Is buying a new GTX 1080 worth it now or wait for Volta?

[phexac]

Thinking of upgrading to the 1080 from R9 290. I am gaming on a 34" ultra-wide 1440x3440 right now and the 290 seems to not quite cut it in many places. With selling my current card plus selling Destiny 2 looks like I can nab the 1080 for $300-330 net. But would be silly if I got this and then Volta came out 2-3 months later...

 

[MagnusTheBrewer]

Neither one's price is based on value. So, it comes down to how long you're willing to wait.

 

[tviceman]

While I think Volta (or whatever is replacing Pascal on the consumer front_ will be out in the May time frame next year, no one knows for sure. If you're needing something now then get the 1080. If you can be content with what you have, wait until at least more rumors surface.

 

[phexac]

1080 is pretty close to its $500 MSRP...

 

[UsandThem]

I think at this point this late into this generation of GPUs, it would make more sense to buy a 1080ti on sale for around $699, as it will last longer gaming wise (at the resolution you stated), and when you do sell it to upgrade, it will hold its value better. The exact same thing happened with the GTX 980 vs the 980ti, which still sells for a good amount from what I have seen in the FS sub-forum.

 

[Guru]

GDDR6 starts production in Q1 2018, its on offer right now, but any real availability will be from Q1 2018, the next iteration of 16nm is also Q1 2018, so until they sort everything out and test it and announce it and make it available, etc... we are looking at a 2080 arriving in May at the earliest.

So it really depends what is "close enough" for you, you might end up waiting 6-7 months and stuck on your 290 or it may be 4 months from now. I mean it really depend on your definition and feeling of what is close enough. Ultimately unless you buy within the first 6 months of a new generation, there is always the looming thought of the next gen gpu coming along and you may end up never buying a GPU that way.

Personally for me I do not upgrade that often, but I always wait for a big change in technology, like going from GDDR3 to GDDR5, going from 28nm to 16/14nm, etc... those are the type of shifts that will provide immense performance improvement.

 

[Borealis7]

whether you live in the US or not has a bearing on the decision. i don't, so i decided to get a used GTX1080 mostly because i don't play the most demanding games, and at 1080p, so i want it to last 3-4 years. who knows, maybe even more.

 

[Qwertilot]

Probably forever (or at least until the silicon gives up the fight ) at 1080p resolution.....

 

[happy medium]

The Volta 2080 gpu that we will see in a few months will be about as fast as a gtx1080ti so why wait? buy a gtx1080ti now and enjoy Volta gtx 2080 performance now.

The new 2080ti usually wont release till 6 months after the new gtx2080, so some time next November is a good guess.

 

[Shmee]

I agree, that buying a 1080Ti would make more sense than a 1080. Or, if you can, wait for a Vega refresh or Volta. But expect Volta to be expensive, and it may be a while.

 

[amenx]

The Volta 2080 gpu that we will see in a few months will be about as fast as a gtx1080ti so why wait? buy a gtx1080ti now and enjoy Volta gtx 2080 performance now.

The new 2080ti usually wont release till 6 months after the new gtx2080, so some time next November is a good guess.

If we go by last 2 gens, the 970 = 780ti and 1070 = 980ti. The 980 and 1080 were quite a bit ahead of last gen Ti's. Would think a 2070 = 1080ti.

 

[Guru]

If we go by last 2 gens, the 970 = 780ti and 1070 = 980ti. The 980 and 1080 were quite a bit ahead of last gen Ti's. Would think a 2070 = 1080ti.

Not even close. With the current generation we went from 28nm to 14nm, thus the huge shift in performance and even then it took a GTX 1070 to match a GTX 980ti.

Ampero would be a refinement of the current gen, with GDDR6 probably replacing GDDR5x on the high end and GDDR5x replacing GDDR5 on the mid range and possibly low end.

So 2080 is likely going to be as fast as the 1080ti at a $550 price point. That is going to be the advantage of the 2080. Don't expect the 2080 to be faster at all or have much better power consumption either, the shift from 28nm to 16nm happened, now its just going to be improvements until we see 7nm and 10nm. My guess is the 2080 will have maybe 20W less power consumption and be as fast as the 1080ti at a lower price point.

But again why would Nvidia essentially sell a 1080ti at a lower price point? So chances are we are going to wait quite a while before next gen is introduced. My theory from the start was about mid Q2 2018 with the high end products like the 2070 and 2080 hitting first, followed in two weeks by the 2060 and 2060ti, then two to three weeks later the low end 2050 and 2050ti, then probably about one month later the Titan XXL hitting the market, sooner than ever since Nvidia wouldn't want to sell the 2080 at $550 or $600 whatever they decide for long, so a Titan XXL at $1000, followed by a 2080ti a month after than at about $750.

 

[Eric1987]

Not even close. With the current generation we went from 28nm to 14nm, thus the huge shift in performance and even then it took a GTX 1070 to match a GTX 980ti.

Ampero would be a refinement of the current gen, with GDDR6 probably replacing GDDR5x on the high end and GDDR5x replacing GDDR5 on the mid range and possibly low end.

So 2080 is likely going to be as fast as the 1080ti at a $550 price point. That is going to be the advantage of the 2080. Don't expect the 2080 to be faster at all or have much better power consumption either, the shift from 28nm to 16nm happened, now its just going to be improvements until we see 7nm and 10nm. My guess is the 2080 will have maybe 20W less power consumption and be as fast as the 1080ti at a lower price point.

But again why would Nvidia essentially sell a 1080ti at a lower price point? So chances are we are going to wait quite a while before next gen is introduced. My theory from the start was about mid Q2 2018 with the high end products like the 2070 and 2080 hitting first, followed in two weeks by the 2060 and 2060ti, then two to three weeks later the low end 2050 and 2050ti, then probably about one month later the Titan XXL hitting the market, sooner than ever since Nvidia wouldn't want to sell the 2080 at $550 or $600 whatever they decide for long, so a Titan XXL at $1000, followed by a 2080ti a month after than at about $750.

I agree completely. Sounds like exactly what I was thinking.

 

[tential]

But again why would Nvidia essentially sell a 1080ti at a lower price point? So chances are we are going to wait quite a while before next gen is introduced. My theory from the start was about mid Q2 2018 with the high end products like the 2070 and 2080 hitting first, followed in two weeks by the 2060 and 2060ti, then two to three weeks later the low end 2050 and 2050ti, then probably about one month later the Titan XXL hitting the market, sooner than ever since Nvidia wouldn't want to sell the 2080 at $550 or $600 whatever they decide for long, so a Titan XXL at $1000, followed by a 2080ti a month after than at about $750.

You're completely right.
It's just that, we aren't "waiting" any longer than we normally do. Nvidia is releasing the product when it's ready. When it's ready is mid Q2. We've seen this happen how many times now?

It's not that we're waiting any longer than normal, it's Nvidia's release schedule, that they've shown us multiple times in the past.

Mid Q2 isn't a revolutionary prediction, it's standard operating procedure.
Why we have a debate each time around acting like it's something new and hear all these outlandish theories I just don't get.
If I hear one more conversation about how the xx80ti isn't coming this generation and that it will only be the Titan card, I think I'll just have to lose all respect for those members at this point.

 

[phexac]

Interesting discussion here. Thanks guys. I particularly liked Guru's take and the laying out of the reasons for predicted performance improvement.

 

[Qwertilot]

Pretty confident he's wrong though. The predicted tiny energy efficiency jump is more or less known to be wrong - GV100 got a big efficiency jump over GP100.

Carry that over and there's enough leeway for them to push the 2080 comfortably ahead of the 1080ti and the massive revenues from people upgrading provide a very strong motive for them to do so. As noted by tentail they're really 'boring'. Predictable annual releases, predictable performance upticks etc etc.

 

[Harmaaviini]

Pretty confident he's wrong though. The predicted tiny energy efficiency jump is more or less known to be wrong - GV100 got a big efficiency jump over GP100.

Carry that over and there's enough leeway for them to push the 2080 comfortably ahead of the 1080ti and the massive revenues from people upgrading provide a very strong motive for them to do so. As noted by tentail they're really 'boring'. Predictable annual releases, predictable performance upticks etc etc.

Keep in mind that GP100 was not in geforce 1080 and GV100 won't be in 1180 (or is it going to be 2080?). As far as I know nvidia hasn't revealed the architecture of their gaming card lineup.

Sure, the announced theoretical peak flops look good in 300W powerenvelope but are they sustainable?
We don't know how big the efficiency jump is in reality.

They decreased the die size going from GM204 to GP104. What about GV104? They certainly have plenty of room to grow so theoretically they can easily match 1080ti if they choose to. But will they?

 

[Qwertilot]

Probably, because they've always done that in the past few years.

Its been working rather well for them, so it would be a big surprise if they switched strategy before they're really forced to by the silicon die shrinks hitting a wall.

 

[spinejam]

I would buy a used 1080 or 1080ti with a warranty at this point. Life is short and playing the waiting game sucks, ...so go ahead and buy something and enjoy it! Happy National Cookie Day!

 

[Bouowmx]

"Efficiency jump in reality":

Chart-topping LuxMark performance at < 300 W: https://www.reddit.com/r/nvidia/comments/7983zx/tesla_v100_luxmark_results/

 

[TestKing123]

Nvidia needs to sell a faster card on an established cycle to keep investors happy. They are not going to "not release a card because there's no competition or they do not want to sell a 1080Ti at a cheaper price point."

If Nvidia doesn't refresh, no one will have a reason to upgrade. And when that happens? Nvidia's cash flow stops. That is why we will see Volta earlier rather than later. The same reason why Nvidia kept a steady release schedule even though the 8800GTX dominated everything for a long time much like the 1080Ti does now.

Also, even without AMD, Nvidia still has competition in the Xbox 1 X. The console can do 4k gaming at a $500 price point. PC Gamers tired of overly expensive Nvidia cards on a slow refresh schedule will just jump to the Xbox 1 X, another scenario Nvidia doesn't want.

So we will see a refresh in the May-June timeframe and it will be faster than a 1080Ti (how much faster is the question), to keep both consumers and thus investors happy.

 

[Head1985]

We already know from GV100 how many sp will gaming volta have.Same SP but cut from all non gaming stuff:
Volta TITAN 5376SP GV102
GTX2080TI 5120SP GV102
GTX2080-3584SP GV104
GTX2070-2688SP GV104 (one GPC cut off like GTX1070)
GTX2060-1792SP GV106

Lets say the cards will run on 2Ghz reference.
GTX2080 with 3584SP will have 14.3TF+ new delta color compression+better IPC.GTX1080TI only runs at 1750Mhz so it will be like overclocked GTX1080TI+IPC.15-20% above 1080TI FE with no problem.With both cards overclocked it will be only 10%.(only IPC gain)
GTX2070 with only 2688SP will have very hard time even close to reference 1080TI.2688SP at 2Ghz is only 10.7TF(1080TI at 1750Mhz have 12.5TF).Should be 10% slower than reference 1080TI.Aftermarket cards will tie 1080TI.But overclocked 1080TI will be faster.
GTX2060 with 1792SP at 2Ghz it will be 7.1TF.Pretty much same as GTX1070(reference 1070 runs at 1860Mhz).With some IPC gain it should be faster than GTX1070.Propably at 1070TI level.

 

[Qwertilot]

Looks incredibly plausible I guess that the GDDR6 stuff should be worth a little performance bump as well?
(Not that they're sure to use that on everything.).

 

[tential]

They are not going to "not release a card because there's no competition or they do not want to sell a 1080Ti at a cheaper price point."

The amount of people who hold beliefs like this is astounding though.

 

[Harmaaviini]

Ofcourse they're going to realease a new card but it doesn't have to beat 1080ti to be successfull. In my opinion 20%-25% improvement over 1080 will have pretty much the same sales potential as a card that slightly beats 1080ti. If that slightly slower option has better margins it's a decision nvidia might make.