Bottleneck or Benchmark?

[Raves]

unigine heaven 2 will probably not reveal a cpu bottleneck - it is pretty much a gpu bench.

if final fantasy is the game you play, and performance is poor with a 470, then it is cpu bound. really im surprised there is much debate, i found a 5850 with a e8500 at 3.8ghz was slightly bottlenecking the gpu.

i'd recommend getting a q9400 similar and oc'n to 3.6.

 

[HangTime]

Hey, just reading this I think there might be something up with your CPU as I also have an e6300 and stock is 2.8ghz not 1.8ghz? Unless I'm missing something that's only a 100mhz OC your running?

Some genius at Intel decided to release a different CPU with the same name as an old one but different specs. You've got the new E6300, he's got the old one.

 

[M0T]

Updated my drivers and now get 2900 on the FF benchmark which is a bit better, hopefully the game will now be playable on my rig.

 

[aqon]

you're gpu will feed better off a core 2 duo at around 4ghz ' ish rather than a quad core at 3.6ghz. At the minute gpu's need raw speed not more cores. This has been tested.
Im still with an old evga 680sli board and c2d e8500 @ 4.2ghz and its still going strong.

 

[Marine-RX179]

you're gpu will feed better off a core 2 duo at around 4ghz ' ish rather than a quad core at 3.6ghz. At the minute gpu's need raw speed not more cores. This has been tested.

Actually that only reflect on games that only use 2 cores (i.e. Crysis), whereas games that are written to properly use more than two cores, Quad would always have upper hand, especially when paired with a (pair of) fast graphic card. But it's also true that most mmos do not use more than two cores.

 

[drunkenmaster]

While I don't have specific data for GTX470, look at the following data for different CPUs pairing with a 5870 running Crysis:
http://www.bit-tech.net/hardware/cpus/2010/03/03/overclocking-intel-core-i3-530/8
If you comparing the results of E8400 at 3.0GHz to E8400 at 4.25GHz, you can see that the it is minimum 19fps average 36fps vs minimum 27fps average 48fps. Granted not all games are as demanding as Crysis, but a fast graphic card still need to be pair with a decent CPU...and the minimum being consider as decent enough nowadays pair with cards of GTX460 1GB and above has to be either a 3.6-4GHz+ dual-core, or a 3.0GHz+ quad-core.

Do ignore the results for Q6600 on Crysis though, as the game only use two cores, and the Q6600 only has two of its four cores utilized.

Also, the CPU bottleneck is clear, since he only got 2300 mark on the FFXIV bench on low with a GTX470, while I get 4000 mark on low with just a 9800GTX+.

http://www.bit-tech.net/hardware/graphics/2010/08/13/asus-radeon-hd-5870-g-v2-review/5

3.2Ghz cpu, the problem being that, when you play to the settings in the game you're likely to use, those high fps aren't available, because you become GPU limited.

Thats the problem with your assumption, the reviewer has PURPOSEFULLY used LOWER THAN USUAL GPU settings, to highlight the difference between CPU's, thats it, in "real world" situations, you use higher settings and the game stops being cpu limited.

Crysis is a game where 30fps average feels smooth, and 33fps minimums simply aren't required, or will likely be achieved.

 

[aqon]

Dont get me wrong quad core cpu's are great in principal but clock for clock there isnt a lot of difference. Graphics cards are that quick these days, they need raw power, rather than more cores. soz for goin on , bit tipsy.

 

[Rroff]

Control panel settings? wouldn't be the first time I've heard someone complaining about poor GPU performance compared to their old card only to find they had silly levels of AA set in the control panel or similiar.

 

[Duff-Man]

everything u wrote. just no.

Care to elaborate on that, or are you just being obstinate?

What he said is very valid. With few exceptions, games do not fully utilise more than two CPU threads, and as a result, CPU clockspeed is more important than the number of cores. If you believe otherwise then why not back up your viewpoint with data, rather than just saying "no"?

 

[Duff-Man]

No need to reiterate and oversimplify what has already been covered in this thread.

I completely disagree with aqon's unsupported opinion, hence all that is required in response is 'no'.

This thread is about a modern game, and the OP has a 470. IMO his gpu and gaming requirements will be better met by a quad at 3.6 (or higher) than a dual core at 4.0.

Just brilliant... You slam another member for producing an unsupported opinion, and then proceed to do the exact same thing yourself.

Unless you want to produce evidence to the contrary, your baseless opinion is pretty much irrelevant.


...To bring this debate away from pointless opinionated conjecture, take a look at these comparisons:

Guru-3D
- Compare the 3Ghz QX6850 with the 3Ghz E8400. Not a great deal of difference, particularly as you crank up the resolution to realistic levels. Note several pages to this review.

Alienbabeltech
Again, very little difference between dual and quad core CPUs. Unsurprisingly, by far the biggest difference comes from the choice of GPU. Plenty of other pages, but note the e8400 at 4.25Ghz beating out the Q9550 at 4Ghz.

 

[gurusan]

Raves, with the exception of a few games like BC2 there is very little difference between a dual and quadcore in games.

I think FFXIV is optimized for over 2 cores as well, though how much of this actually affects gameplay is another story as there are TONS of multithreaded games that show virtually zero realized benefit from more than 2 cores (UT3 based games, Cryengine games, etc).

So raves, your answer of "no", is pretty silly.

I personally would go for a wolfdale or yorkfield clocked to 4GHz. Otherwise having a GTX 470 with a 2.9GHz Conroe is going to provide very inconsistent performance depending on the game, resolution, and graphical settings.

 

[Raves]

Just brilliant... You slam another member for producing an unsupported opinion, and then proceed to do the exact same thing yourself.

Slam? I disagreed with his opinion. Since it was only an opinion, I only offered mine.

Chill out.

If you guys really want to recommend a 4.0ghz dualie, over a 3.6 quad... that's your opinion. I disagree with your opinions. Clock speed is dependent on architecture too, which complicates the discussion even more.

I could dig up of benches demonstrating my position, and can dig up benches showing yours, but in the end it will come down to the app that the user wants to most use, and getting the tech balance right.

Bottomline the OP is cpu bound in his game and he has a 470. What would you recommend as the best solution to his problem? Me, I say a quad clocked past 3.6 ghz.

 

[Marine-RX179]

http://www.bit-tech.net/hardware/graphics/2010/08/13/asus-radeon-hd-5870-g-v2-review/5

3.2Ghz cpu, the problem being that, when you play to the settings in the game you're likely to use, those high fps aren't available, because you become GPU limited.

Thats the problem with your assumption, the reviewer has PURPOSEFULLY used LOWER THAN USUAL GPU settings, to highlight the difference between CPU's, thats it, in "real world" situations, you use higher settings and the game stops being cpu limited.

Crysis is a game where 30fps average feels smooth, and 33fps minimums simply aren't required, or will likely be achieved.

I understand what you are saying, like if with AA and stuff, the frame rate difference vary between different CPU would be reduce I agree with that. But as I mentioned before, there are games that will hit CPU limit before the GPU, such as mmos. Most of them are not graphic demanding, however because of they are not written to use more than two cores, the CPU would bottleneck and GPU usage simply won't reach near 100%.

Sure you can use a GTX470 with a 3.0GHz dual-core, but "ideally" it's better to use at least a 4.0GHz dual-core or 3.0GHz Quad-core...as it is much more likely to run into game scenes where the CPU hit the roof than with a faster CPU.

Sure the FFXIV bench at low is only running at 1280x720 no AA, but my mark with a Q6600@3GHz and 9800GTX+ is 3400, where OP's mark is with the [email protected] with GTX470 only score 2300, so unless I'm mistaken, the limitation pretty much point to the CPU (unless there are other errors/problem that are the cause of GTX470 underperforming).

 

[Duff-Man]

Slam? I disagreed with his opinion. Since it was only an opinion, I only offered mine.

Chill out.

If you guys really want to recommend a 4.0ghz dualie, over a 3.6 quad... that's your opinion. I disagree with your opinions. Clock speed is dependent on architecture too, which complicates the discussion even more.

Firstly, you did not offer your opinion. Until I called you on it, your only response was "no". This was a completely dismissive response to an arguably valid viewpoint, and was simply uncalled for, as well as being lazy. You mention that you "could dig up benchmarks" to support your position; well I suggest that next time you do just this. Single word responses are rather petty and certainly not constructive, but benchmarks provide solid information.

Secondly, I am not recommending any CPU. As far as I can see the OP has not asked for one. At this stage we are simply trying to ascertain whether the GTX470 is working as expected, and if so, what degree of CPU limitation is present. When the thread is not being derailed by trolling that is... If you want to assist in this, then great. If not, there are countless other threads where you can discuss the merits of dual vs quad core CPUs for all sorts of applications (check the CPU forum).

 

[Marine-RX179]

FSecondly, I am not recommending any CPU. As far as I can see the OP has not asked for one. At this stage we are simply trying to ascertain whether the GTX470 is working as expected, and if so, what degree of CPU limitation is present. When the thread is not being derailed by trolling that is... If you want to assist in this, then great. If not, there are countless other threads where you can discuss the merits of dual vs quad core CPUs for all sorts of applications (check the CPU forum).

So you simply choose what you want to see and ignore what you 'don't want to see' and call whoever not agree with your line of thought a troll?

I can tell you, with a system that has a GTX470 only scoring 2300 on the FFXIV bench on low, there's something definitely wrong. Besides me scoring 3400 on it, my friend with a stock speed Phenom II 555 unlocked to Quad with a 4890 score between 3000~4000 as well.

 

[minicle]

Have a good read of this thread on the cpu forum, http://forums.overclockers.co.uk/showthread.php?t=17929621

Which is a very good comparison of several c2d cpus at stock and o/c'd against various benchmarks.

My gut feeling is that you are bottlenecked by the speed of your cpu. Maybe take the time to run a few 3dmark benchies at various degrees of stock and oc'd and see where you end up?

 

[Duff-Man]

So you simply choose what you want to see and ignore what you 'don't want to see' and call whoever not agree with your line of thought a troll?

Not at all. Quite the opposite in fact. If you read my previous posts anywhere on this forum you will see that I strive only for objective reasoning. I own an i7-930, so to suggest I am some kind of dual-core evangelist would be incorrect.

I consider single-word dismissive responses to arguably valid viewpoints to be trolling. That is all.


And I agree that there is something "definitely wrong" with the benchmarks that the OP is posting. I have said this earlier in the thread, hence my suggestion that he run various GPU-restricted benchmarks (i.e. heaven and furmark) to see whether the GPU itself is working as expected. Once we have eliminated the GPU as faulty, or the drivers as badly installed, he can perform more analytical tests to determine the degree of CPU limitation. From this we can begin to draw conclusions that are based on hard facts, rather than on conjecture and opinion.


edit: I suggest we get back to this, rather than taking the thread further off-topic. As I said, there are plenty of other threads on the dual-core vs quad-core debate. It has been pretty much "done to death".

 

[Marine-RX179]

And I agree that there is something "definitely wrong" with the benchmarks that the OP is posting. I have said this earlier in the thread, hence my suggestion that he run various GPU-restricted benchmarks (i.e. heaven and furmark) to see whether the GPU itself is working as expected. Once we have eliminated the GPU as faulty, or the drivers as badly installed, he can perform more analytical tests to determine the degree of CPU limitation. From this we can begin to draw conclusions that are based on hard facts, rather than on conjecture and opinion

Fair enough if you put it this way.

 

[Duff-Man]

No.

See post #21.

So you made an unrelated post earlier in the thread... That doesn't change the fact that you were abrupt and dismissive of someone making a very valid point. And a brand-new member at that... There is simply no need for such trolling.

Now, can we please just get back to the topic at hand?



M0T; if you are reporting 99% GPU usage and 84C temperatures during the heaven benchmark, that suggests that the GPU is working at close to maximum capacity (as it should be). However your result is still lower than would be expected from a GTX470 in a GPU-restricted benchmark. Can you check in GPU-z that the GPU has flashed properly, and is reporting as a GTX470? (if you haven't already done so).

Secondly, to assess the degree of CPU restriction you can lower your CPU clock by a few hundred Mhz and check how your scores are affected. This will allow you to estimate the scaling you will gain from improved CPU speed. If it is close to linear with CPU speed, you have a solid answer on your CPU bottlenecking question.

 

[M0T]

What should I be getting on the heaven benchmark at 1680x1050 16af 4aa and extreme tesselation?