• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

** The Official Nvidia GeForce 'Pascal' Thread - for general gossip and discussions **

I currently have 2 x R9 280x's in crossfire, Would a single gtx1080 be a decent soloution to replace them and not loose any performance. I've had a google for some comparisons but it seems no-one has done any that i can find.
TIA

its going to be faster and more consistent in the framerate. A 980 ti would also be faster and more consistent just to give you a broader picture
 
I currently have 2 x R9 280x's in crossfire, Would a single gtx1080 be a decent soloution to replace them and not loose any performance. I've had a google for some comparisons but it seems no-one has done any that i can find.
TIA

I have two 290x in xfire. When xfire works they give around 980ti performance.

A 1080 is up to 20% faster.

And from memory my 290x are 10 to 20% faster than your 280x???? Not sure on that bit.

SO I am expecting a fairly decent 20% boost moving to a single 1080 with better minimum framerates, smoother and no crippling of the framerates by half when xfire doesnt work.

You will see a much bigger gain.

You can find reviews of 1080 out there which list a 295x2 card still in the charts which was just two 290x on one card. Your two will be slightly slower than that if you are looking to compare.
 
an article from play3r.net along the line of semi accurate

It has recently come to our attention that the availability of these highly clocked, insanely good performing cards are rather rare and a little difficult to find on the market. We’ve come up with a speculation as to why this might be. The reasons, which TechPowerUp uncovered recently, may have something to do with that.

The samples which were being sent out for review by a number of manufacturers were shipping with (sometimes) significantly higher clock speeds than the retail samples. What’s the reason? Well, one would guess that the reason is to get it to top the charts and then to make every man, dog, and pet furball rush out to buy that particular card. The reality of it is, though, that those incredibly highly clocked GPUs are stupidly difficult to attain high yields from. In fact, we have heard from our sources that there is roughly a three per cent retention rate on such highly clocked samples, which makes them difficult to assemble, and even more difficult to have enough for the retail channel, which is purely down to high demand and low yields. You’d think that would just make them more expensive, which may be true, but the reality is that there is just no way to consistently deliver these cards to the consumer without a long wait. That’s if you are even able to get one at all. You have to remember that three per cent is a very, very small number in the grand scheme of things.

In that article which I linked above, it is clear that they are only showing MSI and ASUS as being the bad guys in this story (not just for this generation, either.) However, do not be fooled. There are other manufacturers that are also having issues (from what we gather) and they include companies such as Inno3D with their ‘iChill’ series, and possibly even EVGA with their Super Clocked cards. In fact, MSI are actually one of the leaders in ensuring stock levels for their cards are high as they haven’t suffered from the same issues. Over zealous clock speeds from certain manufacturers such as ASUS, Inno3D have caused serious delays. I know that there is a k|ngp|n edition coming to market, but no one knows when just yet. It is unfair that only ASUS and MSI are being thrown under the bus in this ordeal. In fact, even GIGABYTE have a GPU with a >1930 MHz boost clock, so I’d expect that these will be strictly on a first come, first served basis only; we expect this model to have the same problems too!

You have probably all seen that the GTX 1080 can do 2100 MHz on the core, and all of the PR stuff that followed. However, the fact of the matter is that this was more than likely done on some cherry picked samples, along with added voltage which the manufacturers should not apply to the core as a default option. They don’t do it either, and this is why: the reasoning for this is pure and straightforward, it will shorten the lifespan of the GPU, something which none of us want.

If you believe we are making this up, just take a look at the amount of pre-orders on the retailer websites and then you will soon see that there may be some truth to this. There is just no way to guarantee such high clocks with a new architecture like Pascal. Unfortunately, it looks like the low yielding dies may be here to stay for a while, and it is unpredictable how long it may last.



edit : some quote from reddit about the semi accurate article content :

Well, someone is saying it says TSMC produced thousands of broken/unusable Pascal wafers, or something along that line.
 
Last edited:
That would be plausible if not for the fact that founders cards are in plentuful supply, yet the stock clocked AIB cards are also out of stock everywhere. If it was a yields issue across the whole lineup then founders cards would be short too, and if it was the top ens clocked AIB cards that were the problem then the stock clocked AIB cards should be plentiful.
 
A fair number of the cores "can" do 2100MHz or there about just not under the conditions nVidia portrayed as if they were running. Though from what I've seen so far over a range of applications on the FE cards no matter what you will hit the powerlimit sometimes bringing clocks down to a max of approx 2025MHz as your highest 100% sustained clock.

Yields are not an issue on TSMC 16nm FF.
 
Sorry if this is a dumb question, but when do you break the warranty with overclocking?

It seems like there is very little difference between the potential of the cheaper reference boards and the more expensive custom boards, so it would make sense to buy the cheap card and then OC it.

I know of course that the warranty will not be valid if I kill the with overclocking, I'm thinking more in general. Will other problems be dismissed if the card has been overclocked?
 
Last edited:
Sorry if this is a dumb question, but when do you break the warranty with overclocking?

It seems like there is very little difference between the potential of the cheaper reference boards and the more expensive custom boards, so it would make sense to buy the cheap card and then OC it.

I know of course that the warranty will not be valid if I kill the with overclocking, I'm thinking more in general. Will other problems be dismissed if the card has been overclocked?

pretty much all GPU makers will cover warranty if you overclock just using PC based software like Afterburner - there is no way they can check anyway, so even if they ask you if it was overclocked, if you say no there is no way they can prove otherwise.

Some makers will honour warranty if you fit a waterblock (MSI, Gigabyte, EVGA to name a few) as long as their is no obvious damage been cause and others won't (Gainward, Zotac)

Flashing a bios (and leaving it there when you send in for an RMA) is a warranty void as well, but again they probably won't be able to tell if you did if you put the original back on (which might not be possible if the card is dead)... EVGA will honour warranty on dual bios cards as long as the original bios is still in the number 1 position, so it doesn't matter what you do to the 2nd bios.

Hardware mods, anything that involves soldering or removing parts, is obviously warranty voiding

it has been a quite common theme that people buy the cheapest (usually reference) card they can get hold of, put a waterblock on, custom bios and they top the charts here on OCUK... when the Pascal bios tool shows up, we should get a better lay of the land. The single 8-pin connector is not an issue - the minifitjnr connector is rated at 9a per pin, so 3x12V is more than 300w... the potential problem is the 30A mosfet they've used - its rated at 30A at room temp, so as it gets hotter that figure will drop - on the stock bios they hit the power limit, if you remove the power limit and don't sufficiently cool the mosfet you could burn it out

However, the fact of the matter is that this was more than likely done on some cherry picked samples, along with added voltage which the manufacturers should not apply to the core as a default option. They don’t do it either, and this is why: the reasoning for this is pure and straightforward, it will shorten the lifespan of the GPU, something which none of us want.

this is rubbish, going from the default 1050mv to 1093mv will have absolutely no effect on longevity, what it does do is increase heat and makes the card fall over, so no they weren't increasing voltage to hit those clocks, you actually need to cherry pick a card that will hit those speeds on lower voltage if you want to do it on a FE card
 
Last edited:
this is rubbish, going from the default 1050mv to 1093mv will have absolutely no effect on longevity

This isn't strictly true, though it's likely to have no effective or significant impact the two main factors in chip degradation are heat and voltage. Increasing volts adds to both of these factors so will have an effect - but if the chip itself isn't going to be the first bit to fail anyway then the card as a whole has the same lifetime.
 
pretty much all GPU makers will cover warranty if you overclock just using PC based software like Afterburner - there is no way they can check anyway, so even if they ask you if it was overclocked, if you say no there is no way they can prove otherwise.

Some makers will honour warranty if you fit a waterblock (MSI, Gigabyte, EVGA to name a few) as long as their is no obvious damage been cause and others won't (Gainward, Zotac)

Flashing a bios (and leaving it there when you send in for an RMA) is a warranty void as well, but again they probably won't be able to tell if you did if you put the original back on (which might not be possible if the card is dead)... EVGA will honour warranty on dual bios cards as long as the original bios is still in the number 1 position, so it doesn't matter what you do to the 2nd bios.

Hardware mods, anything that involves soldering or removing parts, is obviously warranty voiding

it has been a quite common theme that people buy the cheapest (usually reference) card they can get hold of, put a waterblock on, custom bios and they top the charts here on OCUK... when the Pascal bios tool shows up, we should get a better lay of the land. The single 8-pin connector is not an issue - the minifitjnr connector is rated at 9a per pin, so 3x12V is more than 300w... the potential problem is the 30A mosfet they've used - its rated at 30A at room temp, so as it gets hotter that figure will drop - on the stock bios they hit the power limit, if you remove the power limit and don't sufficiently cool the mosfet you could burn it out

Thanks, very good and detailed answer.

I won't be getting into bios flashing or fiddling with the cooler, I would be in over my head :D

But I might go for a card like the Evga SC, which seems to be one of the cheaper cards you can get here in Denmark. I'm also looking at the Palit Gamerock, as the premium version is not that much more expensive, and it has been tested to be the fastest card out of box, but I just cant stand the look of it! Also it would be nice to have the 3 year Evga warranty.
 
Last edited:
This isn't strictly true, though it's likely to have no effective or significant impact the two main factors in chip degradation are heat and voltage. Increasing volts adds to both of these factors so will have an effect - but if the chip itself isn't going to be the first bit to fail anyway then the card as a whole has the same lifetime.

Nvidia have these locked down so tight - they throttle when the heat increases, so increasing the voltage doesnt actually increase heat, it just makes the card throttle sooner - so no it wont significantly affect longevity.

Bear in mind that CPU's that operate at around 1v are routinely run on water at 1.3v without significantly affecting their lifespan, so going from stock 1.050 to 1.093 will not have any effect.
 
Back
Top Bottom