GPU Not Performing?

SickSnappy

Honorable
Mar 9, 2013
37
0
10,530
So I recently purchased Battlefield 4 and have been playing some Hardline as well. The only problem is that I feel as if my performance is not up to par with what it should be... I was under the impression that a GTX 770 would be able to run Battlefield 4 on Ultra Preset in 1080p at 60 fps. Am I wrong? If so, what settings should I be able to run it at with a constant 60? I am currently getting anywhere from 25 frames during demanding points in the game, to 90 while walking down a hallway.

One thing I thought of was that maybe my FX-4300 (Overclocked to 4.4 GHz) was bottle necking the 770. But While playing, my CPU usage was running at about 90-100% using the Core Parking Manager tool set to 90%. My GPU usage fluctuated constantly. I thought that if your CPU was bottle necking it would always be running at 100% usage.

Even if I turn the graphical settings to the lowest preset, I get pretty much the same fps... My GPU Usage just lowers to about 40-50%. I didn't like this, so I reduced res scaling to 25% with next to no change in fps yet again... Usage drops to 20-30%.

Is there any way to fix this, or is my rig incapable of handling it? Thank you for any and all answers.

My PC:
Windows 7 64-bit
AMD FX-4300 Quad Core Processor 4.4GHz
GTX 770 MSI Twin Frozr
Asus M5A97 le r2.0 Mobo
8GB 1600MHz G.Skill Ripjaws x RAM
HX750 PSU
1TB WD Blue
 
Solution
If the CPU is running 90-100%, that's your problem. A CPU bottleneck can occur with 90% usage. It's pegged to the max, you need more CPU power.

SquirrelPhantom

Reputable
Mar 19, 2014
130
0
4,710
I don't think it's your gpu but more your cpu causing the problem. The game is very demanding on the cpu and n FX quad core is probably not meeting the needs he game requires even if it's overclocked. I believe the game does recommend a 6-core AMD processor for best game play.
 

wdmfiber

Honorable
Dec 7, 2012
810
0
11,160


It's actually a tough call. As FX CPU's are old and a weak design(piledriver).
However... you not building from scratch, you already have the board...

 

iron8orn

Admirable
1600mhz is not horrible but the fx is optimized for 1866mhz.

it is actually a measurement of threads.. a 4300 has 2 physical cores and 4 logical threads. a 6300 has 3 physical cores and 6 logical threads.

the fx is known to perform weakly with only 2 cores.

if you take a look in your bios and give me the speed your ram is running at and its values i can see if you can get to 1866mhz to test it out.
 
See the many many many threads on FX-4xxx people pissed because they could perform on BF3 but don't understand the difference with BF4. It is a COMPLETELY different game and yes demands way more then a FX-4xxx can handle (just google BF4 CPU benchmark). At this point 2013 and forward games are topping AMD at requiring nothing less then the FX-8xxx or switch to Intel i5/7 as AMD has abandoned any further gaming CPUs (their focus is has been solely since the FX-8xxx release sticking to the APU low power, low demand, low cost portable marketspace).
 

SickSnappy

Honorable
Mar 9, 2013
37
0
10,530




I will check that out and get back to you. thanks for your help.
 

SquirrelPhantom

Reputable
Mar 19, 2014
130
0
4,710


I can assure you that FX has two threads per module because each module has two cores.
 


INCORRECT. Hyperthreading is EXCLUSIVE proprietary technology patten only to INTEL. AMD solution was to provide more physical cores for MultiCore software, but sadly all software (except CAD or extreme software) is all coded for SINGLE core. What happens to AMD (sighs!!!) is the threading, the part which pulls data to then pass to the processing subcores, then pass back when complete, with so many cores 'standing in one line' all get in the way with each other that they start to actually kill performance / stand around twiddling thumbs because they can't "get" / "get rid of" the data they are processing efficiently as Intel has done. Then Intel did one more and made each 'core' actually perform like 2 cores, so while one core stands in line (remember single core processing) it takes 2x the work load and performs it, effectively doubling the 'processing' the less cores do (hence why a i7 8 'core' (4 cores 8 threads) performs 200% (twice) better then the FX-8xxx (true 8 cores) in all 'normal' applications (games).

IF AMD could code a 'addition' software to do the same performance and push it out to the current FX line, they could 1.5 / 2x the performance of all systems better than any Intel chip (i.e. your cheap FX-4xxx would be faster then a i7 finally). Sadely AMD has decided with it's ATI merger, the future profits (not gaming performance) lays in APUs, a combined CPU and GPU for 'low cost, low demand, low power' portable (read as Tablets/phones) solutions.
 

iron8orn

Admirable




not saying your wrong and i think i have read this argument somewhere before.

 

iron8orn

Admirable



thank you :) i have tried to sell the Kaveri with powerful ram several times but no one seems to bite.
 

4745454b

Titan
Moderator
There is little difference between intels "ht" and amds module approach. Amd threw more resources at the threads. I personally still believe their shared FP core needs more work. To say that amds modules are fine and intels HT isn't good for games shows how little you know. As games needs more cores/threads, even HT well show more gains then it did with older software.
 


http://en.wikipedia.org/wiki/AMD_FX refers to what your saying partially. The design is the two cores are inside a 'module', where as Intel does a single core and threads it into two processing.
Here is how AMD does it http://en.wikipedia.org/wiki/File:AMD_Bulldozer_block_diagram_(CPU_core_bloack).PNG
Where as Intel does it this way http://images.anandtech.com/reviews/cpu/intel/Haswell/Architecture/haswellexec.png
As well as this quote in Wiki http://en.wikipedia.org/wiki/Bulldozer_(microarchitecture)
The modular architecture consists of multithreaded shared L2 cache and FlexFPU, which uses simultaneous multithreading. Each physical integer core, two per module, is single threaded, in contrast with Intel's Hyperthreading, where two virtual simultaneous threads share the resources of a single physical core.[8]

The easiest way to say it (as I seen explained before is) AMD went Serial Processing for the CPUs, which if there was sofware coded to take each processor individually and use them (multicore software like CAD) then eash waiting 'in line' for the next use, it would be highly effective. Where as Intel does Parallel, so that as one core takes in two processing of data, it does twice the work at the same time a single core is actually being used.
 

4745454b

Titan
Moderator
I see where you are going with that, fit some reason it doesn't seem right. Intel first brought HTabout because the p4 architecture was so long that many parts of the chip were idle. Very inefficient. By adding registers intel was able to increase the effiency of their chips.

Amds solution was different as they wanted more cores to solve their IPC problem. They are on the right track because once they figure out how to use the igp as the fp core they should see a huge boost in performance. But to say that one is serialwhile the other is parallel seems off as they do the same thing in a slightly different way.
 

wdmfiber

Honorable
Dec 7, 2012
810
0
11,160

Essentially iron8orn is correct. People... stop reading the AMD box and open the task manager on a PC running a FX chip. AMD is a corporation that was recently very close to going bankrupt. They're marketing department will say anything to sell product.

To the operating system(example: FX 8350) is a 4 core CPU with 8 threads.

Initially when the these FX chips were new, Windows would treat them as advertised. Example: 8 core. It was a disaster. FX cores are weak, but the 2nd core in each module is especially weak. A patch so the cores(realistically... threads) weren't loaded sequentially helped a lot. But regardless, the architecture on FX chips is old and too weak for a gaming machine.

Hyper threading is different and really shouldn't even be in this discussion. But thinking about it and how an OS would assign "work" helps. Windows has too know if your running a real 8 core(Intel Xeon).

If you chip is "handicapped" by a module design; it's not going to be able to handle the assigned "work load".
 


Actually that is exactly how they work. AMD is doing things in like this
Module 1 (CPU1, CPU2), Module 2 (CPU3, CPU4) etc. <<<<>>>> Data
So each CPU has to wait it's turn on the "normal" software out there (Office, BF4, etc.), the exact definition of serial, one thing after another in a straight uniform line.

What AMD wants is
Module 1 (CPU1 + CPU2)<<<>>>DATA
Module 2 (CPU3 + CPU4)<<<>>>DATA2
etc. as provided by CAD and other High end (read thousands of dollars) software that does multithread coding. Each 'module' would whop the hell out of a Intel processor because your putting two cores (like back when we had Dual Core Servers as you remember) to process the data, but because the data comes in one 'land' no matter how many modules, the hardware is just blocking itself.

Where as Intel with HT does it this way for single thread applications:
CPU1 (Virtual Core1+Virtual Core2) <<<<DATA & DATA2>>>
CPU2 (Virtual Core3+Virtual Core4) <<<<DATA3 & DAT4>>>, etc.
So intel is doubling it's work in PARALLEL (two processes at the same time) as compared to AMD.
But when we try to apply this to multithread, the second virtual core (because DATA2 goes to CPU2 NOT CPU1) sits basically idle and the performance seen is really low when scored as compared to AMD
 

4745454b

Titan
Moderator
First we are getting WAY off track from what the OP is asking about. Second, I'm trying to follow what you are writing but I either don't understand, don't know what you mean by some symbols, or some combo/other thing.

What AMD wants is
Module 1 (CPU1 + CPU2)<<<>>>DATA
Module 2 (CPU3 + CPU4)<<<>>>DATA2

What AMD wants is great, but what we have is something else. Some tasks can be highly parallel. Encoding video is a great example. Others, no matter how much we want it to be otherwise, isn't. Sometimes you have to compute C (A + B) before you can add it to D. What I'm not understanding is why you wrote "Module 1 (CPU1 + CPU2)<<<>>>DATA" CPU 1 and CPU 2 are independent. In the example you gave, you should be able to work on DATA 1-4. As long as we are talking INT math. I said above that I'm not sure they have their FPU working correctly as I think in far to many cases it's still limited to DATA 1-2, and doesn't work enough as DATA 1-4 like they claim. If you have any links to what you are talking about I'd love to join in a "deep" conversation about it, but it should be in our own thread.
 

iron8orn

Admirable
Its all good, snappy private messaged me last night.
Ram was running at like 1400mhz and the next step up was like 1700mhz.. my first thought was a fsb overclock but it is just a odd mobo.
I got a stable cas 9 timing for him and the test went well.
As for the gaming performance of the fx4300.. well i think that has already been covered.