[vrzone] GTX670Ti in March

[blastingcap]

I paid $260 for this gtx570 and the way things look now it will be at least 4 years before I can get anywhere near doubling that performance for the same amount of money. heck even at $350 that could be true. we seem to be waiting longer and longer between true next gen cards yet performance is not going up like it used to when the cycles were shorter.

Consolification -> less demand for discrete performance GPUs -> lower profits to reinvest in the performance GPU segment -> less price/perf improvement with each new generation of GPUs.

Also, shrinks look like they aren't going to be as cheap as before to do, now that we're already at the 20nm level.

And cards have been getting a lot more and a lot faster VRAM. Seems like only yesterday that 512MB VRAM was more than enough. Now people look at you funny if you suggest that 1GB VRAM is enough for 1080p+.

 

[Arachnotronic]

The 7970 is a decent card, but it simply isn't that much faster than a GTX 580 (yes, I've owned both...). Driver issues/incompatibility with older games made me take back the 7970, which was a bummer since, well...it was neat having the latest and greatest.

Kepler shouldn't have that hard of a time laying the smack-down on 7970.

 

[3DVagabond]

Q1 is not over yet. 2011 was just not doable with TMSC. Both IHVs launched in Q1 2012. AMD rushed out a few cards with no regard (again) to the reviewers, giving them very little time to test it.

Poor reviewers! I'm sure AMD were holding their children hostage and holding guns to their heads forcing them to do those reviews.

They could not get out many cards with decent voltages/clocks because of process variations, this is why the voltage on Tahiti is relatively high or the clocks low, however you want to look at it.

Voltages doesn't mean squat. It's watts that matter and Tahiti's efficiency and temps are fine.

So no, until Q1 is over, Kepler is not late (by what standards anyway? rushed launch of the competition on an immature process?).

You would have to define late. They are late to the competition for example. They are late by their original roadmaps. They are not yet late to their Q1 revision back in August. We'll have to wait a few more weeks for that deadline. Of course if they release one mobile or one mid/lowend SKU there will be those who proclaim victory while the competition is likely to have rolled out 3 desktop chips and 6 to 8 different cards covering most of the performance segments. At that point all AMD has to do is tweak prices, if they so desire, to control the market.

Couldn't agree more. Tahiti is a performance GPU, a performance/highend hybrid at best. As the successor of the 6970 it should roughly sell for the same price. The only two things permitting these prices right now are lack of competition and great OC potential. If you look at stock performance though, it's a bit underwhelming.

Consider that no company has to put out a significantly faster product than the competition. If they would do that, they would rob themselves of profit (less good chips due to high clock targets) and set the bar too high for their refresh. I could very well see Nvidia bringing a GK110 that holds a bit back because more performance is not needed. Why beat the competition by 40% when you can do it at 20% and higher profit? (just an example!). But this also depends on what AMD does when Kepler arrives. Will they bring a 7970 "Ultra" to compete or will they just lower prices?

Agreed, at stock clocks the performance is only about 40% more than their last gen. The part of your post I've bolded explains why. You even go as far as to say you'll understand it if that's what nVidia does. Understand it when AMD does it as well. There is no competition. AMD has released their product months before nVidia. They are a corporation, not a charity, and they have a legal responsibility to their shareholders.

 

[boxleitnerb]

Poor reviewers! I'm sure AMD were holding their children hostage and holding guns to their heads forcing them to do those reviews.

Ask some reviewers. AMD is notorious for sending out samples very late, the press is really pissed about this. And of course the sites have to test them - that's how they generate clicks and revenue. Do you actually know how much time it takes to properly test a graphics card? Image quality, features, possible hiccups, videos, screenshots, power consumption, noise etc. etc. To do it in a couple of days is almost impossible. Look at computerbase, they do some of the best reviews in the web, very thorough. That takes time.
Also interesting: Several reviewers received cards with thermal paste that was not properly applied. AMD even wrote them and told them so during their test...I wouldn't call that a well prepared launch, but a rushed one so they can claim they released in 2011. Yeah, right.

Voltages doesn't mean squat. It's watts that matter and Tahiti's efficiency and temps are fine.

You realize that wattage largely depends on operating voltage, right? The point is, they could have clocked Tahiti higher on a more mature process, bringing performance to a point that actually justifies the $550 price tag. Or lower voltages and wattage, reducing cost (and price) by using a simpler PCB, less fancy circuitry etc.

You would have to define late. They are late to the competition for example. They are late by their original roadmaps. They are not yet late to their Q1 revision back in August. We'll have to wait a few more weeks for that deadline. Of course if they release one mobile or one mid/lowend SKU there will be those who proclaim victory while the competition is likely to have rolled out 3 desktop chips and 6 to 8 different cards covering most of the performance segments. At that point all AMD has to do is tweak prices, if they so desire, to control the market.

As I said, I believe the competition rushed out Tahiti prematurely. I wouldn't take that as a reference point. There aren't even WHQL drivers for it, close to 6 weeks after launch. That's gotta tell you something. If Nvidia launches a refined product with actually good price/perf, the delay doesn't matter. It's not like AMD is selling their cards by the millions...a couple of ten thousand cards mean nothing in the big game.

Agreed, at stock clocks the performance is only about 40% more than their last gen. The part of your post I've bolded explains why. You even go as far as to say you'll understand it if that's what nVidia does. Understand it when AMD does it as well. There is no competition. AMD has released their product months before nVidia. They are a corporation, not a charity, and they have a legal responsibility to their shareholders.

True that. Problem for both is/will be, however, how the customers receive this behavior. I would definitely not spend 550 bucks on a card that won't be the fastest until the refresh. I pay this kind of money for really high end performance. And if Nvidia releases a card that is on Tahiti XT level for 400-500 bucks, they can keep it. However, if they top the 580 by 60+% for 499-549, that is more in line with my expectations. Months...well 2 months by the latest looks of it. Nothing to worry about.

 

[Arzachel]

As I said, I believe the competition rushed out Tahiti prematurely. I wouldn't take that as a reference point. There aren't even WHQL drivers for it, close to 6 weeks after launch. That's gotta tell you something.

Correct me if I'm wrong, but Nvidia hasn't released WHQL drivers since the end of October themselves and that's on a pretty mature arch.

True that. Problem for both is/will be, however, how the customers receive this behavior. I would definitely not spend 550 bucks on a card that won't be the fastest until the refresh. I pay this kind of money for really high end performance. And if Nvidia releases a card that is on Tahiti XT level for 400-500 bucks, they can keep it. However, if they top the 580 by 60+% for 499-549, that is more in line with my expectations. Months...well 2 months by the latest looks of it. Nothing to worry about.

More like 6 months at best if you expect such an improvement.

 

[3DVagabond]

Ask some reviewers. AMD is notorious for sending out samples very late, the press is really pissed about this. And of course the sites have to test them - that's how they generate clicks and revenue. Do you actually know how much time it takes to properly test a graphics card? Image quality, features, possible hiccups, videos, screenshots, power consumption, noise etc. etc. To do it in a couple of days is almost impossible. Look at computerbase, they do some of the best reviews in the web, very thorough. That takes time.

While I'm not saying it isn't a lot of work to do reviews, it is what it is. It's a job and these guys choose to do it. It's a job that a lot of people wouldn't mind having. I don't know what you do for a living, but if it's without pressure and deadlines, you are either very lucky or it's a mundane job. Pressure is a part of the job for a journalist. Big corporations can't afford to work at a leisurely pace and delay things for days or weeks just so a review journalist doesn't have to work so hard. These corporate types are putting in long days under the gun. You aren't going to find a lot of sympathy.

You realize that wattage largely depends on operating voltage, right? The point is, they could have clocked Tahiti higher on a more mature process, bringing performance to a point that actually justifies the $550 price tag. Or lower voltages and wattage, reducing cost (and price) by using a simpler PCB, less fancy circuitry etc.

No wattage does not largely depend on operating voltage. It's dependent on what the resistance of the load is. I'm not an electrical engineer, but even I know Ohm's law. AMD's GPU's obviously have a higher electrical resistance than nVidia's do. They consistently use less power than nVidia's.

As I said, I believe the competition rushed out Tahiti prematurely. I wouldn't take that as a reference point. There aren't even WHQL drivers for it, close to 6 weeks after launch. That's gotta tell you something. If Nvidia launches a refined product with actually good price/perf, the delay doesn't matter. It's not like AMD is selling their cards by the millions...a couple of ten thousand cards mean nothing in the big game.

There is nothing in that paragraph that either of us has a clue about. We don't know if it was rushed or not. I was under the impression that Tahiti was ready months ago and was just waiting for TSMC to be able to manufacture it for AMD. WHQL is just Microsoft giving the drivers their blessing. I sure wouldn't not sell a product because I was waiting for WHQL certification. Finally, we don't know how many 7000 series cards AMD is selling. No idea what so ever.

True that. Problem for both is/will be, however, how the customers receive this behavior. I would definitely not spend 550 bucks on a card that won't be the fastest until the refresh. I pay this kind of money for really high end performance. And if Nvidia releases a card that is on Tahiti XT level for 400-500 bucks, they can keep it. However, if they top the 580 by 60+% for 499-549, that is more in line with my expectations. Months...well 2 months by the latest looks of it. Nothing to worry about.

Of course your buying decisions are your own. They don't influence the market. Nor, are they necessarily a reflection of it. It's just "you". If nVidia releases a card first or second week of March (2 Mos. later) that's 60% faster than a 580, stock for stock, you will be right, but I'll be really surprised. All rumors now don't have that happening until 3rd/4th Q. You do realize though you can O/C a 7970 right now and have it be that much faster than a 580?

 

[boxleitnerb]

While I'm not saying it isn't a lot of work to do reviews, it is what it is. It's a job and these guys choose to do it. It's a job that a lot of people wouldn't mind having. I don't know what you do for a living, but if it's without pressure and deadlines, you are either very lucky or it's a mundane job. Pressure is a part of the job for a journalist. Big corporations can't afford to work at a leisurely pace and delay things for days or weeks just so a review journalist doesn't have to work so hard. These corporate types are putting in long days under the gun. You aren't going to find a lot of sympathy.

Other companies do things differently (Intel, Nvidia) from what the reviewers here in Germany say. Of course there is pressure involved being a journalist, but as a company delivering samples you should avoid slapping reviewers in the face.

No wattage does not largely depend on operating voltage. It's dependent on what the resistance of the load is. I'm not an electrical engineer, but even I know Ohm's law. AMD's GPU's obviously have a higher electrical resistance than nVidia's do. They consistently use less power than nVidia's.

Sorry mate, you are dead wrong here. There are plenty undervolting and overvolting tests out there, look them up. If you increase voltage, power consumption increases alot (conversely, if you decrease voltage, it drops), this is common knowledge.
I found an example for you here:
http://www.expreview.com/img/review/HD7970voltage/powerrating.png

There is nothing in that paragraph that either of us has a clue about. We don't know if it was rushed or not. I was under the impression that Tahiti was ready months ago and was just waiting for TSMC to be able to manufacture it for AMD. WHQL is just Microsoft giving the drivers their blessing. I sure wouldn't not sell a product because I was waiting for WHQL certification. Finally, we don't know how many 7000 series cards AMD is selling. No idea what so ever.

The point was made amongst others to illustrate product maturity. There are WHQL drivers out there for all other existing cards.

Of course your buying decisions are your own. They don't influence the market. Nor, are they necessarily a reflection of it. It's just "you". If nVidia releases a card first or second week of March (2 Mos. later) that's 60% faster than a 580, stock for stock, you will be right, but I'll be really surprised. All rumors now don't have that happening until 3rd/4th Q. You do realize though you can O/C a 7970 right now and have it be that much faster than a 580?

They won't release such a card in March. I'm expecting GK110 in May or June but no later. O/C counts for people willing to do it, but there are several things to consider. Warranty, noise, necessity for an aftermarket cooler, time and effort spent and no guaranteed clocks. O/C is nice but it doesn't replace a faster product at stock clocks.

 

[Leadbox]

Q1 is not over yet. 2011 was just not doable with TMSC. Both IHVs launched in Q1 2012. AMD rushed out a few cards with no regard (again) to the reviewers, giving them very little time to test it. They could not get out many cards with decent voltages/clocks because of process variations, this is why the voltage on Tahiti is relatively high or the clocks low, however you want to look at it.
So no, until Q1 is over, Kepler is not late (by what standards anyway? rushed launch of the competition on an immature process?).



Couldn't agree more. Tahiti is a performance GPU, a performance/highend hybrid at best. As the successor of the 6970 it should roughly sell for the same price. The only two things permitting these prices right now are lack of competition and great OC potential. If you look at stock performance though, it's a bit underwhelming.

Consider that no company has to put out a significantly faster product than the competition. If they would do that, they would rob themselves of profit (less good chips due to high clock targets) and set the bar too high for their refresh. I could very well see Nvidia bringing a GK110 that holds a bit back because more performance is not needed. Why beat the competition by 40% when you can do it at 20% and higher profit? (just an example!). But this also depends on what AMD does when Kepler arrives. Will they bring a 7970 "Ultra" to compete or will they just lower prices?

If you're in anyway suggesting that Nvdia are waiting for all these little things you are calling AMD on, to come right before they release Kepler, you're having giraffe

 

[boxleitnerb]

The reasons why Nvidia didn't release in 2011 or in January 2012 for that matter are supply constraints (capacity for mobile chips needed to fulfill OEM contracts), bad yields in general and possibly the PCIE 3.0 problem that they needed to fix.

I guess Nvidia just wants a hard launch with good availability. Maybe that was/is not possible before February/March due to their priorities in the mobile segment. There are only so many 28nm wafers and so many chips. As for GK110, I don't know. For such a large chip they might really be waiting until yields are better. Or rather they have waited - GK110 has taped out after all.

 

[Arzachel]

The reasons why Nvidia didn't release in 2011 or in January 2012 for that matter are supply constraints (capacity for mobile chips needed to fulfill OEM contracts), bad yields in general and possibly the PCIE 3.0 problem that they needed to fix.

I guess Nvidia just wants a hard launch with good availability. Maybe that was/is not possible before February/March due to their priorities in the mobile segment. There are only so many 28nm wafers and so many chips. As for GK110, I don't know. For such a large chip they might really be waiting until yields are better. Or rather they have waited - GK110 has taped out after all.

No way are they waiting for a hard launch seeing how they launched the GTX480 with barely any availability. I think it's more likely we'll see a paper launch or the cards will be in short supply.

 

[Crap Daddy]

No way are they waiting for a hard launch seeing how they launched the GTX480 with barely any availability. I think it's more likely we'll see a paper launch or the cards will be in short supply.

What has the GTX480 launch got to do with this one? Things change, you know, smaller chip, good perf/watt, drop the hotclocks, bag for the buck, leaks point out tho these things very unNvidia.

 

[railven]

Wow, people are still arguing this? By my definition, nVidia is late. Period. Care to argue that with me, trust me you'll lose. And if they don't release by April they're too late. Anyways.

So this is officially being dubbed GTX 670 ti or is that just an early rumor still? This card sounds perfect for me. +/- 10% GTX 580 Performance in the $200-350 price bracket.

Now, for nVidia to deliver on this product and I'll be taking my vacation from wadding through all this mud sligging.

 

[96Firebird]

Wow, people are still arguing this? By my definition, nVidia is late. Period. Care to argue that with me, trust me you'll lose.

Right, because your definition of late is the correct one? I think not...

There is this thing called an opinion, everyone has a different opinion as to what late is. That is why I laugh when people start pulling up slides and old rumor articles to prove why they're opinion is the right one.

 

[boxleitnerb]

Wow, people are still arguing this? By my definition, nVidia is late. Period. Care to argue that with me, trust me you'll lose. And if they don't release by April they're too late. Anyways.

So this is officially being dubbed GTX 670 ti or is that just an early rumor still? This card sounds perfect for me. +/- 10% GTX 580 Performance in the $200-350 price bracket.

Now, for nVidia to deliver on this product and I'll be taking my vacation from wadding through all this mud sligging.

Fermi was late by 6 months and Nvidia is still alive. They had good cards with the GTX460 and 470 that sold very well. Now if they're 3 months "late" (again, by YOUR definition which doesn't have to be everyone elses) with possibly even better parts and it's doom and gloom?

If they've been producing for mobile Kepler only until lets say January/February, it's only natural that it takes time to get the desktop chips out. Priorities shift. When/if GK104 has not shown its face in March, then you can be concerned.

Btw Kyle Bennett from HardOCP has most likely signed an NDA. He has hinted at that in their forums this morning.

 

[railven]

Right, because your definition of late is the correct one? I think not...

There is this thing called an opinion, everyone has a different opinion as to what late is. That is why I laugh when people start pulling up slides and old rumor articles to prove why they're opinion is the right one.

Hey kettle, long time no see. And of course my definition is correct. See how that works.

Me: 1/ You: 0

 

[Arzachel]

What has the GTX480 launch got to do with this one? Things change, you know, smaller chip, good perf/watt, drop the hotclocks, bag for the buck, leaks point out tho these things very unNvidia.

While the arch might be totally different, the circumstances are a bit similar(Problems with the process, slower to bring out products on a new node than AMD) but I didn't want to compare Kepler with Fermi, just pointing out that Nvidia can and will launch products with next to no availability if it's in their interests, just like AMD paper launched Tahiti.

 

[railven]

Fermi was late by 6 months and Nvidia is still alive. They had good cards with the GTX460 and 470 that sold very well. Now if they're 3 months "late" (again, by YOUR definition which doesn't have to be everyone elses) with possibly even better parts and it's doom and gloom?

If they've been producing for mobile Kepler only until lets say January/February, it's only natural that it takes time to get the desktop chips out. Priorities shift. When/if GK104 has not shown its face in March, then you can be concerned.

Btw Kyle Bennett from HardOCP has most likely signed an NDA. He has hinted at that in their forums this morning.

Buddy, learn to read - I never once hinted Doom and/or Gloom. Sorry, I'm not your intended target with the Defense Force.

I'm actually looking forward to this card, I just hope it isn't too late, by my definition of course.

 

[OILFIELDTRASH]

AMD missed the boat on the 7XXX series. They had a small window competition free and instead of pricing the 79XX series decently they priced many buyers out of the market. They lost many potential sales. If it was priced better there would not be one 7950/7970 available. Instead a quick look on newegg shows lots of stock on both. In fact most of the 7950's have never went out of stock. Just read through some of these threads and you'll see many cases of people that would of and wanted to upgrade yet they were priced out of the market.

 

[boxleitnerb]

Buddy, learn to read - I never once hinted Doom and/or Gloom. Sorry, I'm not your intended target with the Defense Force.

I'm actually looking forward to this card, I just hope it isn't too late, by my definition of course.

Defence force? Are you implying I have an agenda? We're just discussing here, no need to pull out the fanboy tag. And I don't care for your condescending tone either.

You said "too late" Too late for what? For them to sell anything? If you're not specific, don't complain about the interpretation. Nvidia and AMD have not always launched products so close to each other. In fact this is difficult if not impossible for a large company that plans months and years ahead and is dependent on external manufacturers. So I don't really get why people cry for simultaneous launches and then complain on or the other is late. That is not realistic imo, never has been.

 

[railven]

Defence force? Are you implying I have an agenda? We're just discussing here, no need to pull out the fanboy tag.
You said "too late" Too late for what? For them to sell anything? If you're not specific, don't complain about the interpretation. Nvidia and AMD have not always launched products so close to each other. In fact this is difficult if not impossible for a large company that plans months and years ahead and is dependend on external manufacturers. So I don't really get why people cry for simultaneous launches and then complain on or the other is late. That is not realistic imo, never has been.

Again, where did I call you a fanboy? I read through this read and you've clearly got an agenda to defend against the opinions of others claiming it is late. Who cares? I don't.

If they don't release by April, they're too late as my window to buy would close.

 

[boxleitnerb]

You implied it, that's enough for me. I don't have an agenda, I just have a different opinion and stand by it with arguments and examples.

Fine, that clears it up.

 

[railven]

You implied it, that's enough for me. I don't have an agenda, I just have a different opinion and stand by it with arguments and examples.

Fine, that clears it up.

That's fine and dandy, but when you come off debating someone who said they are looking forward to the card and then implying they are wishing Doom and/or Gloom - well, what would you expect me to say?

 

[boxleitnerb]

I never said anything about "wishing". Obviously it was a misunderstanding because it was not clear what "too late" meant.

Anyway, let's get btt.

 

[Leadbox]

AMD missed the boat on the 7XXX series. They had a small window competition free and instead of pricing the 79XX series decently they priced many buyers out of the market. They lost many potential sales. If it was priced better there would not be one 7950/7970 available. Instead a quick look on newegg shows lots of stock on both. In fact most of the 7950's have never went out of stock. Just read through some of these threads and you'll see many cases of people that would of and wanted to upgrade yet they were priced out of the market.

They brought you a card that's faster than the previous fastest at the same price point, can't get fairer than that.

 

[Arzachel]

AMD missed the boat on the 7XXX series. They had a small window competition free and instead of pricing the 79XX series decently they priced many buyers out of the market. They lost many potential sales. If it was priced better there would not be one 7950/7970 available. Instead a quick look on newegg shows lots of stock on both. In fact most of the 7950's have never went out of stock. Just read through some of these threads and you'll see many cases of people that would of and wanted to upgrade yet they were priced out of the market.

They didn't miss any "potential sales" because they would have no product to sell, how can you not wrap your head around it after all this time? Wafers are short and yield isn't that great, like you'd expect from a new process. Also where are you seeing lots of 7970/50's in stock? 3/10 for 7970 and 5/9 for the 7950 and all of them are priced above MSRP. You'd think that would be the first thing to drop, if they weren't selling.