Anyone get a GTX 670 that's stable over 1300mhz?

[Ieat]

Just wondering if there are a lot of them out there. My Gigabyte got very close at 1293 mhz. But couldn't get that extra 7 mhz completely stable in game. lol It would crash 20 minutes to an hour into gaming.

 

[3DVagabond]

Just wondering if there are a lot of them out there. My Gigabyte got very close at 1293 mhz. But couldn't get that extra 7 mhz completely stable in game. lol It would crash 20 minutes to an hour into gaming.

Is it the O/C version on the 680 PCB? Or, do you have a reference version? Also, if you don't mind, what PSU are you running. I ask only because it also contributes to O/C'ing stability.

Thanks.

Sorry, what mobo as well?

 

[Ieat]

Yep its the Windforce o/c version. I don't think Gigabyte released a reference card for launch. I'm using an Antec 750w Earthwatts. Mobo is in Sig.

 

[3DVagabond]

Yep its the Windforce o/c version. I don't think Gigabyte released a reference card for launch. I'm using an Antec 750w Earthwatts. Mobo is in Sig.

Thanks, mate. I think people should avoid the reference designs. The models on the 680 PCB's just seem so much better. It's nice to see that you didn't need anything "esoteric" to achieve those clocks. You have equipment that is still good value for the performance returned.

Another question. Do you have to run real high fans to stabilize at those clocks?

 

[cmdrdredd]

I found MSI Afterburner and EVGA PrecisionX both crash games when they are running. For what reason I don't know. Even when I don't overclock, just having them in the background breaks the game. A quick search pulled up more than one case of this.

I don't know how else to actually overclock my 670

 

[3DVagabond]

I found MSI Afterburner and EVGA PrecisionX both crash games when they are running. For what reason I don't know. Even when I don't overclock, just having them in the background breaks the game. A quick search pulled up more than one case of this.

I don't know how else to actually overclock my 670

Does Asus have an O/C tool for the 600's?

 

[Ieat]

Another question. Do you have to run real high fans to stabilize at those clocks?

I used Crysis 2 to test stability. Temps were 67c with fanspeed at 58% using a custom profile while playing.

 

[exar333]

I will test mine next week when I get it.

Hoping for 1250+ on my OC.

 

[cmdrdredd]

Does Asus have an O/C tool for the 600's?

Just found it, GPU Tweak. Gonna give it a shot.

 

[SickBeast]

1300mhz seems to be the max that these cards will go without a volt mod. With the volt mod apparently people hit up to 1500mhz on air. With extreme cooling and drastic mods people can hit around 1900mhz.

 

[toyota]

the memory bandwidth is probably going to hold it back in most cases anyway so above 1300mhz might not mean much.

 

[SickBeast]

the memory bandwidth is probably going to hold it back in most cases anyway so above 1300mhz might not mean much.

I doubt that. I know that on my 7850 overclocking the memory does next to nothing. Generally with these newer GPUs it's all about how high you can clock the core.

 

[toyota]

I doubt that. I know that on my 7850 overclocking the memory does next to nothing. Generally with these newer GPUs it's all about how high you can clock the core.

the bandwidth on the 680/670 is most certainly holding it back. some of the people that oc the memory on the 680/670 and claim it does not help dont realize that the memory is already running at its rated speed out of the box. gddr5 will actually do nothing and even at some point slow down performance when pushed to a certain point. the gtx680 has TWICE the shading power and OVER TWICE the texturing power of the gtx580 but can only utilize about 1/4 to 1/3 of that advantage due to lack of bandwidth and rops. it really comes down to the game as that determines how bad the memory bandwidth is an issue.

 

[Don Karnage]

Give me a few days and i'll show you 1300mhz boost clock on a 670

 

[cmdrdredd]

Well, GPU Tweak from Asus doesn't crash but I found out my card is a poor overclocker. The card I have defaults to 1175/6210 by default according to the monitor apps.

 

[Don Karnage]

Well, GPU Tweak from Asus doesn't crash but I found out my card is a poor overclocker. The card I have defaults to 1175/6210 by default according to the monitor apps.

How are temps? Have you tried benching with 100% fan?

 

[SickBeast]

the bandwidth on the 680/670 is most certainly holding it back. some of the people that oc the memory on the 680/670 and claim it does not help dont realize that the memory is already running at its rated speed out of the box. gddr5 will actually do nothing and even at some point slow down performance when pushed to a certain point. the gtx680 has TWICE the shading power and OVER TWICE the texturing power of the gtx580 but can only utilize about 1/4 to 1/3 of that advantage due to lack of bandwidth and rops. it really comes down to the game as that determines how bad the memory bandwidth is an issue.

It can't be holding it back that much seeing as it's able to beat the 7970 in most cases when that card has the 384-bit bus.

Are you saying that the card would be 30% faster with a 512-bit bus?

 

[toyota]

It can't be holding it back that much seeing as it's able to beat the 7970 in most cases when that card has the 384-bit bus.

Are you saying that the card would be 30% faster with a 512-bit bus?

what kind of logic is that? thats like saying the core frequency of the 7970 is not holding it back because its faster than the gtx680 in some games. the gtx680 would be MUCH faster overall if it had more bandwidth.

again it has twice the shading power and over twice the texturing power of the gtx580 yet is only about 30% faster overall.

a few examples that might be bandwidth limited:

the gtx680 is only 10% faster than gtx580 in Alan Wake on max settings at 1080.
http://www.hardware.fr/articles/857-12/benchmark-alan-wake.html

the gtx680 is only 18% faster than gtx580 in Metro 2033 at 1080
http://www.guru3d.com/article/geforce-gtx-680-review/18

the gtx680 is only 18% faster than the gtx580 in Warhead
http://www.anandtech.com/show/5699/n...x-680-review/7

The minimum framerate situation is even worse for NVIDIA here, with the GTX 680 clearly falling behind the 7950, and improving on the GTX 580 by only 10%. At its worst Crysis is absolutely devouring memory bandwidth here, and that leaves the GTX 680 underprepared.

 

[RussianSensation]

It can't be holding it back that much seeing as it's able to beat the 7970 in most cases when that card has the 384-bit bus.

Are you saying that the card would be 30% faster with a 512-bit bus?

1) It's often misleading to compare memory bandwidth limitations across brands. For example, GTX560TI mops the floor with HD4890 despite similar memory bandwidth. Even within the same brand it can often be tricky to compare. Using the same GTX560 Ti vs. GTX285.

2) HD7870 has higher pixel fill-rate than HD7970. Thus while HD7970 has a ton of memory bandwidth, the card could very well be pixel fill rate limited. In other words even when HD7970's memory bandwidth reaches 340 GB/sec with an overclock, it doesn't show substantial increases in performance. When the GPU on the 7970 is overclocked though, it shows massive performance increase suggesting the main bottleneck rests in the GPU.

3) We don't know how much faster GTX670/680 could be with 512-bit bus but given that it's only 30% faster than the GTX580 despite have bucket loads of texture and shader performance, it's reasonable to assume that it's more memory bandwidth starved compared to the GTX580. Of course since NV put the fastest GDDR5 on the 670 and gimped it with 256-bit bus, we won't know how much memory bandwidth was a factor until GTX700 series of cards.

At this point, I'd lean towards HD7970 wasting most of its memory bandwidth advantage. HD7870 OC vs. HD7950 shows that GCN's bottleneck is not memory bandwidth.

 

[cmdrdredd]

How are temps? Have you tried benching with 100% fan?

If I leave any monitoring app open when benchmarking, I get a crash at some point. When I make sure they are not running, everything works.

I did run 3dmark 11 and let it load up the GPU and then exit and check the temp immediately. Not the best test but it wasn't above 70c.

I can't get an accurate reading via software. For me it's alright, it's faster than my 6950 and even at stock (mine is bumped up a bit anyway) it's one hell of a fast GPU. As long as it doesn't crash in stock config I'm alright with it.

 

[Don Karnage]

If I leave any monitoring app open when benchmarking, I get a crash at some point. When I make sure they are not running, everything works.

I did run 3dmark 11 and let it load up the GPU and then exit and check the temp immediately. Not the best test but it wasn't above 70c.

I can't get an accurate reading via software. For me it's alright, it's faster than my 6950 and even at stock (mine is bumped up a bit anyway) it's one hell of a fast GPU. As long as it doesn't crash in stock config I'm alright with it.

What are your scores in 11?

 

[OILFIELDTRASH]

How do you even oc a 670? Does turning up core clock turn up max boost together? Or do I have to tweak each clock speed? I'll report back when I can play around with my 670 some.

 

[toyota]

How do you even oc a 670? Does turning up core clock turn up max boost together? Or do I have to tweak each clock speed? I'll report back when I can play around with my 670 some.

lol, have you not looked at the tons of Kepler reviews?

 

[Don Karnage]

How do you even oc a 670? Does turning up core clock turn up max boost together? Or do I have to tweak each clock speed? I'll report back when I can play around with my 670 some.

 

[Ieat]

How do you even oc a 670? Does turning up core clock turn up max boost together? Or do I have to tweak each clock speed? I'll report back when I can play around with my 670 some.

Each gtx 670 come with a different default max boost clock. You just move the slider to add mhz on top of that max boost clock.