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

[NirK]

980ti is dead now.

Not really bad if you get one cheap on ebay for almost half the amount the 1080 is going to cost, it will likely still beat the 1070 at higher resolutions, or at least "trade blows" with it. I think the difference between the 1080 and 1070 is going to be more than usual eg. 970 vs 980... mostly because of the GDDR5x.

 

[Mei]

So it seems the 1080 can be used in 3 way sli,according to Jaystwocents latest video.

He talks about it starting at the 6 min mark in his video.

https://www.youtube.com/watch?v=hNvfX0QZbYU

well AMD's dont need the bridge no more? so *shrug* maybe it doesnt need them or maybe the old ones work just as well, if so why the new ones...
its a silly thing to be under NDA i dont think you can blame ppl for speculating on that
i mean if some ppl need to plan a big purchase of 4 cards and waterblocks why keep that info from them

 

[ebilcake]

That's a nice improvement, as soon as this is released my two GTX980's will be setting sail on ebay.

 

[karik666]

well AMD's dont need the bridge no more? so *shrug* maybe it doesnt need them or maybe the old ones work just as well, if so why the new ones...
its a silly thing to be under NDA i dont think you can blame ppl for speculating on that
i mean if some ppl need to plan a big purchase of 4 cards and waterblocks why keep that info from them

I agree mate,it does seem stupid to keep the sli stuff under nda.

 

[Smrtka]

So all of you who said it was ugly and fake aren't buying it then?

I ain't buying the card cos it ugly ! i want it to be pretty and pink!

 

[Rexxar]

It doesn't matter what the reference shroud looks like. The vast majority of cards will have 3rd party coolers.

This, i very much doubt if many cards used by serious gamers will look like this. Almost every 3rd party designs their own cooler.

 

[mahius]

why have there been no 1070 benchmarks? this card is likely to massively outsell the 1080.

Because Nvidia wants you to forget that the 1070 exists in an attempt to try and get you to buy the 1080 (lol jk), which would likely be slightly worse price-to-performance than 1070. Ala 970 vs 980. The difference in price (at least before the 980ti) wasn't worth the smaller difference in performance, when I was considering a 980 for <£400, I was advised that the 970 would be better for the money instead.

In all seriousness, this is a new card launch. They want to show of their biggest guns, to try and sell those more, not the ones that they know already will sell well. And also just to show of the raw power really. The 1080 is the best example of what Pascal is capable of.

http://videocardz.com/59882/nvidia-geforce-gtx-1080-3dmark-overclocking-performance

Spoiler

I'm inclined not to be excited about those benchmarks. Only because the 1080 was the only thing overclocked. Overclock all of the cards and then compare. Of course one card will look much better if you only overclock that one and not the others. Thus the more realistic performance gap is more representable by the previous chart where all the cards were at stock (hopefully). Though it is just a benchmark, it would be much more useful to see the performance in actual games.

Despite being rather cynical about these releases, I'd honestly expect the 1070 to beat the Titan X anyway. Just because the 970 did similarly to the Titan and 780ti. Thus the 1070 beating Titan X is nothing special to me, it's the least I would expect from a die-shrink card.

And let's not forget about what's to come. If the 980ti was comparable to 970 SLI (including imperfect SLI optimisation), then we can expect the 1080ti to be something amazing. And considering that Nvidia seem to be more serious about SLI support, narrowing it down to just 2-way, this could be a generation where SLI is more common and provides more actual benefit than it has before. Thus I expect Pascal to scale well. That or Nvidia is just cutting 3/4 way SLI and not bothering with SLI in general. I think the former is more likely.

 

[Mei]

is there even a need for them 3 fan coolers if this thing stays at 67c oc'd on a blower!?
i guess it needs to be kinda long for all them ram chips anyway i duno
we see i guess!!

 

[Legend]

You don't even see that side of the card, so I don't know why people are complaining. Unless you're one of that very small fraction who have their GPU vertically oriented, you will NEVER SEE IT! I'd rather they put more effort in to the back-plate design, as that's what I'm looking at, and the most visible thing through the window of my case. The Zotac 1080 looks interesting in that regard, although they've managed to get the logo the wrong way round!!

 

[Mei]

I'm inclined not to be excited about those benchmarks. Only because the 1080 was the only thing overclocked. Overclock all of the cards and then compare.

what?? they said only the 1080 was overclocked?? or you just make that up? lol
i thought they all stock

 

[beany_bot]

Would I be right in guessing that the bulk of us serious gamers fall into the (970-980, 390-390x) category?

 

[NirK]

Gaining more % performance than the % clock increase? Must be the magic sauce.

Mm that is a good point actually. Maybe it is fake. Hopefully it is legit because that is the amount I was hoping for (30%) vs a 980ti.

It says the 1080 gains 24% performance from a 13-20% increase in clock speed, depending what the stock boost speed was, maybe they OC the memory as well or something or could be fake IDK.

If the boost clock is 1733mhz exactly, and the overclock is 2150mhz, and prob the memory OC as well, and the scaling for the OC is exactly 100% then the 24% could be true.

I just read it again...

It says "The card being really cool and efficient allows clocks to boost up to 1860 MHz"

So a 2.1Ghz overclock from that is only about a 13% overclock, which is says is 24% higher score, But then 3Dmark does make cards hit power limit and throttle so maybe it was not at 1860, seems a bit fishy.

 

[Jono8]

what?? they said only the 1080 was overclocked?? or you just make that up? lol
i thought they all stock

They are all stock in the first set then the 1080 is overclocked and the rest are stock in the second set. In fact the second set just adds the overclocked 1080 to the rest..

 

[NirK]

Has any card in the history of graphics cards ever increased by 24% performance from a 14-20% overclock?!?!?!? Hmmm color me suspicious of that result!!!

1860mhz x 1.24 = 2306mhz

1733mhz x 1.24 = 2148mhz

As the 1080 is supposed to be efficient and at low temps etc. then I guess it would boost to 1860mhz at stock without hitting any power or temp limits and throttling, which would make the 24% at 2.1Ghz highly suspicious!

 

[CAT-THE-FIFTH]

Has any card in the history of graphics cards ever increased by 24% performance from a 14-20% overclock?!?!?!? Hmmm color me suspicious of that result!!!

Seems a bit weird - but was the memory overclocked though?? The GTX1080 might be very memory bandwidth limited,and the current modules are quite slow for GDDR5X too as they appear to be a special run. It makes me wonder if Nvidia will start using faster modules at some point.

 

[NirK]

Seems a bit weird - but was the memory overclocked though?? The GTX1080 might be very memory bandwidth limited,and the current modules are quite slow for GDDR5X too as they appear to be a special run. It makes me wonder if Nvidia will start using faster modules at some point.

If the first stock result, the card was at 1733 stock (lowest boost speed)

Then it was overclocked to 2150mhz, and the memory was overclocked, and it was 100% scaling, overclock = score. Then the 24% would be possible. Otherwise it looks high.

 

[Besty]

Lots of variables, too many to discuss without seeing more testing.

i.e. They could have stuck the fans on high and the card boosted harder for longer.

 

[D.P.]

Mm that is a good point actually. Maybe it is fake. Hopefully it is legit because that is the amount I was hoping for (30%) vs a 980ti.

It says the 1080 gains 24% performance from a 13-20% increase in clock speed, depending what the stock boost speed was, maybe they OC the memory as well or something or could be fake IDK.

If the boost clock is 1733mhz exactly, and the overclock is 2150mhz, and prob the memory OC as well, and the scaling for the OC is exactly 100% then the 24% could be true.

I just read it again...

It says "The card being really cool and efficient allows clocks to boost up to 1860 MHz"

So a 2.1Ghz overclock from that is only about a 13% overclock, which is says is 24% higher score, But then 3Dmark does make cards hit power limit and throttle so maybe it was not at 1860, seems a bit fishy.

The stock clock is 1.6GHz, 2.1/1.6=30%. When over clocked to 2.1GHz thete is still boosting going on.

 

[Rroff]

i.e. They could have stuck the fans on high and the card boosted harder for longer.

That is one thing we have little idea of yet - how much it actually holds its boost clocks and/or the behaviour of them - it could be that it is boosting more in some sections than the 2.1GHz clock even or it could be the out the box boost is tanking, etc. etc.

 

[smallzz]

I was looking at buying a 980ti to replace my R9 290 2 weeks ago... the timing couldnt be more perfect.