7970 GHz edition incoming!

[tviceman]

normally i wouldn't care, if the 680 was more efficient but slower than the 7970 across the board, to me, the 7970 would be better. But when it comes with also being faster in the games i am going to play it is nice.

Even though i plan to watercool the card, it sill means less heat for my loop to dissipate and less hot air blowing out of my system. Coming from 480s i think i could stay on air and still be more than satisfied with either a 680 or 7970. I couldn't believe how quiet these cards under loud.

http://www.youtube.com/watch?v=-svrvajgflq&list=uuxuqsblhae6xw-yeja0tunw&index=6&feature=plcp

starts at 7:10 in.

2:58 lol

 

[Tempered81]

Groove, there is almost no reason to water cool a 680 at this time. First of all, the reference card has a 4 phase VRM (ie the suck)
But as far as water cooling? I don't see a reason. The reference VRM is cheap

no doubt. wasting money on a waterblock for that pos pcb is like waxing a pile of scrap metal

 

[rgallant]

What? I have a 680, it doesn't overclock itself to 1200mhz. It does a 50mhz boost from time to time, which is a stock feature. Stock vs stock is all the graph needs to show. Having to show multiple graphs for every time the boost kicks in is really dumb. Reviewers didn't disable turbo mode on SB processors when running the review. Stock is stock is stock, there isn't much more to say about that.

-I see a gpu-z screen shot showing a clock value but yet the benches STATE the 680 is faster by this much[clock in gpu-z] but at what clock's and some reviews have stated high clock's during benching.
-when they say you can oc 30% on top of the benches you just viewed =bs
-if the card was using part of that 30% when it was benched how can you count that twice or am I missing something

just saying
re sb what when reviewers did benches and if the sb turbo went to 4.8 ,with a cpuu-z shot showing clock's at 3.4 but when the benches were run sb was at 4.8 , then they show you can oc a sb 40% showing a new cpu-z shot at 4.76[40% oc],leading people to think they can get 40% more[points \speed] and if fact it would be slower. but people would not be fooled with a cpu ,but are any fooled with these gpu benches.

{You're completely missing the point of James's post.

-The reviews were sometimes shown with cards peaking over 1300MHz (H), and running 1200MHz+ (H,whitepaper).
-People buy based on reviews.
-The boost that the typical consumer will get is supposedly on average 1058MHz.
-The typical card purchased would appear to not even perform up to stock review performance.
-It is therefore not a great feature for the average joe, since he gets less out of his card while still having his warranty.

We could be wrong here, and the dynamic nature could be static for all cards running the same software (example:all cards run BF3 at 1200MHz). That would appear to be questionable though, as an average of 1058MHz is being thrown around.}
http://www.semiaccurate.com/forums/showthread.php?t=6262&page=63
post 628.

 

[Grooveriding]

no doubt. wasting money on a waterblock for that pos pcb is like waxing a pile of scrap metal

LOL

It would be cool if someone with a 680 could try testing GPU boost's effect with the fan forced to 100% vs at normal with overclocking. I'd like to know if lower temperatures have an affect.

Trouble for me with waiting on non-ref cards, is getting a block for them. There are only 1 or 2 that will get a block made for them and it will be a long while out. The first option will probably be a $700 evga classified with block, and eff that.

From what I've read the 680's pwm is completely hard locked and can't be controlled via software and has a hard cap of 1.21v. So 680 OC comes down to what you can get at 1.21V and if your card will even allow you to use all those 1.21V within whatever confines gpu boost defines for that voltage being safe. The only way around this is a hard mod of the card and I'm not interested in that.

I'm going to watch the forums after the blocks are available and see what sort of results people get using them. Would also like to see some more reviews on OC results/scaling.

I'm without cards atm though. So just throwing the 680s up on fleabay to wait for better is not really an option as I have nothing to use right now.

 

[Tempered81]

Yeah they boost higher with lower temps

I posted a chinese "boost analysis" article somewhere here on this forum. Just cranking his fan up to 100% gave him the same boost in 3dmarks as overclocking the core and increasing power threshold with stock fan.

 

[GodisanAtheist]

amd was years fighting nvidia with a mid range chip....now that nvidia have an eficient chip, the sweet-spot strategy dies ( if bigK exist )

-Literally nothing would make me happier than seeing AMD backing down from their "sweet spot" strategy (which they have steadily done with creeping die sizes since the 3870->4870 transition onward) and just go balls out with the biggest monster of a die they could assemble and scale down from there, like the old days. Sweet spot cards might make sense for the initial node transition to keep prices in check, followed with a no holds barred juggernauts at the re-spin after the node has matured. This is just me fantasizing...

I think they aren't getting as much flack because we all know Big K is coming. 7970 is the likely the top single GPU part AMD is releasing for the 7000 series (baring higher clocked versions of the same chip) If there was no Big K then GTX 680 would be much more disappointing. Probalby if NVIDIA launched first with $500 "mid range" part they would get more flack too, but since they came second they don't get as much crap for that.

-Problem with this logic is given enough time, the BigK has to prove itself against the 7970 and 680, NOT the 580. Given the 680's surprisingly lack-luster compute performance, we can assume a good portion of the BigK's extra die space will go to a larger bus and compute performance and will not yield a proportional increase in gaming performance for its die size (like the 560Ti to the Gtx 580). It might end up being as disappointing a bump over the 680/7970 as those cards were over the GTX580...

 

[badb0y]

I think they aren't getting as much flack because we all know Big K is coming. 7970 is the likely the top single GPU part AMD is releasing for the 7000 series (baring higher clocked versions of the same chip) If there was no Big K then GTX 680 would be much more disappointing. Probalby if NVIDIA launched first with $500 "mid range" part they would get more flack too, but since they came second they don't get as much crap for that.

It really depends on when it lands, if it lands by the end of Q2 early Q3 then we can say that AMD is screwed but if BigK is released by the end of Q3 early Q4 then it isn't competing with the 7970 anymore but probably a 7980 or 8970 or whatever AMD decides to call their refresh.

I am assuming AMD will be releasing something by December/January.

 

[3DVagabond]

Update: Ocaholic has an article with more info. According to them it's not merely an O'C'd card.

About the new chip itself, there are two major improvements from the Tahiti XT to the Tahiti XT2. First of all the clock frequency will be at least 175 MHz higher for the Stock models as the XT is clocked at 925 MHz and the GHz Edition should be clocked at 1'100 MHz. Then the default voltage of the chip changes too going from 1.175 volts for the Tahiti XT to only 1.020 volts for the Tahiti XT2 which is almost 0.15v.

That last part could mean that with some voltage they'll O/C substantially higher than current cards. Here's hoping. :crossfingers:

 

[toyota]

so a 175mhz increase but voltage drops from 1.175 to just a minuscule 1.020? I find that very hard to believe for just a new stepping.

 

[Don Karnage]

They just need to release some 1200 core 7970's so Nvidia could counter with MORE VOLTAGE and higher clocked 680's

 

[toyota]

They just need to release some 1200 core 7970's so Nvidia could counter with MORE VOLTAGE and higher clocked 680's

Nvidia will counter with a gk114. lol

sadly they can do nothing about their memory bandwidth issues. 7000mhz is the fastest gddr5 made but supposedly 6000mhz is the fastest memory that is actually available in high enough production.

 

[Don Karnage]

Nvidia will counter with a gk114. lol

sadly they can do nothing about their memory bandwidth issues. 7000mhz is the fastest gddr5 made but supposedly 6000mhz is the fastest memory that is actually available in high enough production.

I can't wait to see what GK114 can do

 

[Homeles]

Nvidia will counter with a gk114. lol

sadly they can do nothing about their memory bandwidth issues. 7000mhz is the fastest gddr5 made but supposedly 6000mhz is the fastest memory that is actually available in high enough production.

All it takes is AMD or Nvidia improving their memory controller enough to utilize 7GHz GDDR5, and suddenly production issues will disappear...

 

[HurleyBird]

so a 175mhz increase but voltage drops from 1.175 to just a minuscule 1.020? I find that very hard to believe for just a new stepping.

If that's true then that's pretty damn impressive, and if that story is accurate 175 MHz is just the minimum clock speed improvement we can expect.

Looks more or less like the new 7970 will be to the GTX 680 as the GTX 680 is to the current 7970.

Best part is that if this is true Nvidia will likely respond in kind. Given a better PCB and a higher power target they could probably squeeze quite a bit more performance into a hypothetical GTX 685.

It's about time things started heating up again.

 

[Jacky60]

The clock race is on! First one to blow up loses. haha.

LOL Why does it take more than 6 months to achieve what pretty much all 7970's can do out of the box?

 

[3DVagabond]

LOL Why does it take more than 6 months to achieve what pretty much all 7970's can do out of the box?

Reading a bit deeper than the clocks, it looks like TDP is what has changed. 1100MHz @ 1.02v instead of 1.175v

Supposedly this little darling is to be clocked at 1000MHz. Makes the 1.02v sound believable. TDP is more important in the professional field. I don't think they would have upped the clocks if they couldn't lower voltage.

 

[blackened23]

Reading a bit deeper than the clocks, it looks like TDP is what has changed. 1100MHz @ 1.02v instead of 1.175v

Supposedly this little darling is to be clocked at 1000MHz. Makes the 1.02v sound believable. TDP is more important in the professional field. I don't think they would have upped the clocks if they couldn't lower voltage.

TPU stating that the revised tahiti is 1100mhz at 1.02V, not sure if this is a typo..http://www.techpowerup.com/167711/AMD-Radeon-HD-7970-GHz-Edition-quot-Tahiti-XT2-quot-Detailed.html

Now,posts from random green zealots about drivers that will destroy your PC, power consumption that will make your house to catch fire, etc, to follow.

 

[3DVagabond]

TPU stating that the revised tahiti is 1100mhz at 1.02V, not sure if this is a typo..http://www.techpowerup.com/167711/AMD-Radeon-HD-7970-GHz-Edition-quot-Tahiti-XT2-quot-Detailed.html

Now,posts from random green zealots about drivers that will destroy your PC, power consumption that will make your house to catch fire, etc, to follow.

TPU is quoting Ocaholic as the source.

If rumors are true, avg. power usage will be below Kepler while being faster. If this happens I predict there will be a major flip-flop on efficiency, just as there has been with compute. (This is on both sides of the fence. Not just team Green.) Always entertaining to watch.

 

[KompuKare]

TPU stating that the revised tahiti is 1100mhz at 1.02V, not sure if this is a typo..http://www.techpowerup.com/167711/AMD-Radeon-HD-7970-GHz-Edition-quot-Tahiti-XT2-quot-Detailed.html

Now,posts from random green zealots about drivers that will destroy your PC, power consumption that will make your house to catch fire, etc, to follow.

Isn't drivers which destroy cards, fires and having to use ovens with graphic cards a trademark of Nvidia Corporation?

 

[railven]

I'm interested to see how this pans on. My current HD 7970 gets a little toasty. I WAS going to buy a Arctic Cooler, but if these things pan out nicely I might sell my card for $400 and snag one of these - IFFFFF.

Might just be smarter to go the arctic cooler route, but I personally like the stock shroud, just looks sexier in my case (and yes I can see into my case )

 

[blackened23]

I'm interested to see how this pans on. My current HD 7970 gets a little toasty. I WAS going to buy a Arctic Cooler, but if these things pan out nicely I might sell my card for $400 and snag one of these - IFFFFF.

Might just be smarter to go the arctic cooler route, but I personally like the stock shroud, just looks sexier in my case (and yes I can see into my case )

I have an arctic cooling 7970 extreme if you want one, i'm not out for $ no worries there, I just don't need it anymore. You'd have to get thermal adhesive but it will still work (and you could throw it on your ghz edition or whatever)

I wouldn';t suggest it for crossfire but in single card configs it is downright amazing and silent even at 100% manual fan (with really good temps - around 55c load while heavily oc'ed and overvolted)

 

[railven]

I have an arctic cooling 7970 extreme if you want one, i'm not out for $ no worries there, I just don't need it anymore. You'd have to get thermal adhesive but it will still work (and you could throw it on your ghz edition or whatever)

I wouldn';t suggest it for crossfire but in single card configs it is downright amazing and silent even at 100% manual fan (with really good temps - around 55c load while heavily oc'ed and overvolted)

PM Sent....

 

[tviceman]

TPU stating that the revised tahiti is 1100mhz at 1.02V, not sure if this is a typo..http://www.techpowerup.com/167711/AMD-Radeon-HD-7970-GHz-Edition-quot-Tahiti-XT2-quot-Detailed.html

Now,posts from random green zealots about drivers that will destroy your PC, power consumption that will make your house to catch fire, etc, to follow.

1100mhz will be enough to pull it slightly ahead of the gtx680 almost entirely across the board; my question is why did it take AMD so long to get off their horse to release this? If they actually did a small revision to Tahiti, and as such had to respin it, then fine that makes sense. But if these are just the exact same chips with new binning procedures, then there are people at AMD that need to be strung up and beaten like a pinata.

 

[tviceman]

Nvidia will counter with a gk114. lol

sadly they can do nothing about their memory bandwidth issues. 7000mhz is the fastest gddr5 made but supposedly 6000mhz is the fastest memory that is actually available in high enough production.

I can't wait to see what GK114 can do

Exactly what Toyota said. Unless they couple it with 7000mhz vram, a GK114 chip might be more efficient but it won't be much faster than 5-8% over GK104. Depending on what AMD prices these "new" hd7970 and hd7950 parts at, Nvidia might not have to do anything (if AMD prices it back to $549) or they could drop their prices slightly and swallow losing the performance lead by a few percent.

 

[lopri]

Is it a new stepping or same thing better binned?