Forum Settings
       
This Forum is Read Only

Performance Question/OC'ingFollow

#1 Sep 15 2010 at 4:26 PM Rating: Good
Avatar
***
2,448 posts
Okay, I know the PC stuff has been thrown around a lot lately. But I wanted to get the opinion of some of the more tech-savvy peeps here.

I have moderate knowledge with PC stuff. I know a good performance/$ when I see it for the most part, and am comfortable putting PC's together.

I put together my current rig for the game:
Biostar TA890GXE mobo
AMD 1055t x6 processor
4gb 1066 ddr @ 7-7-7-20
Corsair F40 SSD boot drive/ffxiv
Asus GTX 460 1gb TOP edition
Seasonic 520w Breeze - bronze certified
Coolermaster Hyper 212+ HS/F

anyway, to the real question.

With this rig at stock clocks, (2.8ghz cpu, 775/2000 gpu) I get roughly 3600 on the benchmark 720p. 2700~ on the 1080p high.
I overclocked the CPU successfully to 3.78ghz(270x14 locked multi) but kept the ram same speed and NB just around 2100~.
I was also able to OC my gpu further to 850/2100, but I know it can go higher. That's not the point at this moment in time though.

With both gpu/cpu oc'd, I score a whopping increase of 4700 on the low res bench. Have yet to try higher one. For a curious george, I put my GPU back to stock 775/2000 and ran low res bench again. 4500.

So my question is...what does this all mean? %clock increase wise, do i get more out of the CPU or GPU being OC'd? Another thing that irks me is the benchmark will push GPU usage up to 70-90% according to MSI afterburner, but will only peak around 54-55% in the beta. Even when I run through town and fps dips to 27~ish.(20-25 if I turn AA up past x4) Does this usage cap in-game effect the amount of performance increase I get in OCing the GPU?

____________________________
Currently Playing: FFXIV:ARR
Lacaan Vasiim:Cactuar
Free Company:Cactuar Corp<CCorp>
catwho wrote:
If you need a bard to get "good exp" in a merit party, you're the weakest link.
#2 Sep 15 2010 at 4:36 PM Rating: Good
Sage
***
1,675 posts
For low 720p the CPU OC gets more gains; as it is CPU dependent. High 1080p is more GPU dependent.

Post your high score with the OC. I'm willing to bet its around 3500.

So OC your GPU and you'll see more of an increase in high with BOTH OC'd CPU and GPU OC'd.


Example: My 8800GT and Q9550 (OC'd of course) are doing really well in beta. Conversely I get 3700 LOW and 2000 HIGH on the benchmark. On low my CPU is maxed out at 3.4ghz I DO NOT see any gains on LOW with higher CPU OCs. (Partially meaning that my GPU cannot take advantage of the higher clocks) However I do see higher gains when I OC my GPU when my CPU is at 3.4ghz.

So on LOW if I were to buy let's say a 5850 GPU (combined with my 9550) it would raise my score to about 4500, maybe 5000 (Especially IF I OC'd my 9550 to 4.0ghz) and if I was lucky 3000-3500 on HIGH.

This is probably the most confusing post I've posted in a while, sorry!


Edited, Sep 15th 2010 6:46pm by Kierk
#3 Sep 15 2010 at 4:36 PM Rating: Good
Sage
*
56 posts
The benchmark program will push your system harder than the actual in-game gameplay, in order to get a good understanding of what your PC can handle. Just because the game isn't using all the GPU power, it doesn't mean it's doing anything wrong or that the game is capping it. I don't think there's a cap in-game regarding performance, it just depends on what your system can handle.

As to your question about what has the most impact, I've seen numerous times people say this game uses more CPU power than graphics power, so yeah, you would see a better score with a higher CPU speed.

I don't see any particular problems you're having, though. Those OCs seem pretty good, and if they give you that much of an increase, then awesome, stick with them and be proud :)
#4 Sep 15 2010 at 4:39 PM Rating: Good
**
723 posts
I have an SLI setup, and while digging around how to get SLI to actually work during this beta I read something that I can't source at the moment, but it said that the benchmark actually is quite bad and doesn't translate properly. So you can score low on it and have a decent FPS in game, or score high and lag, etc. I wouldn't go based on that benchmark.

Your results however, if you did want to look more deeply into them, would indicate that your CPU is bottlenecking your PC. I highly doubt that's the case though. I have 2 GTX 285's in my system, and now even with SLI working they're both at 70-80% usage (Seen using eVGA's precision tool). When I could only get 1 GPU to work, it was almost always at 100%.

What most people have been doing is letting Fraps run to check your FPS in a particular spot, make whatever overclock changes you want then check the same exact spot again. I ultimately got my PC to a point where it runs smooth at all time, and I'm very satisfied with how the game looks. I hope you can get there too, good luck!

Edit: Isn't 520W kind of pushing it in terms of wattage for your power supply? It's probably okay, but I'd personally feel more comfortable with a high watt PSU.

Edited, Sep 15th 2010 6:42pm by Jeraziah
#5 Sep 15 2010 at 4:54 PM Rating: Good
Avatar
***
2,448 posts
One interesting note: Turning on AA and/or putting AA to a higher setting raises GPU usage. I just put on x8MSAA-Q and distant landscape checkbox thing, logged in, and GPU jumped up to 100% and didn't drop. I was in the field during a sand storm by camp drybone, staring at a lone dodo. fraps read a steady 50fps, with spikes up to 55 and dips to 47.

Running hi res bench w/ OC settings...
(side note while bench is running- I play at 1600x1050 so the game fits all but one side of the screen to watch IM windows, gpu-z, taskmanager, afterburner and whatnot.)

2995. But, I realized a mistake in my OP. I now clearly remember the high res bench being around 2300-2400 at stock clocks. I took screenshots, but unfortunately erased them when I moved my installations over to my SSD and reformatted the 500gb drive.

However I am quite happy with the results. I think I can push the graphics card a bit higher, but want to monitor temps first a little more.

Thanks for the help guys. I just needed to get my thoughts organized and get second opinions. Any other thoughts or tips??

Edit: The 520w is actually fine for the system and for stock speeds. Because of it's performance in efficiency at least. I might upgrade the mobo and PSU in the future once bulldozer comes out. I wondered about how the PSU would handle the system OC'd but I haven't had any stability issues. I don't know of any free way to monitor this though.

2nd edit: I believe if you calculate the CPU at 125w and the GPU the same(not sure what it's rated for but I assume its in the same area, maybe 150w max) then the mobo, 1 dvd drive, 3 case fans, 1ssd 1hdd shouldn't add up to more than 400ish. The PSU itself is actually highly rated from johnny guru, a somewhat legend of sorts for stressing PSU reviews.
Edited, Sep 15th 2010 5:57pm by FenrirXIII

Edited, Sep 15th 2010 6:00pm by FenrirXIII
____________________________
Currently Playing: FFXIV:ARR
Lacaan Vasiim:Cactuar
Free Company:Cactuar Corp<CCorp>
catwho wrote:
If you need a bard to get "good exp" in a merit party, you're the weakest link.
#6 Sep 15 2010 at 4:59 PM Rating: Good
**
723 posts
FenrirXIII wrote:
Any other thoughts or tips??


Honestly, the game is poorly optimized. Some very high end systems can't even max the game out. While this is a bad thing, it gives me hope that as time advances either SE will do something about it or nVidia/ATI come out with better drivers for the game.

At the moment my advice to you and everyone is to find a FPS you're comfortable with, then just toy with your graphics till you can get the performance you want and the image quality you want. Everyone is different, I'm one of those people who needs 50-60 FPS or it bothers me. While I don't have to sacrifice image quality with my system specs, if I had a lesser system I would. So yeah, just find the sweet spot so you're ready to play in a week and hope something gets better to make more use of your hardware.
#7 Sep 15 2010 at 5:18 PM Rating: Good
**
472 posts
I haven't OC'd in several years, but I used to do that to every machine I previously owned. I recommend that you run a burn in test once you have the settings you want, and see if it is actually stable. Oh and I think it has a built in measurement of your system, so you could run it and save the file and then compare to the OC'd settings.

http://www.passmark.com/products/bit.htm

"BurnInTest will bring intermittent or hidden problems to the surface so that after a successfully run the computer can be used with a much higher level of confidence. BurnInTest can also be used by overclockers to verify system stability at high clock speeds."
#8 Sep 15 2010 at 5:30 PM Rating: Good
**
723 posts
Parsalyn wrote:
I haven't OC'd in several years, but I used to do that to every machine I previously owned. I recommend that you run a burn in test once you have the settings you want, and see if it is actually stable. Oh and I think it has a built in measurement of your system, so you could run it and save the file and then compare to the OC'd settings.

http://www.passmark.com/products/bit.htm

"BurnInTest will bring intermittent or hidden problems to the surface so that after a successfully run the computer can be used with a much higher level of confidence. BurnInTest can also be used by overclockers to verify system stability at high clock speeds."


This is good advice, I kind of assumed the OP knew about this already since most of his questions don't relate to the process of overclocking itself. Prime95 is one of the best torture testing programs if you need something to test a CPU's stability. Furmark is one of the better GPU testers.
#9 Sep 15 2010 at 5:35 PM Rating: Good
**
472 posts
Jeraziah wrote:
Parsalyn wrote:
I haven't OC'd in several years, but I used to do that to every machine I previously owned. I recommend that you run a burn in test once you have the settings you want, and see if it is actually stable. Oh and I think it has a built in measurement of your system, so you could run it and save the file and then compare to the OC'd settings.

http://www.passmark.com/products/bit.htm

"BurnInTest will bring intermittent or hidden problems to the surface so that after a successfully run the computer can be used with a much higher level of confidence. BurnInTest can also be used by overclockers to verify system stability at high clock speeds."


This is good advice, I kind of assumed the OP knew about this already since most of his questions don't relate to the process of overclocking itself. Prime95 is one of the best torture testing programs if you need something to test a CPU's stability. Furmark is one of the better GPU testers.


Prime95, I haven't heard of that one, I'll have to get my geek tools out and go find it :). Love stuff like that for computers. Furmark or futuremark?

If he does know about all this, anyone else that is viewing this topic: If your computer stops working during a burnin test, stalls, reboots, or catches on fire, you might have a problem.
#10 Sep 15 2010 at 5:40 PM Rating: Good
Sage
**
743 posts
Parsalyn wrote:
Jeraziah wrote:
Parsalyn wrote:
I haven't OC'd in several years, but I used to do that to every machine I previously owned. I recommend that you run a burn in test once you have the settings you want, and see if it is actually stable. Oh and I think it has a built in measurement of your system, so you could run it and save the file and then compare to the OC'd settings.

http://www.passmark.com/products/bit.htm

"BurnInTest will bring intermittent or hidden problems to the surface so that after a successfully run the computer can be used with a much higher level of confidence. BurnInTest can also be used by overclockers to verify system stability at high clock speeds."


This is good advice, I kind of assumed the OP knew about this already since most of his questions don't relate to the process of overclocking itself. Prime95 is one of the best torture testing programs if you need something to test a CPU's stability. Furmark is one of the better GPU testers.


Prime95, I haven't heard of that one, I'll have to get my geek tools out and go find it :). Love stuff like that for computers. Furmark or futuremark?

If he does know about all this, anyone else that is viewing this topic: If your computer stops working during a burnin test, stalls, reboots, or catches on fire, you might have a problem.

Probably a good idea to monitor your temps while running Prime95.
____________________________
I think you've been smoking the Moko...
http://na.finalfantasyxiv.com/lodestone/character/350413/
http://na.finalfantasyxiv.com/lodestone/character/1628942/
http://www.nerdist.com/
Angus of Cerberus (retired)
#11 Sep 15 2010 at 8:18 PM Rating: Default
**
412 posts
I would not use Furmark to test for stability.
Furmark pushes your GPU to insane temperatures. A lot higher than what a game does. I'm talking 20C degrees higher.
I don't think it's worth trying to stabilize it on furmark for video games.
I suggest ATITool or MSI Kombustor. Very close to video game settings and GPU usage is about 94 percent I believe.

Edited, Sep 15th 2010 10:18pm by lambon
____________________________
They appeared from the sky, the one without a heart... "Omega"...And the one who followed it..."Shinryuu"...
Final Fantasy 5
#12 Sep 15 2010 at 8:24 PM Rating: Decent
Avatar
**
421 posts
i have a quick question;

with AA & dept in land opnion on, my hd5850 card is always at 100% with temperature close to 200F (reading from the ati gpu window7 gadget addon)

is that normal for a gpu to run @ 100%?!? or would it hurt my system?!
____________________________
<a href="http://www.enjin.com/" alt="guild hosting"><img src="http://sigs.enjin.com/sig-ffxiv/39704_48d6215903dff562.png"></a>
#13 Sep 15 2010 at 8:30 PM Rating: Good
**
723 posts
lambon wrote:
I would not use Furmark to test for stability.
Furmark pushes your GPU to insane temperatures. A lot higher than what a game does. I'm talking 20C degrees higher.
I don't think it's worth trying to stabilize it on furmark for video games.
I suggest ATITool or MSI Kombustor. Very close to video game settings and GPU usage is about 94 percent I believe.

Edited, Sep 15th 2010 10:18pm by lambon


I and many others have seen their GPU's hit 100% usage playing FFXIV, so if you're going to overclock you may as well push your GPU as hard as you can to make sure it's stable. I'll stick with my Furmark =)

Ggrab wrote:
i have a quick question;

with AA & dept in land opnion on, my hd5850 card is always at 100% with temperature close to 200F (reading from the ati gpu window7 gadget addon)

is that normal for a gpu to run @ 100%?!? or would it hurt my system?!


Yes, a lot of cards are running at 100% right now. I personally like keeping Depth of Field off. I have 2 GTX 285's in SLI and with it on, the game runs well and is playable, but with it off the game is completely smooth 100% of the time. I do consider myself to be a bit of a graphics *****, but even I find it acceptable to turn DoF off. You're not missing too much.

Also, ATI cards in my experience tend to get hotter than nVidia's. You can manually set the fan speed in your Catalyst Control Center, but the other downside of ATI cards is that many of them are quite noisy when you boost that fan up. 200 F is around 93 celsius, which is certainly hot but actually within tolerable range for GPUs these days. I would boost your fan speed though, while the card can work at that temp, it's not something I'd be comfortable with if you game often.

Edited, Sep 15th 2010 10:35pm by Jeraziah
#14 Sep 15 2010 at 9:29 PM Rating: Decent
Avatar
**
421 posts
Yea and with FF I'm looking at 6-8hours a day of gameplay , I have a coolermaster HAF322 case so mycomputer overall is cool , on my side fan is it good to push air in the case or set it to blow air out?!? I'm ganna dig around the CPU fan setting when I get home , thanks
____________________________
<a href="http://www.enjin.com/" alt="guild hosting"><img src="http://sigs.enjin.com/sig-ffxiv/39704_48d6215903dff562.png"></a>
#15 Sep 15 2010 at 9:47 PM Rating: Default
**
412 posts
I'm telling you, Furmark is not a good choice to test for stability, especially for MMORPG where you spend a lot of time on it.
The moment you start it up, it shoots you in to extreme temperatures. Temperatures you will never see in gaming. I OC'ed my 460 1gb from MSI to 900/2200
I get temperatures of about 60-65C playing FFXIV.

Furmark would push my temperatures to 80-90C. Simply not worth stressing my card out so much.
The best benchmark is FFXIV itself (not the FFXIV benchmark, I mean the game itself).

Just rechecked Kombustor, it uses 98 percent of the GPU.

Edited, Sep 16th 2010 8:43pm by lambon
____________________________
They appeared from the sky, the one without a heart... "Omega"...And the one who followed it..."Shinryuu"...
Final Fantasy 5
#16 Sep 15 2010 at 10:06 PM Rating: Good
Avatar
***
2,448 posts
I personally do not like furmark either. For CPU stability, I used Prime 95 just for temperature threshold testing, but for the real meat you wanna use LinX which runs the famous Linpak. Linpak will find errors(indicating unstable cpu) in 30 minutes what Prime95 might take 5 hours to find(simply because all prime basically does is heat it up. Same reason for not liking furmark for fool-proof assurance. Stable under high heat!= stable under all conditions, with all workload types thrown at it.

Used OCCT for GPU stability testing.

To each his own I guess? o.ob
____________________________
Currently Playing: FFXIV:ARR
Lacaan Vasiim:Cactuar
Free Company:Cactuar Corp<CCorp>
catwho wrote:
If you need a bard to get "good exp" in a merit party, you're the weakest link.
This forum is read only
This Forum is Read Only!
Recent Visitors: 13 All times are in CST
Anonymous Guests (13)