Vanilla EVGA GTX 670 Testing Anyone?

[Don Karnage]

Contact of mine hooked me up for FREE :biggrin:

Going to see what she can do on stock air then water

 

[realjetavenger]

Free? Dang, I need some new friends.
Crank that sucker up and tell us how it does. And how loud does the fan get?

 

[Broheim]

I'm really curious to see how it performs rendering with V-ray RT using the Cuda engine, but I doubt there's a free benchmark out there for it.

 

[Don Karnage]

Install went smooth and easy. Just loaded drivers. Going to see what she pulls stock

 

[Don Karnage]

First run

http://3dmark.com/3dm11/3560903

Score

P8629 3DMarks

Graphics Score

8684

Physics Score

9411

Combined Score

7363

 

[toyota]

First run

http://3dmark.com/3dm11/3560903

Score

P8629 3DMarks

Graphics Score

8684

Physics Score

9411

Combined Score

7363

you can compare this to out of box run on the FTW card. your cpu is better so its not a direct comparison though.


Score

P8701 3DMarks

Graphics Score

9025

Physics Score

8078

Combined Score

7547

 

[Don Karnage]

you can compare this to out of box run on the FTW card. your cpu is better so its not a direct comparison though.


Score

P8701 3DMarks

Graphics Score

9025

Physics Score

8078

Combined Score

7547

Boost clock is hitting 1097mhz and the card is maxing at 75C so with better cooling it should perform much better.

 

[toyota]

Boost clock is hitting 1097mhz and the card is maxing at 75C so with better cooling it should perform much better.

out of box I hit 1189 below 70 C and 1176 over 70 C. I raise the power target to 135% and offset to 53 and hit 1242 under load. the highest I have tried so far for memory is 6418 and its 6208 out of box. I have not even tried to see what it will really do but even 1242 and 6418 is not too bad.

 

[Termie]

you can compare this to out of box run on the FTW card. your cpu is better so its not a direct comparison though.


Score

P8701 3DMarks

Graphics Score

9025

Physics Score

8078

Combined Score

7547

Interestingly, his physics score is about 17% higher than yours, almost exactly in proportion to the 10% higher clock and ~7% higher IPC.

Boost clock is hitting 1097mhz and the card is maxing at 75C so with better cooling it should perform much better.

You'd be at 1110 if you were under 70C. Just up the fan profile to get there if you're interested.

out of box I hit 1189 below 70 C and 1176 over 70 C. I raise the power target to 135% and offset to 53 and hit 1242 under load. the highest I have tried so far for memory is 6418 and its 6208 out of box. I have not even tried to see what it will really do but even 1242 and 6418 is not too bad.

Just curious - any reason you're not picking round numbers for OC'ing? Why 53 and not 50 or 55? Why 210 on the memory and not 200 or 250? Is there some strategy there, like conforming to certain standard nVidia thresholds?

By the way, your FTW edition allows much higher power targets than the vanilla, more like the 680s. That probably won't make a difference, however, because your card will lock up long before it hits 135% power. Keep in mind that all power does is maintain a given clock at a certain TDP, rather than auto-underclock. If the board doesn't have the power to give, it will lock up, even if you have headroom in your power offset.

 

[Don Karnage]

out of box I hit 1189 below 70 C and 1176 over 70 C. I raise the power target to 135% and offset to 53 and hit 1242 under load. the highest I have tried so far for memory is 6418 and its 6208 out of box. I have not even tried to see what it will really do but even 1242 and 6418 is not too bad.

Setting up the loop as I speak. Im hoping for 30C loads

 

[Termie]

Setting up the loop as I speak. Im hoping for 30C loads

This will only give you 13MHz more speed. Kepler is only temp-limited in that it auto-undervolts. At high boost, temps are not what cause lockups.

 

[toyota]

Interestingly, his physics score is about 17% higher than yours, almost exactly in proportion to the 10% higher clock and ~7% higher IPC.



You'd be at 1110 if you were under 70C. Just up the fan profile to get there if you're interested.



Just curious - any reason you're not picking round numbers for OC'ing? Why 53 and not 50 or 55? Why 210 on the memory and not 200 or 250? Is there some strategy there, like conforming to certain standard nVidia thresholds?

By the way, your FTW edition allows much higher power targets than the vanilla, more like the 680s. That probably won't make a difference, however, because your card will lock up long before it hits 135% power. Keep in mind that all power does is maintain a given clock at a certain TDP, rather than auto-underclock. If the board doesn't have the power to give, it will lock up, even if you have headroom in your power offset.

it only has certain numbers that it goes to when moving the slider in PrecisionX.

 

[guskline]

First run

http://3dmark.com/3dm11/3560903

Score

P8629 3DMarks

Graphics Score

8684

Physics Score

9411

Combined Score

7363

Don: To give some perspective, my bone stock GTX 680 running in rig 2 below (SB 2500k@4532) gave a performance 3dmark11 score of 8936. Unfortunately, I'm still at work and don't have the balance of the details. I know with an OC by EVGA pecision to 135% power;core ocd 145 and memory 500, my score jumped to @9650. Some people who really push get over 10,000. I run my card stock. It's already fast. The GTX670s are very close and for the $100+ difference the GTX670 has to be the best buy in graphics cards at present for high end cards.

 

[cmdrdredd]

Just curious - any reason you're not picking round numbers for OC'ing? Why 53 and not 50 or 55? Why 210 on the memory and not 200 or 250? Is there some strategy there, like conforming to certain standard nVidia thresholds?

By the way, your FTW edition allows much higher power targets than the vanilla, more like the 680s. That probably won't make a difference, however, because your card will lock up long before it hits 135% power. Keep in mind that all power does is maintain a given clock at a certain TDP, rather than auto-underclock. If the board doesn't have the power to give, it will lock up, even if you have headroom in your power offset.

I do that too because the way Boost clock is calculated with TDP % increase. I run +145% TDP and +86Mhz core gives me 1250 boost clock under load even when it says my boost clock is 1165. It's kind of weird, can't set a specific number and the reporting is pretty strange too. GPU-Z doesn't seem to report any changes when you change 2 or 3Mhz. When I go up 5 it will report 4 sometimes.

 

[toyota]

Don: To give some perspective, my bone stock GTX 680 running in rig 2 below (SB 2500k@4532) gave a performance 3dmark11 score of 8936. Unfortunately, I'm still at work and don't have the balance of the details. I know with an OC by EVGA pecision to 135% power;core ocd 145 and memory 500, my score jumped to @9650. Some people who really push get over 10,000. I run my card stock. It's already fast. The GTX670s are very close and for the $100+ difference the GTX670 has to be the best buy in graphics cards at present for high end cards.

we have basically the same cpus and just adding a 53 mhz offset and 210(effective) memory increase to my FTW card gave me a score P9188. I went with the gtx670 because clock for clock its only about 3-4% slower and many gtx680 cards have no better overclocking than a gtx670.

 

[cmdrdredd]

If all goes well tomorrow I'll have my Ivy Bridge system setup and I'll be interested to see the difference that has on the performance of this card.

 

[guskline]

we have basically the same cpus and just adding a 53 mhz offset and 210(effective) memory increase to my FTW card gave me a score P9188. I went with the gtx670 because clock for clock its only about 3-4% slower and many gtx680 cards have no better overclocking than a gtx670.

toyota that's an incredibly good score. When I get home, I'll set my GTX680 to those settings and report to you. I also have surround setup so I'll set it back to 1920 x 1080 and run the test on only 1 monitor to keep it as equal as possible. I have a PNY GTX 680. I just saw that the EVGA GTX 670 FTW has a stock core of 1006 and a boost of 1084 which is much closer to the reference GTX 680 than the reference GTX 670 (915, 980).

 

[WiseUp216]

I grabbed the same EVGA model at Microcenter because the FTW model was out of stock. I wasn't patient enough to order online, especially with the long holiday weekend.

Definitely not disappointed. I was only slightly hesitant because so many people complained about the noises the fan made, but I don't hear it at all. I'm guess you haven't either.

 

[Termie]

it only has certain numbers that it goes to when moving the slider in PrecisionX.

I do that too because the way Boost clock is calculated with TDP % increase. I run +145% TDP and +86Mhz core gives me 1250 boost clock under load even when it says my boost clock is 1165. It's kind of weird, can't set a specific number and the reporting is pretty strange too. GPU-Z doesn't seem to report any changes when you change 2 or 3Mhz. When I go up 5 it will report 4 sometimes.

Hmm, I haven't noticed what Toyota mentioned, but I have seen that sometimes you dial in a round number like 100 and you get an actual boost of 99, or something like that, but that's usually when the card is approaching its limits. Punching in 100 on my card takes me straight from 1110 to 1210, and 150 gives me 1260.

If all goes well tomorrow I'll have my Ivy Bridge system setup and I'll be interested to see the difference that has on the performance of this card.

Congrats - I look forward to your results. If you have a chance to bench before the upgrade, make sure to get minimums as well as averages. I've also down some GPU usage and CPU load screenshots to get an idea of how the two interact.

 

[Termie]

I grabbed the same EVGA model at Microcenter because the FTW model was out of stock. I wasn't patient enough to order online, especially with the long holiday weekend.

Definitely not disappointed. I was only slightly hesitant because so many people complained about the noises the fan made, but I don't hear it at all. I'm guess you haven't either.

I have had two EVGA vanilla cards (the first RMA'd), and both made the same noise at idle. It really bugs me, but I have tuned my heavily ventilated and poorly sound deadening case to be very, very quiet, with low RPM Scythe fluid dynamic bearing fans throughout. Above stock fans in a closed-up case, I probably wouldn't hear the 670.

Don, what do you think of the fan noise?

 

[guskline]

When I see these really impressive scores, I have to laugh at the poster who complained that these were only mid-ranged chips. I think a fairer chracterization is "high-end performance" from a chip that wasn't supposed to be Nvidia's fastest GPU.

 

[guskline]

Termie, what video card did you run in your setup before the GTX 670 and what fan system did it have?

 

[WiseUp216]

I have had two EVGA vanilla cards (the first RMA'd), and both made the same noise at idle. It really bugs me, but I have tuned my heavily ventilated and poorly sound deadening case to be very, very quiet, with low RPM Scythe fluid dynamic bearing fans throughout. Above stock fans in a closed-up case, I probably wouldn't hear the 670.

Don, what do you think of the fan noise?

I have a Corsair 400r, which is a very "open" case with screens all around. If the 670 was making any sounds out of the ordinary, I would definitely hear it.

I thought they were giving FTW versions to those who RMA'd the reference model?

 

[Termie]

I have a Corsair 400r, which is a very "open" case with screens all around. If the 670 was making any sounds out of the ordinary, I would definitely hear it.

I thought they were giving FTW versions to those who RMA'd the reference model?

No dice, that was only for the SC model, which was immediately discountinued after launch. And interesting that you don't hear anything - I'd think I was imagining it but I've had two cards with the same unique buzzing sound.

Termie, what video card did you run in your setup before the GTX 670 and what fan system did it have?

5850 crossfire, one with a reference fan, one with a Sapphire mid-mounted fan, both running at 20% at idle. It was quieter at idle than my single 670 (which unfortunately only goes down to 30% fan minimum), but of course much louder at load.

 

[toyota]

If all goes well tomorrow I'll have my Ivy Bridge system setup and I'll be interested to see the difference that has on the performance of this card.

btw what is your idle fan speed? mine is 40% but a guy on another forum says his 670 FTW card has a 30% idle fan speed.