Why/How is the FX 8350 a bottleneck?

[TymmVonD]

 

I'm sitting between 120-200 fps with my 780's while I've never ever hit 99% on any gpu because 1) cpu bottleneck 2) fps cap of 200

 

 

Rust taking a 8350 to 100% doesnt sound normal. 

 

When I would join an empty server, the CPU usage would drop and the game would run right. On full servers, max load. That's at stock though. The 8350 was meant for OC'ing. So the bottleneck was my fault.

[cthomas1489]

My 290x's don't go to 100%. I'm kinda new at reading and monitoring this information but this is what it looks like during BF4.

 

[Buckley]

My 290x's don't go to 100%. I'm kinda new at reading and monitoring this information but this is what it looks like during BF4.

Some scenes are less intensive and also the pause/spawn screen shoots usage down but yours is all over the place, that's odd

Also use GPUZ for monitoring, its much much cleaner and easier to read imo

[Faa]

My 290x's don't go to 100%. I'm kinda new at reading and monitoring this information but this is what it looks like during BF4.

Monitor it without vsync. Because the gpu could be enough at 30% load to provide 60 fps (your cap) for a moment

[Apach3]

Any half decent $120+ CPU will give you a fantastic gaming experience these days. Most folks get i7s just because they can not really because they need them.

I have personally tested an FX 8320 & an i7 2600 with my own freaking hands & they performed the same in every single game I've benchmarked except one & even then I wasn't able to tell the difference in the gameplay experience.

Most folks don't care that much about their CPU purchase to research for more than a few days that's if they even end up doing any research at all & in the end the brand name & brand loyalty are still major factors.

Intel has always moved more product than AMD & will likely always will that doesn't mean that the AMD product is bad it just means that Intel runs a more successful business. Look at CryTek, they've made the most advanced graphics engine in the industry & the best looking games of all time & now they're facing serious financial perils it doesn't mean that CryTek is a bad studio, in fact I would argue that they're one of the best. You can have a great product & still have a tough time trying to sell it.

 

when i get my R9 290 in a few days, i'll run some tests with my own FX-8350 to add to this. i've already run tests with a GTX570 in Unigine's Valley benchmark; i can confirm that only when you get down to 1 module/2 cores running at 3GHz do you notice any slowdowns, and even then it's only seen during scene transitions and the intense weather situations.

[cthomas1489]

Monitor it without vsync. Because the gpu could be enough at 30% load to provide 60 fps (your cap) for a moment

 

V-sync is off.

 

 

Still looks all over the place after a few minutes of running around. Same with when I switch to the 2nd card.

 

Managed to capture it at 1% lol. It definitely flucuates in the higher end region though.

[i_build_nanosuits]

Just like Faa said, monitor your GPU load when you play games (VSYNC must be turned off!!) if it's not around 95% to 99% most of the time then it's sign of a bottleneck somewhere in the system, most likely a CPU bottleneck!

Some games are not as well optimised as other and in some of them it would be normal to have quick drop to 85% for example lasting a second or so and then peaking up at 98% again, this i would consider normal it's likely a micro stutter due to poor game optimisation, but if some of your games are running constantly in the bellow 90% than you are bottlenecked.

The actual load on the CPU does not indicate anything, a low CPU load won't mean it's processing the crucial data for the game fast enough, though most of the time a CPU at 100% load means it can't process any more data.

I used to run an FX-8320@4.6GHZ along with my GTX 780 and the CPU was still bottlenecking my card in many games...hence the reason i upgraded to an i7...a quick i5 would have done the same job as of now in current game but i'm scared of future multi-threaded game optimisation and if that occured i still i'm better covered with the i7 hence why i picked it!

when i get my R9 290 in a few days, i'll run some tests with my own FX-8350 to add to this. i've already run tests with a GTX570 in Unigine's Valley benchmark; i can confirm that only when you get down to 1 module/2 cores running at 3GHz do you notice any slowdowns, and even then it's only seen during scene transitions and the intense weather situations.

unigine is a GPU benchmark, it uses so little processing power from your CPU it is normal that you experienced that, you need real games to test CPUs...not GPU benchmarking tools...

[Faa]

V-sync is off.

 

 

 

Still looks all over the place after a few minutes of running around. Same with when I switch to the 2nd card.

 

Managed to capture it at 1% lol. It definitely flucuates in the higher end region though.

The gpu-z screenshot is with one card? The msi afterburner screenshot shows an average of 50-60% so you have tons of gpu headroom left to be filled.

Anyways this is a video I recorded when BF4 got out. Some dips to 80/80% but it was 90/90% consistently it didnt seem to be wanting to above 90% at all Started to raise resolution scale & POV somewhere in the middle of the video. Still havent figured it out why but I dont need SLI anymore to get 120 fps. I'm hitting the 200 fps cap at much lower loads though

 

[Buckley]

V-sync is off.

Still looks all over the place after a few minutes of running around. Same with when I switch to the 2nd card.

Managed to capture it at 1% lol. It definitely flucuates in the higher end region though.

Maybe your card just is shitting bricks and hates you

But idk man, mines always at 99% other than at spawn screen, I also use riva tuner with the in game overlay

[i_build_nanosuits]

personaly i use rivatuner within msi afterburner to show the GPU load while i'm playing games that way i can monitor it constantly but it kind of maked me a control freak on this i no longer play games i'm constantly monitoring stuff...

[Buckley]

personaly i use rivatuner within msi afterburner to show the GPU load while i'm playing games that way i can monitor it constantly but it kind of maked me a control freak on this i no longer play games i'm constantly monitoring stuff...

I put it in the lower right corner so its always in my view since I'm always constantly looking at my ammo count

[TechFan@ic]

V-sync is off.

 

 

Still looks all over the place after a few minutes of running around. Same with when I switch to the 2nd card.

 

Managed to capture it at 1% lol. It definitely flucuates in the higher end region though.

Disable CrossfireX via the Catalyst Control Center, if GPU usage goes significantly up in the same gameplay sequence then you probably have got a CPU bottleneck.

Now that isn't necessarily a bad thing, if you're already getting constant 60+ FPS then the CPU bottleneck is obviously not affecting your gaming experience and it also means that you can raise the graphics settings even more to take advantage of that GPU headroom.

[Faa]

Maybe your card just is shitting bricks and hates you

But idk man, mines always at 99% other than at spawn screen, I also use riva tuner with the in game overlay

Well lets say you're getting in a game averages of 30% and nothing more, just in a stupid old cpu bound game - if you add a 2nd card what basically will happen is each gpu would run at 15% so the load actually just splits but your performance is the same because you were cpu bound and not supposed to add another gpu but a better cpu/overclocking instead. Hence why I mentioned people getting 3 titan blacks for WoW (loads are probably 10/10/10%).. If you ever want to go CF/SLI for the love of god make sure you get 99% on your gpu or atleast very close to it or you wouldn't gain any performance. Now the misleading part; I had on a single gtx 670 99% consistently so I decided to add another what happened I didnt gain a fuck really just 50/50% and nothing more (often it went to 70/70% eg when scoping with a 12x scope). Got rid of the 2nd 670 the first day rofl.

His averages pulled from that Afterburner screenshot around 50-70%, that proves he gets 99% easily on a single 290x. Because you get 99% on a single, it's not guaranteed you'd push the max out of two cards He doesnt take full advantage of his CF config but 4K might burn them to 99% or 200% res scale at 1080p with a big hit on your performance.

In my experience, I dont play singleplayer games, I'm usually cpu bottlenecked I usually don't see the cards ever going above 50/50% so I just disable SLI for that game. Not only multiplayers are cpu bound, games like "lets release it now and we'll fix it later" are usually cpu bound to fuck like Watch dogs/Assassins creed black flag. Actually its been awhile I used my 2nd 780 for a game, it just sits there for aesthetics only.

Like nano said, the gpu load is the source of your performance - if its hitting 99% and you arent happy upgrade it or get a 2nd - if its not hitting close to that get a new cpu. You can calculate fps though with the gpu load so eg if you have 50% at 60 fps that means you'd get 120fps at 99% scales pretty much lineairly. 

 

 

Disable CrossfireX via the Catalyst Control Center, if GPU usage goes significantly up in the same gameplay sequence then you probably have got a CPU bottleneck.

That screenshot was with CF disabled. And since when does the cpu become a bottleneck if the gpu usage goes significantly up?

[Faceman]

@Buckley

 

I have an i5-4670k OC'd to 4.7Ghz + a single GTX780(not overclocked, lost the silicon lottery big time on my GPU) when I play BF4, I get 120-150fps.  My settings are Ultra everything, and Texture Resolution 140%.

 

I eventually turned V-Sync on because there is no sense running at more than 60fps on my monitor.

[cthomas1489]

Hmm so those first screens were from multiplayer but I just loaded up some single player screens. I'm watching the fps using 'perfoverlay.drawfps 1' from the console (correct me if I should be using fraps or something other).

 

Anyway, afterburner shows the same jaggedness in the GPU1 and GPU2 usage. It's never consistant but mostly up and down up and down.. Also my fps shown are not capped at 200 when on Ultra (which I would think they should be). I have crossfire enabled in CCC. When I run GPU-Z it still shows each card to pick from separately.

 

I'm going to try Skyrim real quick and see if the % are more consistant.

 

EDIT: Skyrim was a bad choice.

[Faceman]

@cthomas1489

Up and down, up and down is not good. When looking at benchmarks, highs are not important, its the lows. Lows are going to give you a baseline for consistent performance. It seems like your performance is jumping around.

Check out this review of the i3-4340. Read the entire article.

Here is one of my favorite quotes when talking about Assassin's Creed Black Flag

"This is a huge result – it wasn’t until we used a Haswell core CPU that the R9 280X was able to deliver consistent frame times and a 60 FPS frame rate in Assassin’s Creed IV. All three AMD CPUs we used – even the FX 8350 – and the Ivy Bridge Core i3 would deliver a sub 60 FPS frame rate, with frame spikes throughout the benchmark run."

@Buckley

Run Cinebench and do the OpenGL test. Let us know what your FPS is.

[cthomas1489]

@Buckley

Run Cinebench and do the OpenGL test. Let us know what your FPS is.

Will do.

So here's the cinebench test. I'm pretty sure it only tested one GPU though. It gave 87.09 FPS. The CPU test got 700.

[Faceman]

So here's the cinebench test. I'm pretty sure it only tested one GPU though. It gave 87.09 FPS. The CPU test got 700.

 

Ya..... your CPU is bottlenecking your GPU.

 

My i5-4670k @ 4.7 scores 702

 

For my GPU, on record it shows 127.31, but I have gotten it as high as 134.

 

When my i5 is at stock, my GPU score was 118, CPU score at stock was like 566 I think.

[cthomas1489]

Ya..... your CPU is bottlenecking your GPU.

 

My i5-4670k @ 4.7 scores 702

 

For my GPU, on record it shows 127.31, but I have gotten it as high as 134.

 

When my i5 is at stock, 3.4Ghz, my GPU score was 118.

 

Ugh okay.

[Faceman]

Ugh okay.

You are still over 60fps, so it doesn't matter.  But there is a bottleneck there.  The 290 and 780 are very close, the 780 *maybe* being 5% more powerful.  That is way too big of a score discrepancy when comparing a 290 and 780.

[cthomas1489]

You are still over 60fps, so it doesn't matter.  But there is a bottleneck there.  The 290 and 780 are very close, the 780 *maybe* being 5% more powerful.  That is way too big of a score discrepancy when comparing a 290 and 780.

 

Yeah but I have a feeling it's going to be a big problem when my 4k monitor gets here.

 

Another topic is that of the PCI-E lanes. AM3+ are still 2.0. The mobo I'm using allows x16 x 16 and there are plenty of posts showing that should not make too much a difference but I'm thinking it may factor in as well.

[Faceman]

Yeah but I have a feeling it's going to be a big problem when my 4k monitor gets here.

 

Another topic is that of the PCI-E lanes. AM3+ are still 2.0. The mobo I'm using allows x16 x 16 and there are plenty of posts showing that should not make too much a difference but I'm thinking it may factor in as well.

I am not too familiar with PCI-E, but from my browsing, I have heard that we aren't even making full use of 2.0, so I wouldn't worry about having 3.0.  When you do make the switch to 4k, your R9 290s *should* run better than the 780 because of the 4GB of VRAM, and all benchmarks I have seen show AMD cards scaling much better at higher resolutions.

 

I don't know for sure, but based on your current results, it is in your best interests to make the switch to Intel, an i5 or greater.  A single 290 is already being bottlenecked, so when you try and scale higher, that bottleneck will likely be more prominent. What you could do is buy the 4k monitor when you decide to make the switch and see if you are satisfied with the results.  If not, then you know what needs to be changed.

[Priller]

Will do.

So here's the cinebench test. I'm pretty sure it only tested one GPU though. It gave 87.09 FPS. The CPU test got 700.

 

You should be in the 720 range for that clock speed.

 

EDITED

[cthomas1489]

You should be in the 720 ram for that clock speed.

Hmm. So....what does that mean that it's not?

I got this in 3DMark

[FusionBlitzx]

That CPU isn't bottlenecking anything don't worry. Just because the CPU isn't at 100% load on all cores doesn't mean that it's bottlenecking your gpu, the 8350 can easily handle most gpu configurations but the most extreme ones