Nvidia GeForce GTX 1000 Series (Pascal) MegaThread: FAQ and Resources

[U6b36ef]

i literally just posted a test of pcie 3.0 with the 1080 up about 5 posts!! so that question is answered now

Whooops, I didn't check the link, and must have skim read your post and forgotten. Sozzy.

Anyway yeah thank you. What did you make of it. That GamersNexus dude seems to think the 1080 doesn't choke on a PCI-E 2.0 x16 (or a PCI-E 3.0 x8 which I think he used).

The PCI-E 2.0 x16 results were a frame or two slower pretty much on all tests. It's either the case that the 1080 is the borderline of PCI-E 2.0, or just that testing this showed up a characteristic.

What I said was I wonder if someone knew the stats. Someone must know the data output of the 1080 Ti when it comes is released, which was the card I mentioned. Not the 1080 which your link did. Then it could be compared to data speed of the cards. Or tested again.

Unless output data of the 1080 is known we can't conclude what the tests in that link was saying. Or I can't anyway, so that's why I asked what you thought.

 

[U6b36ef]

I am staggered at why the Asus Strix is listed at £60 more than the Gigabyte.

https://www.amazon.co.uk/Gigabyte-GeForce-Windforce-PCI-Express-Graphics/dp/B01GCAW1IA/ref=sr_1_1?s=computers&ie=UTF8&qid=1466796307&sr=1-1&keywords=gtx+1080

https://www.amazon.co.uk/NVIDIA-GeForce-Gaming-Express-Graphics/dp/B01G92E9ZY/ref=sr_1_2?s=computers&ie=UTF8&qid=1466796307&sr=1-2&keywords=gtx+1080

It's possible prices may drop though when once in stock. I think the Maxwell's did.

Definitely though I hope AMD have something for this because if not, I worry Nvidia prices may rise. Maybe I am worrying too much.

 

[Math Geek]

i literally just posted a test of pcie 3.0 with the 1080 up about 5 posts!! so that question is answered now

Whooops, I didn't check the link, and must have skim read your post and forgotten. Sozzy.

Anyway yeah thank you. What did you make of it. That GamersNexus dude seems to think the 1080 doesn't choke on a PCI-E 2.0 x16 (or a PCI-E 3.0 x8 which I think he used).

The PCI-E 2.0 x16 results were a frame or two slower pretty much on all tests. It's either the case that the 1080 is the borderline of PCI-E 2.0, or just that testing this showed up a characteristic.

What I said was I wonder if someone knew the stats. Someone must know the data output of the 1080 Ti when it comes is released, which was the card I mentioned. Not the 1080 which your link did. Then it could be compared to data speed of the cards. Or tested again.

Unless output data of the 1080 is known we can't conclude what the tests in that link was saying. Or I can't anyway, so that's why I asked what you thought.

attempts have been made before, but i don't think there is any way to measure actual bandwidth usage from a card. tom's did a test last year i believe it was. it showed a VERY SMALL amount of data traffic, like less than 10% of the available bandwidth if i recall right. but it was not a test designed for such a use and probably not very accurate. i have not seen anything since then trying to measure it. if we knew numbers like that, then it wold be easy to say what slots would work since we know the pipeline's capabilities. so we are stuck with tests like nexus did with simply trying cards in various slots. last big one i saw took a titan x all the way down to 2.0 x4 before it started losing fps in a big way. but 1-2 fps is simple margin of error and can't really be used to conclude anything.

i can look back and see if i can find the article but it was a long time ago.

 

[Doomjoon]

The thing is though, all of these tests that I have seen seem to consistently show a 1-2fps difference, in favor of 16x 3.0 (vs 8x 3.0). I mean, it could be nothing, but such a consistent difference seems to point at something, no matter how small it is. Still, this obviously doesn't mean that the current PCI-E standards will bottleneck this generation of GPU's in any significant way.

 

[turkey3_scratch]

PCIex16 3.0 vs 2.0 has virtually no difference in performance. However, PCIe in x4 mode affects performance pretty significantky.

 

[U6b36ef]

The thing is though, all of these tests that I have seen seem to consistently show a 1-2fps difference, in favor of 16x 3.0 (vs 8x 3.0). I mean, it could be nothing, but such a consistent difference seems to point at something, no matter how small it is. Still, this obviously doesn't mean that the current PCI-E standards will bottleneck this generation of GPU's in any significant way.

That's the point. it is consistent. It's why I wondered if it was just a characteristic.

I do however remember that the GTX 680 was the first card to hit a choke on PCI-E 1.0, (or 1.1; can't remember). It was only up to about 5% bottleneck.

Yet I think it's safe to say a 1080 easily doubles a 680. Though if I recall, that 680 was a beast.

 

[TechyInAZ]

The thing is though, all of these tests that I have seen seem to consistently show a 1-2fps difference, in favor of 16x 3.0 (vs 8x 3.0). I mean, it could be nothing, but such a consistent difference seems to point at something, no matter how small it is. Still, this obviously doesn't mean that the current PCI-E standards will bottleneck this generation of GPU's in any significant way.

That's the point. it is consistent. It's why I wondered if it was just a characteristic.

I do however remember that the GTX 680 was the first card to hit a choke on PCI-E 1.0, (or 1.1; can't remember). It was only up to about 5% bottleneck.

Yet I think it's safe to say a 1080 easily doubles a 680. Though if I recall, that 680 was a beast.

LOL the GTX 1080 easily doubles a 980, let alone a 680. I'd probably say a GTX 1080 = 4 680s.

 

[turkey3_scratch]

The 1080 is not quite double the performance of a 980. It seems to be more about 80% better, but less better under lower resolutions and more better under higher resolutions.

 

[TechyInAZ]

The 1080 is not quite double the performance of a 980. It seems to be more about 80% better, but less better under lower resolutions and more better under higher resolutions.

Sorry your right. I keep getting VR performance mixed with non VR benchmarks.

 

[17seconds]

The 1080 is still almost 4X the performance of a GTX 680 (a few percent less than a GTX 770), at least for the Gigabyte G1.

 

[TechyInAZ]

MERGED QUESTION
Question from SoleSurvivor111 : "Most powerful gtx 1070"

What gtx 1070 is the most poweful without overclocking like the evga ftw or msi gaming x

 

[melenie]

A question for you more knowledgable people out there.

So FE overclocks to the same limits as current custom cards (cooling aside). Do you think the 1 power connector will be a limiting factor if a custom bios was to surface which would enable us to push passed 1.25v? Or is there enough headroom with the FE gpu's that again, dual power connectors does not provide better overlocks with the 10 series gpu's.

The reason I ask is I currently have the opportunity to obtain reference cards at aprox 10% cheaper then the price i can purchase a custom. Cooling is no issue for me as I will strip whatever card I end up with as they are going into an open watercooled loop. Not to mention I can't even get my hands on a custom board partners card.

Just really want to know if im going to be substantially limited with the one power connector etc.

 

[U6b36ef]

A question for you more knowledgable people out there.

So FE overclocks to the same limits as current custom cards (cooling aside). Do you think the 1 power connector will be a limiting factor if a custom bios was to surface which would enable us to push passed 1.25v? Or is there enough headroom with the FE gpu's that again, dual power connectors does not provide better overlocks with the 10 series gpu's.

The reason I ask is I currently have the opportunity to obtain reference cards at aprox 10% cheaper then the price i can purchase a custom. Cooling is no issue for me as I will strip whatever card I end up with as they are going into an open watercooled loop. Not to mention I can't even get my hands on a custom board partners card.

Just really want to know if im going to be substantially limited with the one power connector etc.

I would watch this vid as JayzTwoCents explains where and how the FE 1080 throttles. It can be worked around, but I dunno if it's worth it. Probably though, but up to you. https://www.youtube.com/watch?v=OXUo1S55ZUM

 

[Math Geek]

if i recall right, gamers nexis took a 1080 FE and put a custom cooling set-up on it. one of those aftermarket 3 fan ones. they were still able to push it to the same numbers but the temps were way down. so it obviously was limited by the voltage and not thermals. i would assume that a small voltage increase via a custom bios would raise the temps a bit but with watercooling it would be easy to handle.

the FE was hitting 220w or so which is about the max for the 8 pin,. so i'd assume more voltage would hit power limits for the 8-pin as well. personally, i'd want that extra 6-pin to be able to take advantage of the voltage increases of a custom bios.

 

[U6b36ef]

I found a chart that seems the GTX 680 and other cards on different PCI-E slots.

http://www.tomshardware.co.uk/forum/386054-33-would-pcie-slot-bottleneck

I tried to copy the graph in. Didn't work, sorry. Anyway I hope it answers something. It does seem to show a scaling from PCI-E 2.0 by x8 and x16. Faster performance on x16. However it might not be a card bottleneck and more of a characteristic. I suspect that is what is being seen in the data on the 1080 on different slots.

Anyway I don't know but the reason I recall PCI-E 1.0 and 1.1, was because I was buying a GTX 650 Ti Boost once. I was wondering what the limit was on 1.1.

 

[melenie]

A question for you more knowledgable people out there.

So FE overclocks to the same limits as current custom cards (cooling aside). Do you think the 1 power connector will be a limiting factor if a custom bios was to surface which would enable us to push passed 1.25v? Or is there enough headroom with the FE gpu's that again, dual power connectors does not provide better overlocks with the 10 series gpu's.

The reason I ask is I currently have the opportunity to obtain reference cards at aprox 10% cheaper then the price i can purchase a custom. Cooling is no issue for me as I will strip whatever card I end up with as they are going into an open watercooled loop. Not to mention I can't even get my hands on a custom board partners card.

Just really want to know if im going to be substantially limited with the one power connector etc.

I would watch this vid as JayzTwoCents explains where and how the FE 1080 throttles. It can be worked around, but I dunno if it's worth it. Probably though, but up to you. https://www.youtube.com/watch?v=OXUo1S55ZUM

Thanks for the link. Although it is not directly related to my concern, it provided some good info overall.
Thermal throttling will not be an issue for me. My card will be water cooled. My concern is with throttling due to power limitations. Would be cool if there was a video or some in depth review of that (there probably is stashed away somewhere)
=]

 

[melenie]

if i recall right, gamers nexis took a 1080 FE and put a custom cooling set-up on it. one of those aftermarket 3 fan ones. they were still able to push it to the same numbers but the temps were way down. so it obviously was limited by the voltage and not thermals. i would assume that a small voltage increase via a custom bios would raise the temps a bit but with watercooling it would be easy to handle.

the FE was hitting 220w or so which is about the max for the 8 pin,. so i'd assume more voltage would hit power limits for the 8-pin as well. personally, i'd want that extra 6-pin to be able to take advantage of the voltage increases of a custom bios.

Ahh as always, great info here! I did watch the frankenstein build with a 980ti EVGA water block on the FE 1080 a couple weeks back. Looking at power draws wasn't on my mind back then though haha

 

[-Fran-]

http://techreport.com/review/30281/nvidia-geforce-gtx-1080-graphics-card-reviewed

TechReports 1080 review. They were really disappointed with the cooling of the FE; other than that, in-line with all non-DX12 benchmarks out there.

Now we're just missing AnandTech's.

Cheers!

 

[Math Geek]

wow, they are a bit behind. was expecting a custom card review not a month old FE review. i'm a little surprised at how few custom reviews are out there. especially for the 1070 cards. i just expected more reviews by this point after release.

 

[jeffler383]

https://youtu.be/mWcsaociTjE
This is a very interesting video from hardware unboxed about the high bandwidth sli bridge. From his results it appears it does actually make a difference, but only on some titles at 4k under max settings. To me, this seems like there aren't really any games out there that require that much graphics horsepower, even at 4k, to make the sli bridge a bottleneck. Maybe in the future it will change, but for now the conclusion is still the same; not really needed.

Also interesting is that he got the same results using 2 legacy bridges at the same time.

 

[turkey3_scratch]

I'll be sure to remember next year that the actual release date of Nvidia's 1100 series will be 2 months after the release.

 

[kinnecal]

I've tried posting elsewhere so I guess I'll try here too. I just got a GTX 1070 MSI FE today and it is seriously underperforming. like 20 fps in the witcher 3, 40 fps in DS3, FFXIV, Overwatch and GTA, so way under where I expected it to be at. I have reinstalled the drivers using DDU, I have changed all of the power settings in my PC and Nvidia control panel, I have turned vsync on and off, I have tried pretty much everything, but it is still underperforming. Do you think my card is defective? What step should I take next?

Also this may be unrelated, but when I go to the led visualizer, it says 0 mV for the voltage reading, is that my problem?

A pic of the led visualizer

 

[melenie]

First,
What compnents are you running?
Power supply
Mother board
Cpu

Power supply is my suspicion. Anyway.


https://www.youtube.com/watch?v=OXUo1S55ZUM

Watch this. Follow what he does to overcome thermal throttles.

 

[kinnecal]

First,
What compnents are you running?
Power supply
Mother board
Cpu

Power supply is my suspicion. Anyway.


https://www.youtube.com/watch?v=OXUo1S55ZUM

Watch this. Follow what he does to overcome thermal throttles.

Was this for me?

Power Supply: Thermaltake 850W
Motherboard: AsRock Z97E-ITX/ac
CPU: i7-4790k
and 16 GB of ram.

I'm going to try the settings from the video you posted and I'll report back after that

 

[Deleted member 217926]

This is the wrong place for help. Make another thread and I'll allow you to post a link to it in here.

These Megathreads need to stay mostly on topic.