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

Page 14 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
The HB SLI uses both SLI fingers on each card for double the bandwidth.

NVIDIA-GeForce-GTX-1080-HB-SLI-Bridge-1-635x847.jpg

http://wccftech.com/nvidia-geforce-gtx-1080-pcb-gp104-gddr5x/
 


It's a good idea, since SLI bridges and CFX bridges only use 1 finger on each card. Allowing both fingers to work together now allows better performance. But I'm confused as to why SLI bridges haven't been doing this for years.
 


Wait what? Well, I guess it's understandable, given how bad scaling is with 3 and 4 way SLI previously.

Hopefully DX12 fixes that. 😀
 
MERGED QUESTION
Question from culzone : "1070 sli vs 1080"

How much better do you think 1070 SLI would be compared to a single 1080?

Maybe same difference as 970 SLI vs 980? HMm? thoughts?
 


Twin 1070s will definantly be quicker than a single 1080 for sure.
 
i'm interested what this new bridge can do as well. though scaling on two cards already quite good. i think what most people want to see is improved scaling on 3 way or 4 way setup. for example adding the third card not just minor performance increase from two but actually triple your FPS compared to single gpu only. and i'm interested to know why nvidia still using bridge.
 
i don't think there is any other option but abridge for sli. the tesla p100 will use the nvlink but that is still way to new to make it to consumer cards. is there even another way for the cards to talk to each other besides a bridge? i on't think the pcie slots are connected directly to be a fast enough way for them to talk back and forth
 


Well actually, the R9 390s and 390Xes don't use bridges at all. They run ALL they're communication thru the PCI-E link.

I haven't heard of any issues with doing crossfire that way, it seems to be working well. This is mainly because the PCIE 3.0 bus has so much extra bandwidth that AMD can do this with no penalty.

However, with how powerful Pascal is, it makes me wonder if SLI bridges were a need? Maybe?
 


They said at the Nvidia Live Event the other day that the GTX 1080 is faster than a SLI of 980ti and also faster than a single Titan X, and they show a performance chart.... and it looks like it works a LOT faster than a Titan X....

 
they said 1080 was faster than 980 sli not 980 ti. big difference.

plus the "faster than a titan x for the 1070 and 2x faster than a titan x for the 1080" seems to be more related to VR than regular gaming. at least that is what people are starting to think since the leaked benchmarks are nowhere close to those levels of performance.
 


Take that with a grain of salt for now. From what others have said, this is supposedly only for VR performance. The performance gap is smaller if your using a regular monitor.
 


Regarding the 1070, During the CEO's 1080/1070 'overview' towards the end (1hr 11mins in) he said the 1080 was twice the perfomrance of the Titan X. This figure clearly relates to his previous VR Slides. He then goes on to introduce the 1070... he said verbatim "6.5 GFLOPS, 8 GB G5 memory, Faster than a Titan X Faster than a titan X" - Yes he said it twice, was this intentional? perhaps....

I HOPE that the 1070 can perform the same as a titan X and it was NOT a VR comparison. Does anyone have any information about this specific claim?
 
i am expecting between 980 and 980ti performance for the 1070. but this is just an educated guess. looking at the leaked benchmarks from AoS, the 1080 matched the Fury X for the most part. this is obviously not even close to 2x a titan x which we all pretty much agree refers to VR and nothing else.

i see no reason to think that the comment for the 1070 should be taken any other way either. if it bests a titan in gaming then it would be stronger than the 1080 in this benchmark. not very likely that a 1070 is stronger than a 1080 so we can pretty much infer that the 1070 is not gonna beat a titan x in real world gaming outside VR.

now i realize there will be a lot of improvements made in drivers and game patches and this is not the only game out there to be tested. but it is all we have to go by until the NDA expires on the 17th. So it's all we got to make guesses about.
 


Just because you might not like them doesn't mean they don't exist. There are allot of youtubers out there like Jayz two cents, Linus tech tips, awesome sauce network, hardwarecanucks, pauls hardware and quite a few more. I know jayz got two, and i think luke said LTT got two as well. Paul got one.
 


And yet, this youtube confirmation does not yet exist.
It will soon (for various values of soon), but not today.

"I know jayz got two". You know this how? You skyped with jayz and he showed you the cards? Or you were the UPS guy that delivered them?
Or you heard it from a friends cousin's gardeners boyfriend?

All I see here are dozens of random questions that no one can answer yet.
"Can I...?"
"Should I...?"
"Will it...?"
 
it should be interesting to see how each site reviews them. we should get every possible game benchmarked in every possible way with the amount of cards out there. and since they all should be reference cards hopefully there will be some consistency in the results. personally i am hoping that at least one site does a cpu comparison test to see how the card does with a variety of cpu's. this is something people are already asking here and will be of interest to those thinking of upgrading. will a new i3 be enough? older i3/5? gets old seeing a bunch of test systems with x99 cpu's and tons of ram and such. i wanna see a test system like i am likely to have and not super high end stuff. how about a solid 4690k build like i have. i volunteer mine if tom's wants to send me a 1080 to test 😀