Recommended Interim GPU before Pascal

[MrTeal]

The OP is spending $750 on a 1440p 144Hz GSync monitor. He could get by on a 970 for now, but that isn't going to be the gaming experience you're looking for when you spend that kind of coin on a monitor.

If he picks up a $600 980Ti now, he'll probably lose $200 when he goes to resell it. If the buys a 970, he might lose $100. Considering we don't know when GP104 will launch and how much of an improvement of the 980Ti it will be, let alone when we might see GP100 in a consumer GPU, it's a question of whether spending an extra $100 it worth it to have almost twice the GPU while we wait possibly a year to get our hands on GP100.

 

[Headfoot]

No way he'd lose $100 on a used 970 sold at Pascal launch. $40-50 tops. If he's buying new, then maybe, but I wouldn't ever buy new for a card you know is a temporary card. Thats just a bad deal.

 

[MrTeal]

No way he'd lose $100 on a used 970 sold at Pascal launch. $40-50 tops. If he's buying new, then maybe, but I wouldn't ever buy new for a card you know is a temporary card. Thats just a bad deal.

Yeah, if he buys used the delta will obviously be lower. Not everyone likes buying used though. Either way, the argument is the same; we might be waiting a year to get the true replacement to the 980Ti. If you're looking to spend $2k on a new rig to play at high framerates, is it worth it to you to spend the extra $15 a month to go with the 980Ti over the 970?

 

[Headfoot]

maybe. For me at least, yes. Buying the best only to have the best get obsoleted shortly thereafter feels different to me than buying midrange, knowing you're waiting for the real high end. That feels like two upgrades! Not logical but hey

 

[poofyhairguy]

maybe. For me at least, yes. Buying the best only to have the best get obsoleted shortly thereafter feels different to me than buying midrange, knowing you're waiting for the real high end. That feels like two upgrades!

Nvidia approves this message.

 

[LTC8K6]

I figured he wouldn't be bothering with GSYNC with the interim card, so I had no problem recommending an AMD card.

 

[skipsneeky2]

maybe. For me at least, yes. Buying the best only to have the best get obsoleted shortly thereafter feels different to me than buying midrange, knowing you're waiting for the real high end. That feels like two upgrades! Not logical but hey

Seems the Ti cards usually finalize the product lineup lately,if you want the best of a generation just hold for the Ti.I am pretty much doing this,even if the wait is a year out.Ti makes any Titan look horrible and it offers more then a X80 card.

If the Ti isn't worth it,then hey by then the mid range cards should have finalized their pricing and you buy based on bang for buck.No trickery this time about with 980 cards becoming rebadged GTX1070 cards lol.

 

[SimianR]

I'm pretty sure that Ethereum mining has driven up the cost of used AMD cards. An R9 290, 970 or 390 would all be good temporary solutions. I do think the resale value all of those cards will take a fairly substantial hit when Polaris/Pascal launch because of the performance and perf/watt improvements that they will bring. Not sure a lot of people are going to want the older cards at that point.

p.s. Are you sure you want to lock yourself in on a G-Sync monitor? Seems like a big chunk of change and a lot of commitment to me. But to each his own.

 

[poofyhairguy]

I do think the resale value all of those cards will take a fairly substantial hit when Polaris/Pascal launch because of the performance and perf/watt improvements that they will bring.

I always laugh when people say this, like day one of launch Nvidia and AMD will roll out their entire lineup and make everything obsolete all at once.

We don't really know what AMD will do, but Nvidia isn't going to give you that kind of value unless they have to. The first card Nvidia will probably replace is currently the worst value in today's lineup- the GTX 980- and it will provide a LITTLE bit better value than the card it replaced (so not a 980 ti for $500). Then we get a new Titan, then maybe we get a 750 ti replacement.

The cards people want to buy- the 970 replacement and the 980 ti replacement- will come as late as possible. Probably 2017. Also coming late will be the 960 replacement.

There is just no reason for NVidia to release the 970 replacement or the 980 ti replacement this year. They know they will be stuck on this node for maybe even longer than the last one so it helps them to spread out the "improvement" over as long of time as possible.

The only way this doesn't happen is if AMD comes out the gate very aggressively and launches the 390 and 390x replacement first (and at the same time). Then Nvidia will be forced to give us the cards we want when we want them.

 

[SimianR]

I always laugh when people say this, like day one of launch Nvidia and AMD will roll out their entire lineup and make everything obsolete all at once.

We don't really know what AMD will do, but Nvidia isn't going to give you that kind of value unless they have to. The first card Nvidia will probably replace is currently the worst value in today's lineup- the GTX 980- and it will provide a LITTLE bit better value than the card it replaced (so not a 980 ti for $500). Then we get a new Titan, then maybe we get a 750 ti replacement.

The cards people want to buy- the 970 replacement and the 980 ti replacement- will come as late as possible. Probably 2017. Also coming late will be the 960 replacement.

There is just no reason for NVidia to release the 970 replacement or the 980 ti replacement this year. They know they will be stuck on this node for maybe even longer than the last one so it helps them to spread out the "improvement" over as long of time as possible.

The only way this doesn't happen is if AMD comes out the gate very aggressively and launches the 390 and 390x replacement first (and at the same time). Then Nvidia will be forced to give us the cards we want when we want them.

You're right that it's all up in the air right now, but your last paragraph says it all. NVIDIA doesn't have to be aggressive and give you that value, but AMD is a big question mark at the moment. They definitely have to be more aggressive than NVIDIA, and it's very possible that they will try to reclaim as much market share as possible. If NVIDIA does replace the 980 first as you're saying, then OP will be waiting even longer since he wants BIG pascal (980 Ti replacement?). I can't imagine a 970 being particularly appealing or selling for all that much in 2017 if AMD and NVIDIA have 14/16nm parts available. Who knows though. We've never been in a situation where we've been stuck on the same node for 5 years.

 

[poofyhairguy]

You're right that it's all up in the air right now, but your last paragraph says it all. NVIDIA doesn't have to be aggressive and give you that value, but AMD is a big question mark at the moment. They definitely have to be more aggressive than NVIDIA, and it's very possible that they will try to reclaim as much market share as possible.

Sure. I guess it depends on what AMD sees as a priority- having the top card or hitting the sweet spot?

If it is the former, then we will see a Fiji replacement first. I think there is actually a good chance this might happen, given how they really wanted Fiji to take back the performance crown from Nvidia's Titan (it just fell short). The problem with a Fiji replacement first is Fiji is their newest chip, heck they haven't even rolled out its full lineup yet. One good thing about a Fiji replacement is current Furys can be marked down to where the 390x currently is, killing two birds with one stone. AMD doesn't have the resources to make a new card for every tier, something will be a rebrand.

The sweet spot move is obviously replacing the 390 and 390x. This would match up well to a 980 replacement, and give the 970 (which is currently the biggest hit on the market) a reason to sweat, but it would hang Fiji out to dry (the gap between the 390x and the regular Fury is already too small). I really don't think AMD is going to do this, which means Nvidia gets to sell 970s into 2017. The 390 is doing well right now and wasn't launched that long ago. You have to figure the midrange will eventually have 4GB HBM, but that will look bad next to the 8GB 390 unless it blows the 390 away.

The dark horse is AMD releases the Tonga replacement first. I think this is the best move for them- the 960 is a VERY weak competitor and Tonga is getting a little old at this point (even though we just barely saw it in full form). I think the gap between the 380x and the 390 gives AMD more wiggle room than anywhere in the market. We could get a Tonga replacement that is an upgrade that doesn't knock off the 390 and as a bonus it might be toe-to-toe in Directx 12 to the 970.

We will know soon enough. One thing is for sure, anyone assuming the 980 ti replacement out the gate will be disappointed. No matter what AMD does Nvidia is going to want to milk another Titan before it destroys that value with the 980 ti replacement. Unless AMD is super aggressive (more than a company in their condition can be) OP is waiting until 2017 to buy a new card.

 

[SimianR]

All good points, either way, I'm excited to see what Polaris and Pascal have to offer.

For the OP: I just stumbled across this deal, thought it might be worth a look:

http://www.newegg.com/Product/Produ...4150723&cm_re=290x_8gb-_-14-150-723-_-Product

R9 290X 8GB card for $310 after rebate. With a minor overclock you have yourself a 390X for $310. Might be a decent option until next gen cards are available.

 

[Mahigan]

That R9 290x 8GB is a great buy.

 

[SteveGrabowski]

EVGA BStock currently has a GTX 970 in stock for $250.

http://www.evga.com/Products/Product.aspx?pn=04G-P4-2977-RX

Any lower than GTX 970 or R9 290 isn't going to run Witcher 3 too well.

 

[SteveGrabowski]

I always laugh when people say this, like day one of launch Nvidia and AMD will roll out their entire lineup and make everything obsolete all at once.

We don't really know what AMD will do, but Nvidia isn't going to give you that kind of value unless they have to. The first card Nvidia will probably replace is currently the worst value in today's lineup- the GTX 980- and it will provide a LITTLE bit better value than the card it replaced (so not a 980 ti for $500). Then we get a new Titan, then maybe we get a 750 ti replacement.

The cards people want to buy- the 970 replacement and the 980 ti replacement- will come as late as possible. Probably 2017. Also coming late will be the 960 replacement.

There is just no reason for NVidia to release the 970 replacement or the 980 ti replacement this year. They know they will be stuck on this node for maybe even longer than the last one so it helps them to spread out the "improvement" over as long of time as possible.

The only way this doesn't happen is if AMD comes out the gate very aggressively and launches the 390 and 390x replacement first (and at the same time). Then Nvidia will be forced to give us the cards we want when we want them.

I think you're being really pessimistic. When Nvidia went from 40 nm to 28 nm their GTX 680 was the upper midrange Kepler chip and still destroyed the balls to the wall high end GTX 580. I think there's good reason for them to make big improvements if the technology is there. Look at how many people bought GTX 970s when they brought near 780 Ti level performance to the $330 price range. It made them an enormous amount of money, everyone bought GTX 970s. If I'm Nvidia I'm releasing 70 and 80 series ASAP and then a year to year and a half later I'm releasing the 80 Ti series to get all those 70 and 80 series buyers to buy again just like they did with 680 and 780 Ti. Why would they put out crap that won't outshine the 900 series when they won't be selling 900 series any more?

 

[poofyhairguy]

I think you're being really pessimistic. When Nvidia went from 40 nm to 28 nm their GTX 680 was the upper midrange Kepler chip and still destroyed the balls to the wall high end GTX 580.

I think the market has slowed since then. I think Nvidia COULD produce a 50% leap year one in value, but it knows it will be stuck on the node for a while so the next two generations (on the same node) will only be like 10% improvements. Nvidia knows it would be better to cut that 50% over a few generations to not get stuck in the rut that Intel is in. So we will get a 15-20% leap next gen. And the next gen. And so on.

The question is if AMD will let them. The 980 ti kicks ass at $650 because of the Fury X. Otherwise that is a $800 card I bet.

Maybe I am being pessimistic.

 

[tweakboy]

960 GTX. gl

 

[swilli89]

I think the market has slowed since then. I think Nvidia COULD produce a 50% leap year one in value, but it knows it will be stuck on the node for a while so the next two generations (on the same node) will only be like 10% improvements. Nvidia knows it would be better to cut that 50% over a few generations to not get stuck in the rut that Intel is in. So we will get a 15-20% leap next gen. And the next gen. And so on.

The question is if AMD will let them. The 980 ti kicks ass at $650 because of the Fury X. Otherwise that is a $800 card I bet.

Maybe I am being pessimistic.

You're definitely right. If AMD didn't release Fury X there probably wouldn't even be a 980 ti. I can see an nV 980 replacement that edges out the 980 ti sometime later this year, but that card will be at "high end" pricing. At ;east $600 if AMD doesn't have anything to compete with.

 

[Dribble]

I figured he wouldn't be bothering with GSYNC with the interim card, so I had no problem recommending an AMD card.

Isn't that the whole point of gsync/freesync - to make a slower card that can't keep the high fps feel smoother?

 

[Madpacket]

Isn't that the whole point of gsync/freesync - to make a slower card that can't keep the high fps feel smoother?

Yes. A 970 @ 1440P with Gsync enabled should have no issues with smooth framerate delivery even in demanding games like the Witcher 3. Just keep within the memory footprint and you'll be fine.

 

[Arachnotronic]

You're definitely right. If AMD didn't release Fury X there probably wouldn't even be a 980 ti. I can see an nV 980 replacement that edges out the 980 ti sometime later this year, but that card will be at "high end" pricing. At ;east $600 if AMD doesn't have anything to compete with.

980 Ti came out first.

 

[MrTeal]

I think you're being really pessimistic. When Nvidia went from 40 nm to 28 nm their GTX 680 was the upper midrange Kepler chip and still destroyed the balls to the wall high end GTX 580. I think there's good reason for them to make big improvements if the technology is there. Look at how many people bought GTX 970s when they brought near 780 Ti level performance to the $330 price range. It made them an enormous amount of money, everyone bought GTX 970s. If I'm Nvidia I'm releasing 70 and 80 series ASAP and then a year to year and a half later I'm releasing the 80 Ti series to get all those 70 and 80 series buyers to buy again just like they did with 680 and 780 Ti. Why would they put out crap that won't outshine the 900 series when they won't be selling 900 series any more?

At launch the 680 was ~25% faster than the 580 at 16x10 and 19x12. That's a good bit faster, but it also cost $100 more than the 580 at that point. I don't think that's really destroying it even in performance, definitely not in perf/$. A few years later the numbers looked way different of course, but at the time we didn't see massive increases in perf/$.

I'd agree with you on timelines in the consumer sector though. GP104 will come first, with GP100 trailing after. We're already seeing $600 980Tis, by the time GP104 launches I would imagine we'll see GP104 launch at least at the same price as the outgoing 980Ti. What the x70 launches at will be more interesting. $329 vs $549 for Maxwell 2 was amazing, but the 670 was only $100 cheaper than the $500 680.

 

[SteveGrabowski]

At launch the 680 was ~25% faster than the 580 at 16x10 and 19x12. That's a good bit faster, but it also cost $100 more than the 580 at that point. I don't think that's really destroying it even in performance, definitely not in perf/$. A few years later the numbers looked way different of course, but at the time we didn't see massive increases in perf/$.

I didn't realize it was only ~25% difference at launch, but Techpowerup's benchmark numbers from the 680 launch back that up. Anandtech's bench seems to have it more around 40%, but perhaps that's later on with newer games.

 

[SteveGrabowski]

I'd agree with you on timelines in the consumer sector though. GP104 will come first, with GP100 trailing after. We're already seeing $600 980Tis, by the time GP104 launches I would imagine we'll see GP104 launch at least at the same price as the outgoing 980Ti. What the x70 launches at will be more interesting. $329 vs $549 for Maxwell 2 was amazing, but the 670 was only $100 cheaper than the $500 680.

Considering the 970 was very competitive with the 780 Ti without a node shrink, it wouldn't surprise me to see the Pascal 70 series card beat the 980 Ti. I imagine Nvidia would want to sell it around $370-$400. If they can create that same kind of buzz they had with the 970 by offering previous generation flagship level performance I think they could maximize their profits around $400. It would crush a 970 for $70 more, and the 970 is still selling like crazy at launch price a year and a half after introduction.

 

[SteveGrabowski]

I think the market has slowed since then. I think Nvidia COULD produce a 50% leap year one in value, but it knows it will be stuck on the node for a while so the next two generations (on the same node) will only be like 10% improvements. Nvidia knows it would be better to cut that 50% over a few generations to not get stuck in the rut that Intel is in. So we will get a 15-20% leap next gen. And the next gen. And so on.

The question is if AMD will let them. The 980 ti kicks ass at $650 because of the Fury X. Otherwise that is a $800 card I bet.

Maybe I am being pessimistic.

If I'm Nvidia I'm going for the jugular on AMD. AMD has had two really bad years thanks in no small part to the 970. They were having to practically give away the 290 at $230. Another couple of years like that and AMD could be stuck at the low end and then Nvidia would be able to charge whatever they want.