Are you buying a GTX 1070/1080?

[swilli89]

Eh that was what, a decade ago? Given inflation, moving up to 379 isnt outrageous. That said, if you arent in a rush. Definitely wait and see what AMD offers.

You missed the point then. In that way, the 7900gtx would have been more too with inflation adjustments. The point was that nVidia used to offer substantially more performance from their midrange over the previous gen's high end.

 

[Genx87]

You missed the point then. In that way, the 7900gtx would have been more too with inflation adjustments. The point was that nVidia used to offer substantially more performance from their midrange over the previous gen's high end.

Ok? Did everybody expect 100% improvements until the end of time with each new generation of graphics processors? x86 processors have seen incremental changes <10% for nearly 2 decades. GPUs were bound to have this happen to them as well.

 

[tential]

i had to decide between 660 and 7870. both were almost the same at the time in every metric. now look where the 660 is and the 7870...:thumbsdown:

I can understand your frustration with Nvidia products holding up.
Look at the past:

GTX 660 is on par with the HD7870
750TI is on par with the HD7850

Now lets fastforward to present day.

The R7 370 is the HD7850... rebranded AGAIN (Like jesus just die pitcairn).

So the HD 7870 is even FASTER now. The 8% clockspeed boost in the R7 370 doesn't explain the performance difference between the HD7850 before and now vs the 750ti. And that's on a MAXWELL architecture. With Kepler?

It gets even WORSE.

Look at the previous graph again at the 760. It was 50% above the HD7850. Now it's 10% above the HD7850 in the form of the R7 370 with the tiny clockspeed boost that got of 8%.

Look, I can sit here and tell you the GTX 660 was a great card, the DATA does NOT HOLD. Kepler fell apart against the HD7000 series.

One final look, look at the 780Ti leading the pack, now look where it is....

But hey, this is why no data was provided to backup the claim that the GTX 660 is a great card. It's just said and hoped that people will believe every Nvidia product is amazing.

Nvidia is GREAT at release, and holds up poorly. I expect the 1070 to blow us away at launch just like the previous Nvidia launches. But unless Nvidia changes something DRASTICALLY, people will remember the last time they were burned with Nvidia performance drops, and start to wisen up. Some people anyway.

The MAJOR problem with the 1070/1080 is sit's a refined Maxwell, and if Maxwell isn't holding up, how is refining an architecture that's only getting weaker going to help while AMD is refining an already strong architecture at current games. That's what leads me to be hesitant about a 1070 for ANYONE who holds their GPU more than one generation, or doesn't play Day 1 games, or isn't competitive gaming.

Edit: Kepler dropped off, Maxwell 1 dropped off, Maxwell 2 dropped off slowly, am I supposed to believe the third refined Maxwell in the form of Pascal will hold?

 

[Genx87]

You sold me. If I am going to hold onto a card for 4-5 years and not care about performance when purchased. Then AMD may be the way to go.

 

[swilli89]

You sold me. If I am going to hold onto a card for 4-5 years and not care about performance when purchased. Then AMD may be the way to go.

Great hyperbole. Real world would be:

You sold me. If I am going to hold onto a card for 2-3 years and want equal performance at launch. Then AMD may be the way to go.

 

[railven]

You missed the point then. In that way, the 7900gtx would have been more too with inflation adjustments. The point was that nVidia used to offer substantially more performance from their midrange over the previous gen's high end.

But the same goes to AMD.

4890 to 5870, about 30% performance gain, price went from $250 to $380
6970 to 7970, about another 30% performance gain, price went from $370 to $550
7970 Ghz to 290X, about 40% performance increase, price went from $500 to $550
290X to Fury X, about 30% performance increase, price went from $550 to $650

But I guess as long as the die is bigger, right?

And I know why AMD did it, because them selling a 5870 for $380 was basically charity.

AMD has almost tripled the cost of their top card, and it's still getting beat by NV's top card for that particular generation regardless of the die sizes. Does it make someone feel better if they're paying $550 for a bigger die of it's predecessor if it's still slower than their rivals $500 card?

 

[Keysplayr]

I can understand your frustration with Nvidia products holding up.
Look at the past:

GTX 660 is on par with the HD7870
750TI is on par with the HD7850

Now lets fastforward to present day.

The R7 370 is the HD7850... rebranded AGAIN (Like jesus just die pitcairn).

So the HD 7870 is even FASTER now. The 8% clockspeed boost in the R7 370 doesn't explain the performance difference between the HD7850 before and now vs the 750ti. And that's on a MAXWELL architecture. With Kepler?

It gets even WORSE.

Look at the previous graph again at the 760. It was 50% above the HD7850. Now it's 10% above the HD7850 in the form of the R7 370 with the tiny clockspeed boost that got of 8%.

Look, I can sit here and tell you the GTX 660 was a great card, the DATA does NOT HOLD. Kepler fell apart against the HD7000 series.

One final look, look at the 780Ti leading the pack, now look where it is....

But hey, this is why no data was provided to backup the claim that the GTX 660 is a great card. It's just said and hoped that people will believe every Nvidia product is amazing.

Nvidia is GREAT at release, and holds up poorly. I expect the 1070 to blow us away at launch just like the previous Nvidia launches. But unless Nvidia changes something DRASTICALLY, people will remember the last time they were burned with Nvidia performance drops, and start to wisen up. Some people anyway.

The MAJOR problem with the 1070/108 is sit's a refined Maxwell, and if Maxwell isn't holding up, how is refining an architecture that's only getting weaker going to help while AMD is refining an already strong architecture at current games. That's what leads me to be hesitant about a 1070 for ANYONE who holds their GPU more than one generation, or doesn't play Day 1 games, or isn't competitive gaming.

Sorry. I'd comment, but I'd get warned for posting off topic. You and caswow continue on unfettered though. :thumbsup:

 

[exar333]

But the same goes to AMD.

4890 to 5870, about 30% performance gain, price went from $250 to $380
6970 to 7970, about another 30% performance gain, price went from $370 to $550
7970 Ghz to 290X, about 40% performance increase, price went from $500 to $550
290X to Fury X, about 30% performance increase, price went from $550 to $650

But I guess as long as the die is bigger, right?

And I know why AMD did it, because them selling a 5870 for $380 was basically charity.

I know, right!? I grabbed a 5870 because it was what, $100 more vs. 5850? Both were great, but they were both a steal for how long they lasted...

 

[railven]

I know, right!? I grabbed a 5870 because it was what, $100 more vs. 5850? Both were great, but they were both a steal for how long they lasted...

I counted my stars I got a 5870 before the retailers started price gouging. I remember soon after they were hitting almost $500. Newegg was particularly bad.

 

[tential]

Great hyperbole. Real world would be:

I'm not going to go with equal performance, it's definitely lessers than Nvidia in my opinion. Nvidia is the launch king, and I feel credit should be given where credit is due. NO ONE can beat Nvidia right now in launching a product. They just are amazing at what they do in that regard. If I'm Genx87, I can't buy an AMD product that's behind in performance. I refuse to. I wait until AMD is the better product ALL AROUND, before I get it.

If I buy an AMD product day 1, I'm telling them that the product is good enough. No, I wait until I feel the product is great all around, then I get it, and usually at a discount, and that's AMD's fault for not having good drivers, cooling solution, no coil whine, or whatever foolish mistake they want to make.

The one thing we know for sure is the 1070 will be good at launch. I expect P10 to be amazing for the price, but when that comes to fruition? No one knows...

 

[tential]

But the same goes to AMD.

4890 to 5870, about 30% performance gain, price went from $250 to $380
6970 to 7970, about another 30% performance gain, price went from $370 to $550
7970 Ghz to 290X, about 40% performance increase, price went from $500 to $550
290X to Fury X, about 30% performance increase, price went from $550 to $650

But I guess as long as the die is bigger, right?

And I know why AMD did it, because them selling a 5870 for $380 was basically charity.

AMD has almost tripled the cost of their top card, and it's still getting beat by NV's top card for that particular generation regardless of the die sizes. Does it make someone feel better if they're paying $550 for a bigger die of it's predecessor if it's still slower than their rivals $500 card?

No, I don't care about die size, whether it has HBM, or anything. It's all about performance. Which is why even though I think Fury X is a great card, it's slower than a single 980Ti. I don't care what else it has going for it at ALL. It's slower than it's competing card, so I can't touch it.

 

[railven]

No, I don't care about die size, whether it has HBM, or anything. It's all about performance. Which is why even though I think Fury X is a great card, it's slower than a single 980Ti. I don't care what else it has going for it at ALL. It's slower than it's competing card, so I can't touch it.

Well I wouldn't go that far, because I actually enjoyed AMD playing price / perf wars with NV.

If the CLC is the main cause for Fury X having to sell equal to 980 Ti, that is something AMD might have had to eat a loss on. Or, let their AIBs actually tinker and create something probably just as good for less.

If AMD is going to be slower, they'll have to be cheaper, same goes for NV, and I think NV is rather aware of this. They price cut the 780 so quick when 290X launched.

AMD just can't seem to move fast enough. And when they do, it always feels like it backfires.

EDIT:

Great hyperbole. Real world would be:

What does the edit post even mean? Who wants equal performance from launch 2-3 years later? You want improved performance. You want driver maturity. You want the card you bought to be 1.x faster 2-3 years after you bought it.

That comment doesn't help AMD.

 

[tential]

Well I wouldn't go that far, because I actually enjoyed AMD playing price / perf wars with NV.

If the CLC is the main cause for Fury X having to sell equal to 980 Ti, that is something AMD might have had to eat a loss on. Or, let their AIBs actually tinker and create something probably just as good for less.

If AMD is going to be slower, they'll have to be cheaper, same goes for NV, and I think NV is rather aware of this. They price cut the 780 so quick when 290X launched.

AMD just can't seem to move fast enough. And when they do, it always feels like it backfires.

Nah, I ain't letting AMD pull the "We have a CLC so like pay a ton more!!!!"
Sapphire makes coolers too.
AMD also didn't give their AIBs an option to tinker like you said.

I mean Fiji release as a mess. Nano was way overpriced and had coil whine. Fury X with pump noise.

The funny thing is, Nano is slightly slower than the 980Ti in many of the new DX12 games. If you drop Nano price to $400.... all of a sudden it's a decent competitor to the 1070. Especially at 4K

$425-50 Fury X $400 Nano, and sales on those products might actually make high end fiji competitive with the 1070 and leave the 1080 sit alone at $700 by itself.

I mean is ANYONE excited by this?

390x for $200-250?
GTX 980Ti pricedrop as it drops performance in DX12 games relative to Fury X?

No massive performance jump at all?

I don't care if you're an AMD or Nvidia fanboy, this is NOT what we all got excited about when we heard new nodes were coming as well as architectures.

Edit: Of course newegg shows AMD pricing the nano up and not down.... Same with the Fury X.

AMD should be dropping the price on the high end now. They may convince a prospective 1070 buyer to get a Fury X instead if the price is right or nano is right. Raising the prices doesn't help at all...

 

[exar333]

No, I don't care about die size, whether it has HBM, or anything. It's all about performance. Which is why even though I think Fury X is a great card, it's slower than a single 980Ti. I don't care what else it has going for it at ALL. It's slower than it's competing card, so I can't touch it.

Thank you.

Big die, little die. 'big pascal' or 'little pascal'. Whatever meets my needs for performance and price matter.

 

[Face2Face]

But unless Nvidia changes something DRASTICALLY, people will remember the last time they were burned with Nvidia performance drops, and start to wisen up. Some people anyway.

They did. They designed Pascal a lot like GCN. Hopefully, that means we won't see the performance drop off like we did with Kepler, and soon to be Maxwell.

Also, does anyone know how often are the TechPowerUP performance summary charts are updated?

 

[Josh123]

I'm probably going to pick up a used 980 Ti to tide me over in hopes of a 1080 Ti.

 

[moonbogg]

I'm not being a smartass here, but I fully expect Maxwell performance to sink like a lead titanic.

 

[railven]

I'm not being a smartass here, but I fully expect Maxwell performance to sink like a lead titanic.

That's why I'm bailing that ship. And instead of handing over the 980 Ti to the GF, I'll get her an updated 1070.

Once more DX12 titles role out, even if they dont use crippling levels of Async Compute, Pascal is going to weather the storm better.

 

[Arachnotronic]

I'm not being a smartass here, but I fully expect Maxwell performance to sink like a lead titanic.

Do you think NVIDIA will actively work to degrade performance in future drivers?

 

[railven]

Do you think NVIDIA will actively work to degrade performance in future drivers?

Degrade, no - since it hasn't been proven that NV did that to Kepler.

However, with the API shift, I just expect Pascal to handle DX12/VR better. Thus my bailing of the Maxwell ship.

 

[Arachnotronic]

Degrade, no - since it hasn't been proven that NV did that to Kepler.

However, with the API shift, I just expect Pascal to handle DX12/VR better. Thus my bailing of the Maxwell ship.

I don't doubt that Pascal will be much better for DX12 games than Maxwell, but it's a newer, more robust architecture. Why would we expect anything different?

 

[AntonioHG]

Vega/1080Ti for me. I want something to completely crush my R9 290/x Crossfire setup in a single card.

 

[railven]

I don't doubt that Pascal will be much better for DX12 games than Maxwell, but it's a newer, more robust architecture. Why would we expect anything different?

More so it seems with at least the last three generations, when NV has a strength they play it. With Gameworks we saw that increase more. I would expect the same to continue. If Pascal can do things more efficiently I see NV layering more of it on their Gameworks sponsored titles.

That essentially what crippled Kepler, not that NV was regressing their support for it.

 

[xthetenth]

I don't doubt that Pascal will be much better for DX12 games than Maxwell, but it's a newer, more robust architecture. Why would we expect anything different?

Agreed. It's not NV's architectures that are the exception. You don't need malice to explain the trend. It's that AMD did a fantastic job of building a future for their GPUs. You can make a case for gameworks and so on pushing it harder, but the trend is there for all games.

 

[Grooveriding]

I'm not being a smartass here, but I fully expect Maxwell performance to sink like a lead titanic.

Yeah Maxwell is going to be getting the Kepler treatment in short order. Launch day 1080 reviews will show a 1080 about 20-25% faster on average against a 980ti. Six months from now it will be 50% and a 980ti will be getting 30fps in Battlefield 1 ()

Nvidia cards have the worst longevity. Maxwell will be good for a little while yet. We'll see 1500 core overclocked 980ti scoring about as well as a stock 1080 for a while. Once enough new games release though, it's going to go down like the Titanic, just like Kepler did.