FX 8350 with R9 390 (will it bootleneck?)

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

lionheart051

Reputable
Sep 28, 2014
14
0
4,510
So will this cpu bottleneck a R9 390 gpu? I think i read somewhere, some time back it's ok with a R9 290x but since the R9 390 is a newer card i was wondering if this cpu will cause a bottleneck.
 
Solution
No, it won't bottleneck.

Intel i5 and i7 and also FX 6300 and 8300 series CPUs won't bottleneck a single high end GPU in modern, multi-threaded games.

They also won't bottleneck most 2nd tier Crossfire/SLI setups either, such as R9 280 / HD 7950.

FX 6300 with dual R9 390 or GTX 980ti will start to have a bottleneck in some games.


Example: I had an old Phenom II x6 1090T @ 4.0GHz that would fully load both of my HD 7950 GHz ed. (R9 280) in crossfire with Crysis 3 on max details except for 2x SMAA (* Tom's recommended AA mode for best performance/appearance balance). CPU would be around 90-95% usage on all 6 cores. GPUs would both be at 95-99% usage. No bottleneck, well balanced, average FPS in the 50s.

Midrange OC of 4.4-4.6GHz...

jkteddy77

Honorable
Jun 13, 2013
1,131
0
11,360
As someone who spent HUNDREDS of hours and nearly 50 threads on here nearly 2 years ago on the 290's release, let me shed some light here.
In certain titles, the 8350 bottles the 290 a ton, drops it down to nearly 70% GPU usage in GPU bound games (biggest problem was BF games) That's REALLY bad, you are only getting 70% of the performance you paid for!!! May as well have just bought a slower GPU.
In nearly all games you'll see 5 fps loss vs an i7, and maybe lower minimums, but in other games you can see massive drops.
In certain bottled games, I saw 10fps less on average and saw minimums 25fps lower than I get now with my Intel.
Not to mention turning down the settings didn't give me much more FPS than Ultra did, this is a VERY significant sign of a CPU bottleneck if the game is GPU bound (today, I'd say 95% or more of titles are).

Here is some actual data I RECORDED, not just word of mouth here. Absolutely Unbiased footage, unlike the very varying results I see linked everyday and like in the images above. This isn't something a benchmark can solve, since every game require different amounts of CPU resources (not to mention most benchamrks aren't that CPU intensive in the first place)
Here is some Real World bottlenecking.
Pay attention to GPU and CPU graphs as well as usage levels.

https://www.youtube.com/watch?v=DcTPLMuQ610
https://www.youtube.com/watch?v=Aq9dSLOElX4
https://www.youtube.com/watch?v=IaVKMNL-aS4

In GPU bound games, if your GPU is ever lower than 99% usage with Vsync off, you ARE bottlenecking, and I saw that routinely with the 8350. I show this reading often throughout the videos, and some parts even have it live overlayed on the screen.

If you want me to explain what some of the graphs mean I can, but basically as seen in the gameplay graph in the bottom left corner, the yellow(CPU) and GPU(green) should be right on top of each other. If one is higher than the other, that part is bottling. It is holding the other back, and is a true bottleneck. In the third video I even locked the framerate so the graph more clearly and smoothly showed this correlation. Even though when locked both parts were not at full usage in this video because of vsync, their usages should still match up with each other, which the graph clearly showed they did not.

Also note that the more pixels you are pushing, the less bottlenecking you will experience. Pushing more pixels or higher resolution textures give the GPU more to do before it receives its next instructions from the CPU, effectively alleviating the bottleneck.
If you're in 1600p or 2160p you don't have as large of a problem, the GPU is being very stressed, and the slower CPU becomes less of a problem because the GPU is more preoccupied. But in 1080p? There is not enough for the GPU to render before it craves new instructions of what to render next from the CPU. If the CPU isn't keeping up, the GPU starts to idle, hence it drops GPU usage, and it's not rendering it's maximum potential amount of frames, hence the FPS performance drop. That is the science of bottlenecking.

Since I got my 4790k, I don't ever see lower than 97-99%, it fixed any low FPS issue I was having in all of my games (tested nearly 300 games) OVERNIGHT

So if I had trouble 2 years ago, I am ssuming today's games still aren't faring any better, if not worse with this combination of hardware.

If you can live with bottled performance to cut money, grab an 8320/8350, but I highly recommend an i5 over one anyday
Maybe Zen will be better???
 

g1abhi

Expert
Ambassador
Jun 13, 2015
958
0
11,360


THANK YOU , for the data , I wish people would just calmly analyse this .
 

jkteddy77

Honorable
Jun 13, 2013
1,131
0
11,360


Don't mention it, I love being a help to people on this very cloudy subject, and hope nobody else out there goes through the stress I put myself through all those months trying to figure out why my hardware was behaving as it did. Far too many people blindly blame drivers, or even blame the user themselves for a more common than recognized hardware conflict.
 

Maximus305

Reputable
Jan 7, 2016
2
0
4,510
I have a system with a FX 8350 and 2x R9 390 and i can tell you that with a single card it doesn't bottleneck but in crossfire its a nightmare. In crossfire the GPUs start to downclock because the CPU cant handle it. I will soon jump to a I7, cause now i am starting to notice bottleneck in my system.
 

port27

Reputable
Dec 15, 2015
111
0
4,710
You will regret using that CPU with that GPU. I recently tried talking a friend out of that combo and he went ahead anyway and lets just say hes already upgrading. FX 8350 is good for computing its not ideal for gaming whatsoever. You can get away with it in some games until that game calls for your CPU to kick in.
 

jkteddy77

Honorable
Jun 13, 2013
1,131
0
11,360


You are likely bottlenecked already, you just may not be noticing the performance you are losing.
Check GPU usage, I bet even the single card dips below 99% more often than you think.
In all games is it a problem? no... but when a game DOES require more CPU than the 8350 can give, it turns into a very large problem, especially if it was already a game that you needed all of your GPU's speed to run smoothly in the first place.

port 27 is correct down below, all of my other friends still kept their 8350, and they still are mad that they get nearly 20-30fps less than me in some titles with the same single R9 290.


 

Bem-xxx

Reputable
Sep 20, 2015
163
0
4,710
A FX 8 core won't bottleneck anything.
Here a FX 9590 vs i7 4790X and Titan X SLI at 4K. Both chips perform the same.
https://www.youtube.com/watch?v=wJeESA2JWdM

At any resolution a FX 8370 will perform like an i7 5960X in most games.
http://www.technologyx.com/featured/amd-vs-intel-our-8-core-cpu-gaming-performance-showdown/?hootPostID=282b782e2a8d67e68bbb50aeea079e96

At 4K a FX 8350 performs like or better than an i7.
http://www.tweaktown.com/tweakipedia/56/amd-fx-8350-powering-gtx-780-sli-vs-gtx-980-sli-at-4k/index.html
 

Bem-xxx

Reputable
Sep 20, 2015
163
0
4,710


Drivers downclock the GPU core. Try playing Ryse with Catalyst 15.9.1 and then with the latest Crimson. Crimson will downclock the R9 390 to 400-300Mhz...

The FX 9590 can handle a Titan X SLI flawlessly, no reason for the FX 8350 to underperform with a R9 390 Crossfire.
 

jkteddy77

Honorable
Jun 13, 2013
1,131
0
11,360


Bro you are NOT understanding that bottlenecks more often happen at lower resolutions. At 1080p, this is a very real bottlenecking hardware combination.
You didn't even link to 1080p benchmarks of the R9 290/R9 390 paired with an FX 8-core. The only benchmark in 10870p that you posted was with a much slower 960, which is far less likely to bottlneck compared the the superior 390.

Of course in 4k there is no bottleneck, the GPU has to render more pixels in each frame, slowing its need for instructions from the CPU, thus a slower CPU like the 8350 is enough to keep the R9 290 busy when it spends longer rendering each frame.
But in 1080p, the R9 290 has a much easier time rendering each frame since there are less pixels, so the R9 290 runs out of instructions since it has a much easier time rendering each frame in 1K.
Since the CPU isn't good enough to keep pumping instructions to the speedily running R9 290, it drops GPU usage, and cannot render more frames until the information slowly reaches it. Thus, less frames peer second are rendered.

Think of it like the CPU being the architect, and the GPU being the construction workers. The GPU can't build the house if the architect is designing it too slowly for them to continue. Under a faster architect, the construction workers could build the house much faster.
If the house is very difficult to build, but easy to design (4k) then even though the architect (CPU)is slow at designing, the workers (GPU) are kept busy enough for it not to matter, and they never have to slow down.

Also realize that bottlenecks hardly ever show up in set benchmarks, they happen in real world gaming, especially in hard to run multiplayer games with many players and object to render, as this requires more CPU performance.
THAT is why I posted my ACTUAL OWN gameplay of many situations where a bottleneck is occuring...
GPU usage drops, CPU usage skyrockets, and overall fps lessens greatly

PLEASE
Fully understand why a bottleneck happens before giving out false information,
And especially don't tell the rest of us we're wrong when we have personal recordings and data, as well as experience with this exact hardware combination to support our theories before you come in throwing some random online benchmarks you don't understand at us.
Very disrespectful.
You're just the blind attempting to lead the blind. Hope I taught you something about bottlenecks today.
 

OpTic_STinGsHoT

Reputable
Feb 10, 2016
3
0
4,510


http://gpuboss.com/gpus/Radeon-R9-390-vs-GeForce-GTX-970
 

g1abhi

Expert
Ambassador
Jun 13, 2015
958
0
11,360


Can you read "Real Game Benchmark"??????????????

i can throw stuff too : http://gpu.userbenchmark.com/Compare/Nvidia-GTX-970-vs-AMD-R9-390/2577vs3481