NVIDIA Pascal Thread

[guskline]

What's the max power target?

Deders went home for lunch and fired up MSI Afterburner ver 4.3Beta4 and was able to more the power up to 120%

 

[tviceman]

Guru3d mentioned that all 1080's are basically limited by Nvidia at a hardware or some other deep level preventing high overclocks. That ROG Strix only goes to 2050mhz. Same as the 600 series. Mid range, gimped overclocking. Must wait for real cards to come out if you want good OC's.
This whole thing still pisses me off. These are mid range cards. They don't even OC very well. They are not much faster than existing flagship cards. They are only exciting at a mid range price point. That's the whole reason for them to exist. They are exciting at a mid range price point, but disappointing and lack luster at flag ship pricing because they are very obviously not a flagship caliber of product.

I'm as turned off as the next guy about pricing, but even if they're not maxwell overclocking, to say they don't OC very well is flat wrong. Hawaii didn't OC well. Fiji didn't OC well. Those are my definitions of poor overclockers. With adequate cooling, GP104 is OCing decently.

http://www.techpowerup.com/reviews/ASUS/GTX_1080_STRIX/27.html 17% faster than reference speed is nothing to be kicking the ground over.

 

[tviceman]

Nvidia gimps the OCability of the mid range cards so that they won't step on the toes of their over priced high end chips.

The entire Maxwell lineup overclocked equally on a percentage standpoint. It's not that Nvidia is gimping OCing; they are spacing out the baseline performance at each price tier more and more so as to not have products overlap one another.

 

[antihelten]

I'm as turned off as the next guy about pricing, but even if they're not maxwell overclocking, to say they don't OC very well is flat wrong. Hawaii didn't OC well. Fiji didn't OC well. Those are my definitions of poor overclockers. With adequate cooling, GP104 is OCing decently.

http://www.techpowerup.com/reviews/ASUS/GTX_1080_STRIX/27.html 17% faster than reference speed is nothing to be kicking the ground over.

17% faster than reference is pretty meh, when you could get twice that (34%) from 980 Ti aftermarket cards.

Of course one could argue that this is not so much a case of GP104 being mediocre as it is a case of GM200 being exceptional, but it still puts Nvidia's claims of "crazy overclocker" for GP104 in perspective.

Also if Hawaii is your definition of a poor overclocker then GP104 should also count as a poor overclocker, seeing as Hawaii got roughly the same 12-17% that GP104 is showing. (Hawaii overclocks on TPU: 15%, 20%, 12%, 14%, 10%, 12%)

 

[guskline]

bumped my vcore +300 and got a freeze on the GTX 1080. +200 no problem. That is without any volt mods etc. Same thing happened on my GTX 980 TI.

Truthfully, both of these video cards are so fast you don't need to OC except for bragging rights.

Since I own both a GTX 1080 and GTX 980 TI SC they "feel" similar but the 1080 clearly puts up better benchmark scores. More to follow.

BTW, the "over priced" comments are funny. ANY new item probably is more expensive than we want to pay and some people, myself included, pay more to be first.

My goal was to get the best, or near best, single GPU, for both of my rigs and I think I accomplished that. I don't want CF or SLI.

Your goal may be different. I realize my quest likely will be eclipsed when AMD/Nvidia releases their next greatest gpu.

For me, I'm set for BF1 this fall on both machines.

 

[tviceman]

17% faster than reference is pretty meh, when you could get twice that (34%) from 980 Ti aftermarket cards.

Of course one could argue that this is not so much a case of GP104 being mediocre as it is a case of GM200 being exceptional, but it still puts Nvidia's claims of "crazy overclocker" for GP104 in perspective.

Also if Hawaii is your definition of a poor overclocker then GP104 should also count as a poor overclocker, seeing as Hawaii got roughly the same 12-17% that GP104 is showing. (Hawaii overclocks on TPU: 15%, 20%, 12%, 14%, 10%, 12%)

All points well made!

 

[AdamK47]

HB SLI bridges are now available from Nvidia.

 

[LTC8K6]

HB SLI bridges are now available from Nvidia.

Now for the reviews.

 

[MrTeal]

How does one define a percentage overclock in Boost 3.0 age, when the reference clock isn't really a fixed number? The spec page lists 1607MHz and 1734MHz, but if out of the box you average a boost to 1800MHz and through overclocking you bring that to 2000MHz is your OC 11%, 15%, or 24%? With my Hawaii cards it's pretty simple, when I crank it up to 1200 that's a 27% overclock over the fixed 947MHz base frequency.

Is there a generally agreed upon consensus as to what numbers to use for this?

 

[Termie]

How does one define a percentage overclock in Boost 3.0 age, when the reference clock isn't really a fixed number? The spec page lists 1607MHz and 1734MHz, but if out of the box you average a boost to 1800MHz and through overclocking you bring that to 2000MHz is your OC 11%, 15%, or 24%? With my Hawaii cards it's pretty simple, when I crank it up to 1200 that's a 27% overclock over the fixed 947MHz base frequency.

Is there a generally agreed upon consensus as to what numbers to use for this?

The only way this can be done accurately is to test the actual in-game boost clock of a reference model. HardOCP has kindly done that for us, which you can see here:

Source: http://www.hardocp.com/article/2016...rs_edition_overclocking_preview/#.V2ROYPkrIuU

And because boost is cut as the temperature hits 70C and then 83C, you have to look at it over time. The maximum the Founders Edition can sustain is around 2060MHz. Ironically, this is about what all other custom cards are hitting as well, despite better cooling. So a chip that can hit 2100MHz is a golden sample. Boost at stock clocks looks to be around 1780MHz. That translates to around 15-17% overclocking headroom, regardless of the card you choose.

In my opinion, the printed base and boost clocks are almost completely irrelevant, other than the actual boost offset is based off of them.

 

[AdamK47]

Now for the reviews.

I ordered mine (4 slot). It should be getting to me next week. I'll see if there is a difference with it compared to the EVGA V2 LED bridge.

 

[guskline]

Thanks for the info AdamK47. I got a Zotac GTX 1080 FE and added an EK waterblock. Boy these GTX 1080s are fast.

 

[imaheadcase]

HB SLI bridges are now available from Nvidia.

Now if people could just get 1080 to use them..

Seriously, this was not a good way to launch cards. Not only did founder cards scream "We don't have much in stock", the jumped the price up.

So you think no big deal, i just wait for cheaper ones. But the non-FA cards are being price gouged 50-100 more than normal.

I'm all for waiting for stuff, but if you are going to announce a launch, do it when you have a huge stock on hand already.

I kind of wish cards could only be bought from respective websites from companies, Nvidia/evga/Msi/etc.

 

[antihelten]

How does one define a percentage overclock in Boost 3.0 age, when the reference clock isn't really a fixed number? The spec page lists 1607MHz and 1734MHz, but if out of the box you average a boost to 1800MHz and through overclocking you bring that to 2000MHz is your OC 11%, 15%, or 24%? With my Hawaii cards it's pretty simple, when I crank it up to 1200 that's a 27% overclock over the fixed 947MHz base frequency.

Is there a generally agreed upon consensus as to what numbers to use for this?

There isn't really a straightforward way to do it with boost cards, which is why all of the percentages in my post above were actually based on the performance gain from overclocking. This is really the only straightforward way to compare nvidia's and AMD's cards that I can think of.

 

[Deders]

Anybody have any idea how many watts the led's on the asus strix are likely to take up?

 

[Freddy1765]

Anybody have any idea how many watts the led's on the asus strix are likely to take up?

1? What kind of power-budget are you on that makes it in any way a cause for concern? I'd guess a case fan uses an order of magnitude more power than LEDs on graphcis cards.

 

[Deders]

1? What kind of power-budget are you that makes it in any way a cause for concern? I'd guess a case fan uses an order of magnitude more power than LEDS on graphics cards.

At the moment 3Dmark's stress test is pointing out to me that because my fans rev up to tame the heat after 10 mins or so, this makes enough of an impact on my power envelope that my gpu clockspeed drops by 63-79MHz.

I'm trying to get some hard data on what the different cards base wattage is, and what percentage I can raise that by. Apparently some cards are rated at different tdp's than the founders edition.


Edit:

Apparently the MSI gaming X gives you a 121% power target beyond its base 180W, no led's.

The Asus strix gives you 120%, but I can't find confirmation on its base wattage. If it's 200w like I think I read somewhere then it's a no brainer (My case has no window). If it's 180w. Then the extra 1% and lack of led's clinches it for me. (Also 2 fans vs 3?)

I'm also wondering how much the pumps of the watercooled models will eat into the power target. No point having extreme cooling if it's limited power wise.

 

[wingman04]

The little LED bulb only uses 0.02 Watt/Volt. However, the 8-pin is designed to supply up to 150W and the PCI-E socket supply's 75 watts=225 watts. Graphics Card Power (W) 180watts

 

[Deders]

The little LED bulb only uses 0.02 Watt/Volt.

And it's just 1-3 bulbs that can get all those colours?

 

[kithylin]

And it's just 1-3 bulbs that can get all those colours?

I'm not sure you're understanding. It would take 50 of those bulbs on a card to even use 1 watt of power. And it looks like most cards have like what, 5? 8? maybe 15 of em? 15 would be 0.3 watts used.

Really, LED's or not should not even remotely be anything that you consider in a video card when you're thinking of how much power it uses. They're incredibly insignificant.

 

[Deders]

Good to know, thanks.

 

[kithylin]

For those of you that didn't catch this.. I'll just leave this here: http://videocardz.com/61121/asus-and-msi-accused-of-sending-modified-cards-to-the-press

 

[Freddy1765]

For those of you that didn't catch this.. I'll just leave this here: http://videocardz.com/61121/asus-and-msi-accused-of-sending-modified-cards-to-the-press

This was posted days ago. And it's such a non-issue I cannot fathom anyone cares.

 

[Sweepr]

TechPowerUp's NVIDIA GeForce GTX 1080 SLI Review

www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_1080_SLI/1.html


Edit: Impressive showing for Pascal in The Witcher 3: Blood and Wine. Geforce GTX 1080 is 40-50% faster than Geforce GTX 980 Ti.

http://pclab.pl/art70150.html

 

[guskline]

I'm done with CF/SLI.