What Makes NVIDIA's GeForce GTX 1080 So Fast?

H

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
The crew at VR World think they have a pretty good handle on what makes NVIDIA's GeForce GTX 1080 so fast. Definitely good reading, hit the link to read the full article.

Stock GTX 1080 is clocked at 1.66 GHz, with Turbo Boost lifting it to 1.73 GHz. Founders Edition includes overclocking-friendly BIOS to raise the clocks to at least 2 GHz, and the presentation showed the chip running at 2.1 GHz. The main limiting factor for the overclocking beyond 2.2 GHz is 225 Watts, which is how much the board can officially pull from the power circuitry: 75 Watts from the motherboard and 150 W through 8-pin PEG connector. However, there are power supply manufacturers which provide more juice per rail, and we’ve seen single 8-pin connector delivering 225 W on its own.
 
No benchmarks = tuning out those kinds of marketing messages. Clock speed increases are a little interesting, but not really indicative of performance.

The most interesting part of that GPU so far is its unexpectedly low power consumption.
 
Whoa, this the first time I've heard this said in anything other than hushed whispers:

We were told that the Pascal architecture doesn’t have Asynchronous Compute, but that there are some aspects of this feature which qualified the card for ‘direct12_1’ feature set.
Click to expand...
 
Creig said:
Whoa, this the first time I've heard this said in anything other than hushed whispers:
Click to expand...


Like I said in other threads. No way the card could have had Async in the time frame provided.

However, given how Polaris is potentially just a cheaper 390 (in speed if not in architecture) that uses less power, There really isn't anything standing in Nvidia's way at this point given the significant gain they made over their previous generation, and the better pricing compared to that previous generation.

AMD and Polaris will have to have a real serious surprise in store at the end of May, else the reports on Polaris's speed need to be way off for AMD to at least stem the tide. Perhaps the only other thing that can happen is the card ends up much cheaper than rumored, and you can Tri-fire 3 of them for cheaper than a 1070 (which let's be honest, isn't likely).
 
Basically they made different chips this time. Instead of one chip trying to fit all market segments. So no need to compromise between a chip for gaming and one that does deep learning.
 
AMD has caught Nvidia with their pants down in the past. Remember the 4850/4870 launch? Completely rained on the recently launched 280/260 and forced Nvidia to immediately cut the MSRP on both cards by significant amounts. I'm not saying that's the situation here, but it isn't out of the realm of possibility. We won't know until the benchmarks.
 
I think Polaris is just going to give 390 / 390x performance to the $150 - $200 bracket. I wouldn't be surprised to see the 380 as the entry level $120 card.
 
Stoly said:
I stopped reading at the overclocking friendly BIOS.
Click to expand...
Yea, I only forced myself to finish reading it. I was under the impression that the Founders Edition didn't have anything different other than being the new nomenclature for Reference.
 
I was worried a bit about my 980 ti becoming obsolete overnight....maybe not? The more I hear about the 1080, the more I'm convinced it's barely an upgrade.
 
pxc said:
No benchmarks = tuning out those kinds of marketing messages. Clock speed increases are a little interesting, but not really indicative of performance.

The most interesting part of that GPU so far is its unexpectedly low power consumption.
Click to expand...

Not that unexpected for a Gxxx4 card. The single 8 pin is equivalent to two 6 pins in terms of rated power delivery, and they have historically been rated in the 180 watt range at stock.
 
My current GPU is close in power to a TI83 Plus graphing calculator so I'll for sure be buying into Pascal :)
 
First i have heard about a water cooled version at launch, that is some impressive clocks on that thing if those rumors are true
 
I thought it was because of the special nvidia cakes they were feeding the tiny green hamsters.
 
Last edited:
Hype always makes me think we got a dud coming. Still Ill wait for reviews instead of eating all this hand fed PR Bs that has been coming down the pipes.
 
Last edited:
{NG}Fidel said:
This is so pathetic. Its so fast but we are not willing to give out reviews.
Hype always makes me think we got a dud coming. Still Ill wait for reviews instead of eating all this hand fed PR Bs that has been coming down the pipes.
Click to expand...
Can you taste the marketing? I smell it in my nose, but yet taste it in my mouth. It tastes, green.

linus-likes-the-nvidia-shield-o.gif
 
JohnC0817 said:
Yea, I only forced myself to finish reading it. I was under the impression that the Founders Edition didn't have anything different other than being the new nomenclature for Reference.
Click to expand...
Read somewhere that the chips on Founder's Edition are supposed to be binned differently, thus better overclocking potential. Time will tell if that's true or not.
 
First off, I'm not even convinced yet that it's "so fast".
 
Shameless Liar said:
Read somewhere that the chips on Founder's Edition are supposed to be binned differently, thus better overclocking potential. Time will tell if that's true or not.
Click to expand...

Regular Edition will have up to 3 of the 8 pin power connectors where the Founders Edition will have 1.
 
Shameless Liar said:
Read somewhere that the chips on Founder's Edition are supposed to be binned differently, thus better overclocking potential. Time will tell if that's true or not.
Click to expand...
Nvidia themselves have already stated that the Founders Edition cards are not binned and don't come pre-overclocked.
 
Last edited:
  • Like
Reactions: N4CR
like this
This is the part that got me...
However, there are power supply manufacturers which provide more juice per rail, and we’ve seen single 8-pin connector delivering 225 W on its own.
Click to expand...
I thought it had to do with not overloading that gauge of wire and the resistance of the metal connector pins to the point the wire's insulation and the plastic plug melt or caught fire.

I used this chart and the rating column for "max amps for power transmission".

Just did some math...
150w 8pin plug at 12v = 12.5 amps total provided. That's 3.125 amps per yellow wire (4 of them).

16 gauge wire is typically used on PSU's for most cables.
16 awg is rated for 3.7 amps. This gives a little wiggle room for most cables.

Let's assume the max...
3.7 amps over 4 wires is 14.8 amps provided = 177.6w total...almost 50 watts short of this statement.

To provide 225w on one 8pin connector, you would have to use 14 awg wire.

That's all assuming the metal pins in the plug can handle just as much load as the wire.

Anyone with electrical expertise care to chime in? Please correct any mistakes.
 
Last edited:
16 awg wire can handle more like 15a for short distances, it's the cable pins that may have issue.
 
CaptNumbNutz said:
This is the part that got me...

I thought it had to do with not overloading that gauge of wire and the resistance of the metal connector pins to the point the wire's insulation and the plastic plug melt or caught fire.

I used this chart and the rating column for "max amps for power transmission".

Just did some math...
150w 8pin plug at 12v = 12.5 amps total provided. That's 3.125 amps per yellow wire (4 of them).

16 gauge wire is typically used on PSU's for most cables.
16 awg is rated for 3.7 amps. This gives a little wiggle room for most cables.

Let's assume the max...
3.7 amps over 4 wires is 14.8 amps provided = 177.6w total...almost 50 watts short of this statement.

To provide 225w on one 8pin connector, you would have to use 14 awg wire.

That's all assuming the metal pins in the plug can handle just as much load as the wire.

Anyone with electrical expertise care to chime in? Please correct any mistakes.
Click to expand...

That site is terrible and does not provide useful information.

Quick example, I with my quad copter put about 80amp at 12v through a 12awg cable and it just gets warm.

Here is a decent basic calculator

DC Cable Sizing Tool - Wire Size Calculator - MM2 & AWG - solar-wind.co.uk

You were on the right track and your calculations were correct, it was just the source you referenced that lead you incorrectly.
 
Darunion said:
That site is terrible and does not provide useful information.

Quick example, I with my quad copter put about 80amp at 12v through a 12awg cable and it just gets warm.

Here is a decent basic calculator

DC Cable Sizing Tool - Wire Size Calculator - MM2 & AWG - solar-wind.co.uk

You were on the right track and your calculations were correct, it was just the source you referenced that lead you incorrectly.
Click to expand...
Thank you!

-edit- nvm I used wrong units.
I also see why my original chart was wrong. It was intended for AC not DC. I'm not sure if it even specified stranded or solid wire either. *sigh* big fail on my part.
 
Last edited:
Darunion said:
That site is terrible and does not provide useful information.

Quick example, I with my quad copter put about 80amp at 12v through a 12awg cable and it just gets warm.

Here is a decent basic calculator

DC Cable Sizing Tool - Wire Size Calculator - MM2 & AWG - solar-wind.co.uk

You were on the right track and your calculations were correct, it was just the source you referenced that lead you incorrectly.
Click to expand...
So my original thought still remains...

Was the choice for max of 12.5amps/37.5watts per each of the 4 wires have to do with the wire and/or plug rating or did it have to do with the rails on the PSU?

Using your calculator and assuming 1% loss for a .75 meter cable (Corsair cables are 750mm) and 3.125amps puts the cable size at 18 awg.

Assuming 1% loss for a .75 meter cable for a .75 meter cable and 4.6875 amps puts the cable size at 16 awg. 4.6875 amps per cable would get 225 watts like they mentioned.

I usually see 16 or 18 awg on most PSU's. Hmmm...
 
Last edited:
Actually there are 3 12v wires on an 8-pin PCI-E connector. The other 5 are ground wires, with two of them being sense connections.
 
This graph looks hilarious if you plot back the point where the Titan X is, to below the 1080p.

o04jEnK.jpg
 
CaptNumbNutz said:
So my original thought still remains...

Was the choice for max of 12.5amps/37.5watts per each of the 4 wires have to do with the wire and/or plug rating or did it have to do with the rails on the PSU?

Using your calculator and assuming 1% loss for a .75 meter cable (Corsair cables are 750mm) and 3.125amps puts the cable size at 18 awg.

Assuming 1% loss for a .75 meter cable for a .75 meter cable and 4.6875 amps puts the cable size at 16 awg. 4.6875 amps per cable would get 225 watts like they mentioned.

I usually see 16 or 18 awg on most PSU's. Hmmm...
Click to expand...

I would better assume a 2-3% acceptable line loss because it won't require that high of precision, the power supply should be supplying more than 12v to account for this and so you don't brown out when loading it. As stated you only calculate by the pair so current running from the source line is all you calculate, ground wires do not supply it. The wires really aren't going to be your limiting factor, it will be at the connector as this is where your resistance will be higher there will be heat generated.

I would comfortably put 10 amp through a 16awg wire. Yea as you said many psu manufacturers put 16awg wire on 12v/40A rails. There is also the discussion of them not really specifying anything other than a line graph with wattage at the bottom. We likely are digging too deep into their marketing numbers.
 
GTX 1080 Secret - Looking at the BIOS we can deduce that its run by Hookers, Midgets, and Oreo Cookies! O_O
 
You must log in or register to reply here.
Back
Top