic.com/317a900.jpg

Thanks for all the help. I'm much more happy with this route than keeping the old Phenom.
m
0
l
a c 418 à CPUs
September 16, 2014 6:20:53 PM

McDohl said:


I went ahead and purchased the FX 6300, installed it...

???
You installed the FX-6300 in what?
m
0
l
September 16, 2014 9:27:04 PM

clutchc said:
McDohl said:


I went ahead and purchased the FX 6300, installed it...

???
You installed the FX-6300 in what?

In the m3a78-em.

I'm just messing with ya haha. I got a GA-78LMT-USB3 motherboard and a ram stick as well.
m
0
l
!