*GTX 260/280 Unofficial TPU! Thread*

[chris189]

Actually according to these benchmark reviews. A single GTX580 outperforms 3-way SLI GTX285s @ 2560x1600. Crazy huh? 2-way sli 285s in crysis warhead dx10 @ 2560x1600 get 24fps. A single GTX580 in crysis warhead dx10 @ 2560x1600 gets 40fps. It outperforms any number of 285s @ 2560x1600. It also outperforms the sli 285s at every resolution as well.

A single GTX470 gets 42fps @ 1920x1080 while SLI 285s get 59fp. With some huge overclocks to the 470, it could come very close to an sli'ed 285 setup.

3dmarkvantage scores compared:
14,872 points for the GTX470 while the SLI 285 setup makes a pretty darn incredible 22,223 points. Again the GTX580 wins making an outstanding 24,848 points!!!

GTX 580 bencharmk
http://www.guru3d.com/article/geforce-gtx-580-review/16
GTX285 single/ sli/ and 3-way sli benchmark
http://www.guru3d.com/article/geforce-gtx-285-review--3way-sli/17

**When i get my GTX470 next week, and you have your 285s up and running. We should do some comparison tests, in several different benchmarks. Sounds fun huh?!?

*EDIT*
After reading this review.
http://www.guru3d.com/article/geforce-gtx-570-review/1
I have decided i'm going to get a GTX570 rather. It's actually worth it! plus 21fps on metro2033 for the 470 is way too low. Even after i overclocked it to the max, i may not break 30fps. The 570 on the other hand pulls off 27fps. With minor to major overclock I believe it would easily break 30fps+.

 

[johnspack]

I'll bet I can break 24k! In fact, I'm counting on it, at least once I get my quad. As for the super hi-res scores, of course the cards with more than 1gb of memory will pull ahead. If you are going to game at that high of a res, then yes, the gtx4 and 5s make more sense. I'm limited to 1920x1200, which is where the 285s shine. I'd like a higher res, but try and find a 24" 1920x1200 monitor these days. If you do, they aren't cheap.

 

[chris189]

I'll bet I can break 24k! In fact, I'm counting on it, at least once I get my quad. As for the super hi-res scores, of course the cards with more than 1gb of memory will pull ahead. If you are going to game at that high of a res, then yes, the gtx4 and 5s make more sense. I'm limited to 1920x1200, which is where the 285s shine. I'd like a higher res, but try and find a 24" 1920x1200 monitor these days. If you do, they aren't cheap.

24k is going to require a damn fine overclock. Again those benchmarks are running on a core i7 965 @ 3.75Ghz. I'm running a core i7 930 @ 4Ghz. The gtx570 is so close to the 580 stock , it's a steal. Overclocked it comes with in 2-3fps on the 580 at stock speeds. Impressive in my opinion.

 

[johnspack]

I don't think so, I can almost hit 22k with my 260s! I don't think they tried hard enough....

 

[chris189]

I don't think so, I can almost hit 22k with my 260s! I don't think they tried hard enough....

What's ur rig like and overclock speeds? Do you have a screen shot of what you've managed with the SLI'ed 260s? Including the vantage window, gpuz, and cpuz.

**I managed around 11,800 on a my single 260 192 core. I suppose if your scaling is perfectly linear 20-22k is possible in a perfect world. 3dmark vantage can handle some insane gpu overclocks and not crash, unlike every game ever made. I think i pulled 11,800 with a 780Mhz Core - 1560Mhz Shader - 1225Mhz Memory. I'll post my screen shot when i get home later. Unusual that people with GTX260 216 at stock speeds pulling 12214 total score. I need to look at my gpu score specifically to compare though.

 

[johnspack]

I'm actually in single mode atm, I was having a weird issue where I could only hit 3d low clocks in single mode, pulled a card out, and now I can do 666/1622/1188 fully stable. Thats's on my 55nm card at 1.18v. I think my 65nm card isn't that stable....
Validation: http://www.techpowerup.com/gpuz/ndcks/ I think I could do higher stable. I'll let you know!
Yep, I can, 1656 shaders. It's super chip!

 

[johnspack]

Okay, not as happy now. Seems mostly, under 260 drivers, my cards drop to 3d low. Got frustrated and tried to flash bios to fix that, now it doesn't work. What's the command in nvflash to flash a 2nd card? Man, my 3 drive raid array is failing, unplugged right now, still don't know how I'm going to get a quad, xmas time is great!

 

[chris189]

well i flash my single via "nvflash -5 -6 newbios.rom". You need to allow rivatuner to force 3d all the time, then set it to force performance mode in nvidia control panel, and you will not ever have that problem. It's not in the bios. I used to have that problem. thought it was my card. It's not. Just make it force constant 3d mode and it'll never kick out. Only now when theres a problem itll kick to the desktop saying driver failed when overclocked too much. If it's pushed to hard the screen will artifact up completely and freeze. It requires a manual reset.

 

[johnspack]

Well, that's it. No more experiments for me! I destroyed that 260. It's so bad, when I put it in the 2nd pci-e slot, I ended up having to pull battery and reset bios to boot again. Down to one 260 now... I'll try rivatuner, haven't used it in a while, didn't realize I could use it to force 3d high.

 

[chris189]

and look at this score

 

[johnspack]

Nice score! I can't run vantage right now because my win7 install was on my raid array. I've lost too much hardware in the last few days! Can't wait for my gtx285s to show up, now I really need them.

Trying to find the setting to force 3d mode in rivatuner, can't find it. There is a "Force constant performance level" option, but it's greyed out and I can't select it.
Edit: nope, no way to force 3d in 260 drivers. I'm stuck at 258. Looks like all support for 2xx series has been dropped in the newer drivers. I'm running a dead clean system,
and I've already killed a 260 to prove the point. Hopefully the gtx285s I'm getting will respond differently.

Thanks Chris, for the tip. I looked at the advanced rivatuner settings, and enabled force performance level mode. No more drops in clocks! Wish I'd figured this out before now.......

 

[johnspack]

This old 65nm 260 is plain faster, it beat the 55nm by a bit and with lower temps and shader clocks!:

 

[chris189]

Your Welcome! If u noticed. I made more points at 780 core than 790. I tweaked some settings to improve my score, which in fact made it worst. Set rivatuner direct3d Mip map lod bias to +3 to decrease image quality therefor increase fps and score... Hopefully. lets test it! Also make sure ur nvidia control panel settings are all set to best performance and worst quality possible. I've also messed with the max pre rendered frames. It's best for fast cpu's. Maybe set it to 0 if ur on a dual core to get the best fps. i've seen this help on single core cpu's a lot. Maybe it would help on a dual core. Drivers shouldn't affect anything. i'm gonna do some driver version testing. I think theres a 263 driver out now but it's only for the gtx 500 series. i'm running the 258.96 which i have found to be very stable. I'm gonna run vantage and see what driver performs the best. 258.96 vs 260.99.

 

[johnspack]

Just got my first 285! At 712/1620/2300 it blows a 260 out of the water. My min fps in my ww1 sim just went from 30s to 50s. Dam. Can't wait to get the 2nd one!

 

[chris189]

Nice. I've decided i'm going to get the 470. It's on sale for $220 on newegg with an additonal $20 off mail in rebate. I might buy 2 and SLI these bad ass cards.

I had a question as well. I have a ATI Radeon X1950XT 256MB and i was wondering if you thought a ATI Radeon 4670 would outperform the X1950XT?

 

[johnspack]

As for the ati stuff, no clue, I don't run it. As for the 470, would love to see some benchies! When I get my 2nd 285, I'll do some vantage scores....

 

[chris189]

As for the ati stuff, no clue, I don't run it. As for the 470, would love to see some benchies! When I get my 2nd 285, I'll do some vantage scores....

Actually it's 249.99 with a 12% off code and an additional $30 mail in rebate. 228.78 shipped - $30 = $198.78. I wonder if my Kingwin 850W power supply would hold up with two GTX 470s? It has 70 Amps on the 12v rail... It sucks since this 12% off is only good until the 29th. I won't have enough cash to buy two... Hey what power supply do u have to run 2 gtx 285s?

 

[Mussels]

Actually it's 249.99 with a 12% off code and an additional $30 mail in rebate. 228.78 shipped - $30 = $198.78. I wonder if my Kingwin 850W power supply would hold up with two GTX 470s? It has 70 Amps on the 12v rail... It sucks since this 12% off is only good until the 29th. I won't have enough cash to buy two... Hey what power supply do u have to run 2 gtx 285s?

with 70A on the 12V, it should definitely be sufficient.

 

[chris189]

with 70A on the 12V, it should definitely be sufficient.

Right now when im loaded to the max with my 4ghz i7 and 5 hdd's and gtx 260 overclocked to the max, i saw the 12v drop to 11.862V... I'm not sure how well it would hold up with 2 470s which draw way more than a single gtx 260. My old BFGtech once it dropped below 11.7V it would instantly shut down. Who knows though, it could hold that 11.862 even with 2 470s under load..... I've never had an SLI setup before but it sounds very exciting!!!!!!!!!! By the way i could just return this 850 and get a 1000w or 1200w for one $30-40 more.... Almost totally worth it!!!

 

[johnspack]

I've got well over 70a available with this silverstone 850. It peaks at 900watts, so I'm good. 2 470s, probably not.....

 

[chris189]

Use AIDA64 or everest to monitor your 12v rail under load. Then post your results. Voltage before load, and after.... stress the cpu as well. On your power supply with your cpu, u could probably run 2 470s. The Core i7 is a power hungry monster and sucks down twice as much power as a core2quad overclocked the max. That's why i need a super powerful power supply to supply power to the cpu and gpu(s). This Kingwin pushes much more than 850W under load as well. i've seen 60amps+ load on the cpu and gpu. That's when the 12V line starts to drop big time. It's still worth a shot. Actually it would be a good test to run the 260 and the 470 so i can team up the CUDA cores for Xilisoft Video Converter ultimate. It loves cuda cores. The more the better. 448 on the 470 plus 192 on the 260 = 640 cuda cores! That should speed up my encoding tremendously! I compared my system to our Xeon Server machine at work. It takes 9 minutes to encode 3 videos to my HTC HD2 format using cuda and my 4ghz core i7. I did the same test at work on the server. It took 55 minutes using the Quad Core w/ Hyperthreading Xeon processor. Shows how much more powerful my system is than our Corporate server. They paid a fortune for that damn thing. It's loud, annoying and overpriced. It's an IBM server. The hard drives inside vibrate up against the case inside and make this god awful annoying vibration/ rattle. I'm constantly kicking it at work to shut it up. Again this is a $2,000 IBM Server machine which is 30 days old. What a piece of shit.

 

[Mussels]

Use AIDA64 or everest to monitor your 12v rail under load. Then post your results. Voltage before load, and after.... stress the cpu as well.

oh hell no, NEVER use software to measure voltages, or the readings in the BIOS. get a multimeter or just dont bother.

On your power supply with your cpu, u could probably run 2 470s. The Core i7 is a power hungry monster and sucks down twice as much power as a core2quad overclocked the max.

they use more, but they sure dont use twice as much.

That's why i need a super powerful power supply to supply power to the cpu and gpu(s). This Kingwin pushes much more than 850W under load as well. i've seen 60amps+ load on the cpu and gpu.

i might need to check your specs again, but seriously - you think you've used 720W+ off your 12V rail? how did you measure that?

Actually it would be a good test to run the 260 and the 470 so i can team up the CUDA cores for Xilisoft Video Converter ultimate. It loves cuda cores. The more the better. 448 on the 470 plus 192 on the 260 = 640 cuda cores! That should speed up my encoding tremendously! I compared my system to our Xeon Server machine at work. It takes 9 minutes to encode 3 videos to my HTC HD2 format using cuda and my 4ghz core i7. I did the same test at work on the server. It took 55 minutes using the Quad Core w/ Hyperthreading Xeon processor. Shows how much more powerful my system is than our Corporate server.

yes, but cuda acceleration isnt as multi task as the CPU's. sure its faster in certain circumstances, but i doubt they use their server to encode videos in xilisoft. servers are built to be reliable and multi purpose (well, multi purpose for server tasks)

They paid a fortune for that damn thing. It's loud, annoying and overpriced. It's an IBM server. The hard drives inside vibrate up against the case inside and make this god awful annoying vibration/ rattle. I'm constantly kicking it at work to shut it up. Again this is a $2,000 IBM Server machine which is 30 days old. What a piece of shit.

if they paid a fortune for it, why the hell havent they fired you for KICKING it? no wonder it rattles...

 

[johnspack]

Actually the i7 with it's 8 threads, is much more efficient than the core2quad. I just seriously want one now instead of any more 775 crap. Plus 6gigs of tri-channel ddr3 would rock too. Now where am I going to scrape up $700 to buy these used parts? Hmmm, think my liver still works.....

 

[chris189]

Mussels - According to thermaltake power supply calculator I require much more power than 850W with my current setup. That shows my power supply is being pushed to it's limits. Hints the drop in voltage. I can agree that what AIDA64 reports as current power load in amps is not entirely correct. Much more incorrect than what i've seen in reviews where they monitor total system load power consumption on a similar system. I do believe the 12V line drop is relatively accurate and most likely very close to it's actual value. In regards to the server. Sure it is poorly designed and has no vibration dampening materials inside the case to minimize rattling/ vibration. Also according to everest ultimate/ AIDA64 my system far exceeds the performance of that server. Again being overclocked, helps. They would never fire me for lightly tapping it to shut it up. I could have built them a better server with much better specs and performance. I guess I over exaggerated the whole kicking thing. I would never full on kick a system, i understand what components can't take a beating inside the system therefore would not "hurt" the system.

 

[Mussels]

Mussels - According to thermaltake power supply calculator I require much more power than 850W with my current setup. That shows my power supply is being pushed to it's limits. Hints the drop in voltage. I can agree that what AIDA64 reports as current power load in amps is not entirely correct. Much more incorrect than what i've seen in reviews where they monitor total system load power consumption on a similar system. I do believe the 12V line drop is relatively accurate and most likely very close to it's actual value. In regards to the server. Sure it is poorly designed and has no vibration dampening materials inside the case to minimize rattling/ vibration. Also according to everest ultimate/ AIDA64 my system far exceeds the performance of that server. Again being overclocked, helps. They would never fire me for lightly tapping it to shut it up. I could have built them a better server with much better specs and performance. I guess I over exaggerated the whole kicking thing. I would never full on kick a system, i understand what components can't take a beating inside the system therefore would not "hurt" the system.

those software calculators are utter BS. go run my PC through that same software, and tell me the result. i'll show you the results from my wall meter showing how inaccurate it is.


you have to remember that the sensor reading your 12V rail isnt direcnt connected to the PSU like a multimeter would be - its part of the motherboard, and a few steps in. so while it may record drops, we dont know how precise it is (can it read changes from 12.70 to 12.71, or is it moving in steps of 0.03 or the like) - and because of its placement, its reading only one part of the motherboard. hell we dont even know if the software needs to be calibrated to be accurate on the motherboard, so you may get accurate movement but the readings are off by a few percent.

i'm not saying your voltage drops are irrelevant - it is a sign of the PSU reaching its load... but thats likely just due to it being a crap PSU, and not the wattage. a quality PSU matters more than the (often exaggerated) wattage output.