I wouldn't bother. After his last post it's pretty obvious he's just trying to troll for responses. Reality isn't important. What is important appears to be what is the most inflamitory, even after he's shown to be mistaken.
I wouldn't bother. After his last post it's pretty obvious he's just trying to troll for responses. Reality isn't important. What is important appears to be what is the most inflamitory, even after he's shown to be mistaken.
Hey man, I never tried to troll you. You have done nothing but attack me this entire thread. You called me a liar, said the post was full of "blatant lies" and that I made all sorts of claims. You even posted I made this claim by me quoting HardOCP. Those were Kyle's and Brents words not "claims". I posted benchmarks, and did not say a word, then you come in accusing me of "claiming things". The only thing I ever said was "crushed" and showed the 7970 OC benches and you turned that into something completely different. You called me a troll and Blackend a troll and told Elfear to re-read his own links. You are the only one "mistaken"
Last edited:
I'm going to try to respond to your nonsense without getting angry and flaming you.
Just because the GTX 680 has a "turbo mode" it does not mean that it's overclocked. It essentially dynamically downclocks itself to save on its power consumption.
Overclocked-vs-overclocked, the GTX 680 wins. Period.
http://alienbabeltech.com/abt/viewtopic.php?f=6&t=24550
OP there are people speculating that you and other people here are members of an "AMD focus group" of some sort. I don't appreciate your calling out the nVidia people when it could well be that you work for the other side.
I don't really prefer either brand but I'm sick and tired of this type of flamebait and nonsense ruining the forum.
That's fine, I'll take part in a sensible discussion. No I am not a member of an ATI Focus Group. Yes the GTX680 wins that ABT reivew, except the GTX680's memory is clocked at 7200mhz and is a clear 1000mhz faster than the 7970's. Even given this, the averaged difference is .02% in favor of the 680. and -0.02% if you remove Unigine scores (per BoFox's comments).
You can counter that and say, well the 7970 holds all the top end records (except for Kingpin's 1950mhz record in 11P). But some will argue that tesselation is disabled to get those scores.
I likely come across as one of the pro-ATI people here, sometimes I try to balance the Nvidia Focus Group mentality, I admit it. I think a fair discussion should take place, it shouldn't be so green tinted as it was in the past with our VC&G moderators working for Nvidia.
Fair enough. I used to do that as well. The nVidia people drove me crazy. Now I just ignore them. Most if not all of them are "out" now so we know who they are.
Honestly as hardware enthusiasts we should all be excited because things have really heated up in the graphics business. It has been some time since AMD and nVidia went toe-to-toe like this.
I think it's fair to say that the GTX 680 is the better card for most people by most metrics, albeit slightly. There are cases where the 7970 would be better for some.
But really, I don't think it makes sense to twist facts or to try to paint things in a certain way to make one side look better. It just creates flame wars as you've found out.
So I am going to pay more for a card that has a stock counterpart that is already more expensive than the competition, and it is still slower?
Ehhh.......no.
Ehhh.......no.
Heheh, yes a flame war. How else do you discuss it sensibly? Forgive this post its about to go way off topic with bad typing.
*Firstly i should say that im an ATI fan partial to AMD, my last cards have been a 6970 and 6850, 4890 and GTX 260* So if you plan to read this with tinted glasses, make sure the lenses are green and not red if you want an even perspective please. Thank you.
TL;DR begin:
The 7970 is a good card, and so is the 680. The 680 is designed better for popularity and sales. Thats where Nv beat them. It will be the higher selling card. The seem to give a rats a$$ about the driver support too, so that should help.
One thing we can all agree on as fact is that Kepler IS FASTER than Tahiti. If you were to put both of them at the same clocks, the 680 can be 2-5% faster varying in titles and situations. 1100/1500 is a good comparision, the 680 still takes a slight edge. Thats what made it high end int he first place
The double edged sword or enigma or whatever is that GK104 could blow tahiti out of the water, rather than best it by 10%. If AMD wanted we would have a 1250/1800mhz Tahiti board, and if Nvidia wanted we could have a 1350/1800mhz GK104 board. (Okay if Nvidia really wanted we could have a 1000mhz GK110 board that would make every AMD fan STFU real quick.) The thing is, the 680 comes with 1gb less vram - for agruments sake that makes no difference except Skyrim fan-made HD texture packs and Triple screen 4xAA ultra BF3 or something equally egregious. The 7970 comes with a larger memory bus, and 3gb, but a complete lack of regard for how their fans interpret their multi-gpu and gameplay experiences. Nvidia seems to listen to the gamer crowd, pump them up, give them what they want, and support the gaming community. Of course they're gonna do that and expect an advantage or ROI. The downside is the low-cost pcb on the 680. You can argue that this is/is not low-cost, but if you read instead of whine you will see what I mean.
AMD builds the 7970 board, and seems to not even give a rat's ass if crossfire doesnt work on a triple A title, or if they don't have WHQL drivers out for 4 months. But amazingly the board is incredible. With an asus bios, you can push 1400vddc/1800mvddc through a very impressive array of power components. I won't go into detail on these components, but if anyone requests that I do, I surely will.
If you were to attempt to push 1400/1800 through that 680 ref pcb, it would be a fireworks display, excpet gk104 wouldnt be lighting on fire, it would be the sam-choice mosfets. GK104 can take the heat (or cool). So theoretically, if amd ever put out the super clocked tahiti (1200/1800), NV could easily build a robust board for gk104 and clock it at 1350/1800 and take the win right back from the new "tahiti product".
This whole argument lies in the fact that Kepler is so incredible compared to Tahiti. Kepler is a HUUUUGGEE win for Nvidia. Mainly cause NV was smart enough to survive with ginourmous dies like GF100 and GT200, while still being a money-earning company. If Sony, Nintendo and MS had time to see Kepler in action, then AMD would have no console contracts right now, Much less all of them combined. If Kepler and Maxwell continue this brilliant engineering trend, then AMD is SOL. and PS5, XBOX4, and WII U2 will all be nvidia.
But about 680 & 7970. Well despite the fact that PC gaming is about to be stalled by the next gen consoles (more than it already is today), Nvidia could technically drop in limited quantities their beast GK110 that we could all have a field day with. That chip is going to be so great, and AMD has nothing to respond with that I'm aware of. They have to have a 40CU + GCN in the works (like that rumored Tenerife slide) - which that slide is complete bs anyway, since it came from SA forums and is based off Charlies Canary/Sea/ "C" islands rumors.
Even if 'tenerife' isn't bs, then 40Cu Oc'd gcn2.0 probably still wont be able to take a 384bit/512bit 2304 CC Kepler on a clock-per-clock basis.
Stay focused here - about 680/7970. Here are some bullet points i consider factual:
The 7970 has more potential. board for board comparison there.
Gk104 has mroe potential than Tahitixt, chip for chip, clock for clock
Tahiti has more memory bandwidth (useful for shit like crysis, metro, multi dsiplay, multi-gpu extreme high res)
GK110 puts the memory bandwidht argument to rest.
680 is a more profitable, higher selling (faster performing) product than 7970
Nvidia is about to make a butt ton of money off Kepler, because they are already marketing and distribution magicians. They care about the community and being involved in game development. Their lack of engineering is the only thing that got AMD those console contracts, that and because NV had a costlier bid. Kepler is about to change all that.
Unless ati has some miracle performance metrics coming out of Sea islands, I don't see them beating kepler/maxwell. AMD's downfall is a whole different story, and now my fingers are tired and head hurts, and this is TLDR already. I bet the punctuation is headache-inducing too. sorry
Continuing right along! Amd is like a corporate de-evoluiton in progress, IMO (and far-fetched) 7970 was the "last great graphics board" from AMD. For many reasons... consolitis, weak next-gen consoles, amd/ati merger, small chip design, fusion evolution elements, intel's iGP, and Nvidia's compute strategy.
Nvidia supports HPC more than AMD, AFAIK, they lose ground on the fact that AMD is a veteran tenured company with a history of JEDEC, ISO, MS development. AMD/ATI is the reason for shit like directx and tessellation, but haha, now Nvidia can do that stuff even better than "GCN"
FOK "GCN" i love this crap. Virge disagreed with me when i told him Cayman was going to drop the T-unit and go for a 4-way very long instruction. He said that I should "humor him along with the rest of the folks in the peanut gallery". Well haha! I bet he knew all along it was VLIW-4 anyway, and was just egging me on back then. GCN is nothing more than an evolution of the freaking shader design resulting from the disaster R600. Rv670 if i recall happened right around the time AMD took over, and their design got smaller and more efficient. That rv670 vliw-5 design was great just like G80, and Kepler are. That design evolved into the 3870,4870,5870, and was altered at its biggest point at 6970/cayman. That is about the extent of development for them. I ask why the hell dont they make a 28nm cypress or a 28nm phenom II x4 for that matter. (oh yeah and faildozer doesnt make them seem any more ultra intelligent like something great is gonna come from the last few remaining bitboyz at ATI). ATI's falling apart. Well crap i mean AMD, ati is done already. Killbrew is gone, and look at all the management changes and ceos leaving and people being fired, that company is TOTALLY REFORMING. I hope something like the bitboyz R300 "respawns" from the pile of ashes when the dust settles, but somehow I doubt it will. Ati Canada is gone. DoomIII and R300 and G80 are gone. GK110 is the next BIG thing.
I remember anands old article about Dave Bauman and his elite dinner discussing (sunshine) or whatever was the codename of what was to become "Eyefinity". Thats a cool article for anyone to read. Check out the discussions on b3d, or just go enjoy your fav game on your 680, you lucky bastard. Historically ATI was always on the leading edge for engineering milestones, and I could be wrong but i dont see that happenign again. They were the first on every node transitiion at TSMC, they invented dx standards and tess, but now theres GF and TSMC, now theres Tri-gate and 20nm at intel's inhouse. Now theres the end of the road for standard silicon, the problems with tscm 40nm, 28nm and Nvidia's slides asking for new ways to earn money in lithography relationsihp with tsmc. Things are changing a lot. Nvidia could have their own in-house Tri-gate 20nm fab making SOC smartphone tegra shit in the next year or two. These are some reasons i thought the 7970 to be the last great AMD gpu. Gk110 better, and maybe the last big nvidia thing, untill the maxwell beast arrives. The environment is changing a lot around whats gonig on in the GPU landscape.
Nvidia markets better, we all knwo this. They have PAX, GTC, PDX, blah blah blah, and tons morer where that came from. They have "way its meant to be played" and a lot of other cool things that are really ntoable from a technicaly perspective.
My favs. That new ambient occlu in all the games from driver panel, that super quiet fan, and very compatible NVCP, fxaa, smaa, txaa, 20 different aa modes, 3d vision, and the glasses, even that stacked pci-e power connector thingy, Surround (copied from eyefinity) and now Kepler uArch with regular driver updates. I think Jensen was suprised himself when he said at the conference call that "kepler was tyhe best architecture they every made" It actually is really frickign good. The other good thing is that gpu boost that we're arguing about as auto-overclocking in thsi thread. That is a SMART EVOLUTION in graphics card design. They are squeezing the most out of their product which is in almost all ways better for us as the consumers. We win on every level, and the losers are the folks clamoring for ATI to add boost to their next gen 7970, when the reality is theyd ont give a shit and are busy making next-gen console gpus and reforming their company ranks. Just like nvidia is smart at evolving the cards themselves. Like begging for a change in the layout of intels motherboards, and the power limits of pci-sig. But this only lasted long enough until they had a faster card than the competition. Now who gives a crap about an inverted case where the HSF is ABOVE the gpu, and you can safely shoot 800 watts onto the video card. Nvidia makes Better designed coolers that allow for intake of air in a sensible way, Aerodynamic fan shroud. Nvidia card designs always looked like a ferrari to me, while amd's cards look like Mack dumptrucks. then if you were to compare Cypress to Fermi, Cypress is a ferrari while Fermi is a dump truck. Then if you compare Kepler to GCN, they're both high end ferraris, while kepler is a little faster and higher quality limited edition "Orange Lambo" if you will. But the cards, 7970 is still a dump truck and the 680 is an abosolule ferrari.
Now if you look at the PCBs... hehe 680 is a golf cart/moped/kia, and 7970 is a rolls royce, thats the only way i can safely put that, without trying to hurt someones feelings here. 680 owns in many metrics, but that PWM is not one of them. The power delivery is weaksauce, try and debate me on it, anyone im begging you to make a sensible argument about that power design. PLEASE!
Kingpins' billy-badass 680.5 board is a perfect example. He soldered a video card ON TO A VIDEO CARD for crying out loud. Yeah it set the record, but is it even a 680 anymore?... I mean jesus. LOL. It's a friggin space ship, not a graphics card.
Okay 680/7970 war, one more time (stay focused man) The 680 wins... But is it that easy? The 7970 has the world record. Andre Yang and Peter beat Vince and his alien 680 space ship technology, but only because Vince couldnt put 4 gtx 680 space ships on one Motherboard. With just that one sample at the uber 2Gz, he beat Andre's previous world record (which Vince argues is a cheated record anyway because the of the tessleation giving andre an extra 1-2k points). But hey Andre is a convicted cheater any from years past, and he was only last year unbanned from the XS community. But his cheating was because he knew how to get high scores and he would share them with others for money, not because he was faking the score he is a LEGIT bencher. Monety- Thats why he got banned. EVGA and Vince getting that 680 score above Andres 1.7g and Elmors 1.8G 7970's was a Press Statement more than anything. He said, let us inject this gk104 board with a friggin lightning bolt and see what happens *although you knwo behind the scenes he was begging Nvidia for a GK110 Engineering sample* but they said no, fok you You cant have a sample because if you set some super high retarded score, you're gonna want to show everyone, and there will be a massive riot cuz everyone wants that kind of performance. Yes folks, Kepler is a beast.
The very fact that he was able to beat a 7970 with a gk104 should tell the tale, regardless of how weak the 680 reference board is. And the 680 reference board's pwm is not all that "WEAK" per se really since it does its job and makes people happy by not exploding. Its not weak in that sense, its more like the pwm is "just enough to get by" with some little bit of overclockign headroom to boot. It OCs some, but they still stop you at 1.2v for the F of it. Make room for the primo models, like the Classified, doesnt hurt my feelings.
In the end if I had both a 7970 and a 680, somebody may wonder, what would this lunatic do with these two cards?
Well I would overvolt both of them. I would first off want the 4GB classified with the extra power and 14phase pwm for the 680, and i would go with a dual x sapphire 7970 or a watercooled lightning or something. I would probably put them on water blocks and oc the fark out of them and would want to sample many cards of each, since I know overclockablity differs by sample. But sadly I have neither, and my opinion is invalidated. Ride it out on my 6850 while i play borderlands at medium res, med qual...
you got 4 ppl
guy 1: My 680 ocs like shit, only does 1095 on baseclock
guy 2: my 7970 ocs like crap, only does 1080 on baseclock
newbie3: woah guys my 680 must be good, runnin fine at 1320/1800
newbie4: check my 7970, its at 1375/1950.
The factors at play are #1 VARYING SAMPLES. Somebody whos really good at overclocking could get a sample that stops them at 1080mhz - MURPHYS LAW. Someone who sucks and is newbie at overclockign could get a sample that is golden, best every seen, rarest, and be stuck at 1050 because he cant figure out how to unlock CCC limits or something dumb, or doesnt evne know how to OC in the first place. Like "oh i run stock" and that guys card might be the 1400mhz jewel diamond in the rough.
So you have to knwo what you're doing and push it to the limits, but you also have to get a good sample in the first place. Thats why those guys have 20 GTX 680s or 20 HD7970s in the same box, cuz they're testing them all for the rarest finest sample they can find.
Nvidia is winning right now. The strategy is great $499 680 GK104 (which they could only produce 200 of them before TSMC problems ramping big lots) Props to anyone who got one early. 256bit bus, and it still rocks the competition. Tahiti is good, and going by that OC review from Hard, they can both OC really high, but in the end, the OC 680 still bests the OC 7970. HOW-EVER the WORLD RECORDs are still held by the 7970. Not sure how to approach this, but it's definitly not as clear a win for as say 4870 versus 280 or 5870 versus 480 or 6970 versus 580. In those generations, the X80 beat the X870 eight times outa 9, not 50/50. Even when you overclocked both to their limits, the Nvidia top end still won, and they set all the HWBOT world records, this is the real MAJOR difference.
The 680 is not doing that. They could sell it at 460/560 prices and there would be no difference on the company spreadsheets. The last minute OC and name change was genius. The oem bios + 300.63 driver clocked it at 705mhz, the finalized + 300.99 at 1006/1058. It is engineered to be a high-selling, high-margin product that will be enjoyed by the masses and covers all bases. It overclocks, it does 3d, surround, plays all the games. Its a good product. It's not the Gx1x0, it is the Gx1x4. This says a LOT. It says Kepler is radical, game-changing and billy-badass in general. Kepler is the FTW graphics technology of the future, and is the nail in the coffin for the cursed GCN. I would love to have a 680, and thats part of the reason I hate the damn thing so much! Call me jealous or whateva/ I dont care. Its a wonderful product. It is a tad expensive, but you can put some of that on AMD for ripping us off on the 7 series (compared to their awesomely priced 3870/4870/5870/6970 that is). If Nvidia wants they could play this poker game and not show their all their cards... WHAT? NO WAY... GK110 is what AMD better watch out for. It's going to be a death blow for sure.
And the 20-25% rumors are because
1, gk104 is clocked higher than normal,
2, 110 will have compute centric built in because its being duplicated to Quadro, Tesla workstations and HPC,
3, the rumoured ratio of specs for 110/104 is not as big a gap in Kepler as it was in Fermi,
4, the probably wont have to make 110 a 1ghz sku since there wont be much comp.
Kepler was the big Nail in the coffin. Now time will tell the story of next-gen consoles, and GK110/maxwell/Sea islands.
Props if you read this. head hurts now. gg.
*Firstly i should say that im an ATI fan partial to AMD, my last cards have been a 6970 and 6850, 4890 and GTX 260* So if you plan to read this with tinted glasses, make sure the lenses are green and not red if you want an even perspective please. Thank you.
TL;DR begin:
The 7970 is a good card, and so is the 680. The 680 is designed better for popularity and sales. Thats where Nv beat them. It will be the higher selling card. The seem to give a rats a$$ about the driver support too, so that should help.
One thing we can all agree on as fact is that Kepler IS FASTER than Tahiti. If you were to put both of them at the same clocks, the 680 can be 2-5% faster varying in titles and situations. 1100/1500 is a good comparision, the 680 still takes a slight edge. Thats what made it high end int he first place
The double edged sword or enigma or whatever is that GK104 could blow tahiti out of the water, rather than best it by 10%. If AMD wanted we would have a 1250/1800mhz Tahiti board, and if Nvidia wanted we could have a 1350/1800mhz GK104 board. (Okay if Nvidia really wanted we could have a 1000mhz GK110 board that would make every AMD fan STFU real quick.) The thing is, the 680 comes with 1gb less vram - for agruments sake that makes no difference except Skyrim fan-made HD texture packs and Triple screen 4xAA ultra BF3 or something equally egregious. The 7970 comes with a larger memory bus, and 3gb, but a complete lack of regard for how their fans interpret their multi-gpu and gameplay experiences. Nvidia seems to listen to the gamer crowd, pump them up, give them what they want, and support the gaming community. Of course they're gonna do that and expect an advantage or ROI. The downside is the low-cost pcb on the 680. You can argue that this is/is not low-cost, but if you read instead of whine you will see what I mean.
AMD builds the 7970 board, and seems to not even give a rat's ass if crossfire doesnt work on a triple A title, or if they don't have WHQL drivers out for 4 months. But amazingly the board is incredible. With an asus bios, you can push 1400vddc/1800mvddc through a very impressive array of power components. I won't go into detail on these components, but if anyone requests that I do, I surely will.
If you were to attempt to push 1400/1800 through that 680 ref pcb, it would be a fireworks display, excpet gk104 wouldnt be lighting on fire, it would be the sam-choice mosfets. GK104 can take the heat (or cool). So theoretically, if amd ever put out the super clocked tahiti (1200/1800), NV could easily build a robust board for gk104 and clock it at 1350/1800 and take the win right back from the new "tahiti product".
This whole argument lies in the fact that Kepler is so incredible compared to Tahiti. Kepler is a HUUUUGGEE win for Nvidia. Mainly cause NV was smart enough to survive with ginourmous dies like GF100 and GT200, while still being a money-earning company. If Sony, Nintendo and MS had time to see Kepler in action, then AMD would have no console contracts right now, Much less all of them combined. If Kepler and Maxwell continue this brilliant engineering trend, then AMD is SOL. and PS5, XBOX4, and WII U2 will all be nvidia.
But about 680 & 7970. Well despite the fact that PC gaming is about to be stalled by the next gen consoles (more than it already is today), Nvidia could technically drop in limited quantities their beast GK110 that we could all have a field day with. That chip is going to be so great, and AMD has nothing to respond with that I'm aware of. They have to have a 40CU + GCN in the works (like that rumored Tenerife slide) - which that slide is complete bs anyway, since it came from SA forums and is based off Charlies Canary/Sea/ "C" islands rumors.
Even if 'tenerife' isn't bs, then 40Cu Oc'd gcn2.0 probably still wont be able to take a 384bit/512bit 2304 CC Kepler on a clock-per-clock basis.
Stay focused here - about 680/7970. Here are some bullet points i consider factual:
The 7970 has more potential. board for board comparison there.
Gk104 has mroe potential than Tahitixt, chip for chip, clock for clock
Tahiti has more memory bandwidth (useful for shit like crysis, metro, multi dsiplay, multi-gpu extreme high res)
GK110 puts the memory bandwidht argument to rest.
680 is a more profitable, higher selling (faster performing) product than 7970
Nvidia is about to make a butt ton of money off Kepler, because they are already marketing and distribution magicians. They care about the community and being involved in game development. Their lack of engineering is the only thing that got AMD those console contracts, that and because NV had a costlier bid. Kepler is about to change all that.
Unless ati has some miracle performance metrics coming out of Sea islands, I don't see them beating kepler/maxwell. AMD's downfall is a whole different story, and now my fingers are tired and head hurts, and this is TLDR already. I bet the punctuation is headache-inducing too. sorry
Continuing right along! Amd is like a corporate de-evoluiton in progress, IMO (and far-fetched) 7970 was the "last great graphics board" from AMD. For many reasons... consolitis, weak next-gen consoles, amd/ati merger, small chip design, fusion evolution elements, intel's iGP, and Nvidia's compute strategy.
Nvidia supports HPC more than AMD, AFAIK, they lose ground on the fact that AMD is a veteran tenured company with a history of JEDEC, ISO, MS development. AMD/ATI is the reason for shit like directx and tessellation, but haha, now Nvidia can do that stuff even better than "GCN"
FOK "GCN" i love this crap. Virge disagreed with me when i told him Cayman was going to drop the T-unit and go for a 4-way very long instruction. He said that I should "humor him along with the rest of the folks in the peanut gallery". Well haha! I bet he knew all along it was VLIW-4 anyway, and was just egging me on back then. GCN is nothing more than an evolution of the freaking shader design resulting from the disaster R600. Rv670 if i recall happened right around the time AMD took over, and their design got smaller and more efficient. That rv670 vliw-5 design was great just like G80, and Kepler are. That design evolved into the 3870,4870,5870, and was altered at its biggest point at 6970/cayman. That is about the extent of development for them. I ask why the hell dont they make a 28nm cypress or a 28nm phenom II x4 for that matter. (oh yeah and faildozer doesnt make them seem any more ultra intelligent like something great is gonna come from the last few remaining bitboyz at ATI). ATI's falling apart. Well crap i mean AMD, ati is done already. Killbrew is gone, and look at all the management changes and ceos leaving and people being fired, that company is TOTALLY REFORMING. I hope something like the bitboyz R300 "respawns" from the pile of ashes when the dust settles, but somehow I doubt it will. Ati Canada is gone. DoomIII and R300 and G80 are gone. GK110 is the next BIG thing.
I remember anands old article about Dave Bauman and his elite dinner discussing (sunshine) or whatever was the codename of what was to become "Eyefinity". Thats a cool article for anyone to read. Check out the discussions on b3d, or just go enjoy your fav game on your 680, you lucky bastard. Historically ATI was always on the leading edge for engineering milestones, and I could be wrong but i dont see that happenign again. They were the first on every node transitiion at TSMC, they invented dx standards and tess, but now theres GF and TSMC, now theres Tri-gate and 20nm at intel's inhouse. Now theres the end of the road for standard silicon, the problems with tscm 40nm, 28nm and Nvidia's slides asking for new ways to earn money in lithography relationsihp with tsmc. Things are changing a lot. Nvidia could have their own in-house Tri-gate 20nm fab making SOC smartphone tegra shit in the next year or two. These are some reasons i thought the 7970 to be the last great AMD gpu. Gk110 better, and maybe the last big nvidia thing, untill the maxwell beast arrives. The environment is changing a lot around whats gonig on in the GPU landscape.
Nvidia markets better, we all knwo this. They have PAX, GTC, PDX, blah blah blah, and tons morer where that came from. They have "way its meant to be played" and a lot of other cool things that are really ntoable from a technicaly perspective.
My favs. That new ambient occlu in all the games from driver panel, that super quiet fan, and very compatible NVCP, fxaa, smaa, txaa, 20 different aa modes, 3d vision, and the glasses, even that stacked pci-e power connector thingy, Surround (copied from eyefinity) and now Kepler uArch with regular driver updates. I think Jensen was suprised himself when he said at the conference call that "kepler was tyhe best architecture they every made" It actually is really frickign good. The other good thing is that gpu boost that we're arguing about as auto-overclocking in thsi thread. That is a SMART EVOLUTION in graphics card design. They are squeezing the most out of their product which is in almost all ways better for us as the consumers. We win on every level, and the losers are the folks clamoring for ATI to add boost to their next gen 7970, when the reality is theyd ont give a shit and are busy making next-gen console gpus and reforming their company ranks. Just like nvidia is smart at evolving the cards themselves. Like begging for a change in the layout of intels motherboards, and the power limits of pci-sig. But this only lasted long enough until they had a faster card than the competition. Now who gives a crap about an inverted case where the HSF is ABOVE the gpu, and you can safely shoot 800 watts onto the video card. Nvidia makes Better designed coolers that allow for intake of air in a sensible way, Aerodynamic fan shroud. Nvidia card designs always looked like a ferrari to me, while amd's cards look like Mack dumptrucks. then if you were to compare Cypress to Fermi, Cypress is a ferrari while Fermi is a dump truck. Then if you compare Kepler to GCN, they're both high end ferraris, while kepler is a little faster and higher quality limited edition "Orange Lambo" if you will. But the cards, 7970 is still a dump truck and the 680 is an abosolule ferrari.
Now if you look at the PCBs... hehe 680 is a golf cart/moped/kia, and 7970 is a rolls royce, thats the only way i can safely put that, without trying to hurt someones feelings here. 680 owns in many metrics, but that PWM is not one of them. The power delivery is weaksauce, try and debate me on it, anyone im begging you to make a sensible argument about that power design. PLEASE!
Kingpins' billy-badass 680.5 board is a perfect example. He soldered a video card ON TO A VIDEO CARD for crying out loud. Yeah it set the record, but is it even a 680 anymore?... I mean jesus. LOL. It's a friggin space ship, not a graphics card.
Okay 680/7970 war, one more time (stay focused man) The 680 wins... But is it that easy? The 7970 has the world record. Andre Yang and Peter beat Vince and his alien 680 space ship technology, but only because Vince couldnt put 4 gtx 680 space ships on one Motherboard. With just that one sample at the uber 2Gz, he beat Andre's previous world record (which Vince argues is a cheated record anyway because the of the tessleation giving andre an extra 1-2k points). But hey Andre is a convicted cheater any from years past, and he was only last year unbanned from the XS community. But his cheating was because he knew how to get high scores and he would share them with others for money, not because he was faking the score he is a LEGIT bencher. Monety- Thats why he got banned. EVGA and Vince getting that 680 score above Andres 1.7g and Elmors 1.8G 7970's was a Press Statement more than anything. He said, let us inject this gk104 board with a friggin lightning bolt and see what happens *although you knwo behind the scenes he was begging Nvidia for a GK110 Engineering sample* but they said no, fok you You cant have a sample because if you set some super high retarded score, you're gonna want to show everyone, and there will be a massive riot cuz everyone wants that kind of performance. Yes folks, Kepler is a beast.
The very fact that he was able to beat a 7970 with a gk104 should tell the tale, regardless of how weak the 680 reference board is. And the 680 reference board's pwm is not all that "WEAK" per se really since it does its job and makes people happy by not exploding. Its not weak in that sense, its more like the pwm is "just enough to get by" with some little bit of overclockign headroom to boot. It OCs some, but they still stop you at 1.2v for the F of it. Make room for the primo models, like the Classified, doesnt hurt my feelings.
In the end if I had both a 7970 and a 680, somebody may wonder, what would this lunatic do with these two cards?
Well I would overvolt both of them. I would first off want the 4GB classified with the extra power and 14phase pwm for the 680, and i would go with a dual x sapphire 7970 or a watercooled lightning or something. I would probably put them on water blocks and oc the fark out of them and would want to sample many cards of each, since I know overclockablity differs by sample. But sadly I have neither, and my opinion is invalidated. Ride it out on my 6850 while i play borderlands at medium res, med qual...
you got 4 ppl
guy 1: My 680 ocs like shit, only does 1095 on baseclock
guy 2: my 7970 ocs like crap, only does 1080 on baseclock
newbie3: woah guys my 680 must be good, runnin fine at 1320/1800
newbie4: check my 7970, its at 1375/1950.
The factors at play are #1 VARYING SAMPLES. Somebody whos really good at overclocking could get a sample that stops them at 1080mhz - MURPHYS LAW. Someone who sucks and is newbie at overclockign could get a sample that is golden, best every seen, rarest, and be stuck at 1050 because he cant figure out how to unlock CCC limits or something dumb, or doesnt evne know how to OC in the first place. Like "oh i run stock" and that guys card might be the 1400mhz jewel diamond in the rough.
So you have to knwo what you're doing and push it to the limits, but you also have to get a good sample in the first place. Thats why those guys have 20 GTX 680s or 20 HD7970s in the same box, cuz they're testing them all for the rarest finest sample they can find.
Nvidia is winning right now. The strategy is great $499 680 GK104 (which they could only produce 200 of them before TSMC problems ramping big lots) Props to anyone who got one early. 256bit bus, and it still rocks the competition. Tahiti is good, and going by that OC review from Hard, they can both OC really high, but in the end, the OC 680 still bests the OC 7970. HOW-EVER the WORLD RECORDs are still held by the 7970. Not sure how to approach this, but it's definitly not as clear a win for as say 4870 versus 280 or 5870 versus 480 or 6970 versus 580. In those generations, the X80 beat the X870 eight times outa 9, not 50/50. Even when you overclocked both to their limits, the Nvidia top end still won, and they set all the HWBOT world records, this is the real MAJOR difference.
The 680 is not doing that. They could sell it at 460/560 prices and there would be no difference on the company spreadsheets. The last minute OC and name change was genius. The oem bios + 300.63 driver clocked it at 705mhz, the finalized + 300.99 at 1006/1058. It is engineered to be a high-selling, high-margin product that will be enjoyed by the masses and covers all bases. It overclocks, it does 3d, surround, plays all the games. Its a good product. It's not the Gx1x0, it is the Gx1x4. This says a LOT. It says Kepler is radical, game-changing and billy-badass in general. Kepler is the FTW graphics technology of the future, and is the nail in the coffin for the cursed GCN. I would love to have a 680, and thats part of the reason I hate the damn thing so much! Call me jealous or whateva/ I dont care. Its a wonderful product. It is a tad expensive, but you can put some of that on AMD for ripping us off on the 7 series (compared to their awesomely priced 3870/4870/5870/6970 that is). If Nvidia wants they could play this poker game and not show their all their cards... WHAT? NO WAY... GK110 is what AMD better watch out for. It's going to be a death blow for sure.
And the 20-25% rumors are because
1, gk104 is clocked higher than normal,
2, 110 will have compute centric built in because its being duplicated to Quadro, Tesla workstations and HPC,
3, the rumoured ratio of specs for 110/104 is not as big a gap in Kepler as it was in Fermi,
4, the probably wont have to make 110 a 1ghz sku since there wont be much comp.
Kepler was the big Nail in the coffin. Now time will tell the story of next-gen consoles, and GK110/maxwell/Sea islands.
Props if you read this. head hurts now. gg.
RussianSensation
Elite Member
- Sep 5, 2003
- 19,458
- 765
- 126
:hmm:
So what you are saying is because you owned a bunch of ATI/AMD cards in the past, it's logical to put a maximum overclocked 7970 against a stock 680 and proclaim that it 'crushes' a 680? We should expect that to be the case, especially since GTX680 is a mid-high end Kepler part and not the highest GK110 part. It's rather amazing that a re-badged GTX670Ti with 1536SPs, 256-bit memory bus (70GB/sec lower memory bandwidth), smaller die, and lower power consumption isn't that much worse than a 38% manually overclocked 7970.
Shouldn't we wait for aftermarket 680s though for a fair comparison?
Just because you are a fan of AMD cards doesn't mean you shouldn't be objective. I should point out all of the advantages AMD had during HD4870/5870/6970 series no longer apply:
- Not a leader in price/performance
- Not a leader in performance/watt
- Not a leader in performance/die
So if you were a fan of AMD cards during the last several generations based on the reasons mentioned above, GTX680 is clearly a superior card.
The question then is are you a fan of AMD or a fan of what AMD's graphics division used to stand for?
We would hope so considering that HD7970 has a 38% manual overclock (1280 from 925) and costs $580.
No it's not. It's an overclocked vs. stock review. An overclocked 680 would mean manually overclocking it from stock speeds.
Using your logic, putting an FX8150 with a 38% overclock against a stock Core i7 2600K is overclocked vs. overclocked since 2600K has Turbo Boost at stock speeds. :sneaky:
Also, using your logic, all Intel's Core i5 and i7 mobile processors are "overclocked". How does that make any sense?
Core i7 3920XM is rated at 2.9GHz start clock and end of the turbo line at 3.8GHz for a single core, 3.7GHz for two cores and 3.6GHz for all four cores.
Factory GPU Boost = CPU Boost. The only difference is that GPU Boost doesn't operate on a per core basis, but on a larger variety of variables. Whatever performance boost GTX680 delivers from the factory is stock speeds. It's reasonable to accept a 1-3% variation among cards but to imply that reference GTX680's magically boost to 1200mhz in every game is nonsense.
Ok but if you wanted to tests both cards at their absolute maximum operating frequencies, you would have needed to adjust they power tune and the manual GPU offset for GTX680. Thus, GTX680 cannot be considered to have been operating at its reasonably maximum potential, while HD7970 in this review was cranked to 1280mhz @ 1.3V!
Crushing a reference 680?
- $80 more expensive
- Required a 38% 1280mhz overclock @ 1.3V on air, one of the highest we've ever seen from an air cooled 7970 vs. a stock 680 that was not manually overclocked
- Consumed on average 540W in games vs. 435W, despite hardly being much faster
What happens when the 680 is manually overclocked?
What happens when unlocked BIOS/voltages of aftermarket 680s will allow them to scale beyond 1.2V?
What happens when an aftermarket 680 with a beefier cooler and 8-pin x 2 connectors is put to the test with its maximum overclock?
I don't know why it's so hard for people to accept that at current prices HD7970 is simply overpriced unless you are running some specific program (GPGPU dual precision tasks) that benefits significantly from a 7970. For current 7970 users there is no need to side-upgrade to a 680, but for new buyers, if they can find a 680 for sale, it's the better card in almost every way imaginable.
If the card in this review was $499, then it could be argued that's it's a good alternative.
Last edited:
Zanovar
Diamond Member
- Jan 21, 2011
- 3,446
- 232
- 106
FFs,do you expect anybotdy to read that?,ffs man,shame as you probably made good points in there somewhere.holy fook
Last edited:
RussianSensation
Elite Member
- Sep 5, 2003
- 19,458
- 765
- 126
That's going to be a sweet card but will it cost $499? Probably $649.
I find it remarkable that when HD6950 @ 6970 for $250 and even HD6970 $370 were recommended last generation against a $499 GTX580 (that itself had a 20% overclocking headroom that went largely ignored on these forums), this generation people are willing to pay more for a card that needs a huge overclock to beat a reference $499 card. It's like the bang-for-the-buck philosophy has left the forums this generation.
The HD7970's amazing overclocking characteristics are not enough to justify the $80-100 premiums that aftermarket versions have considering a manual overclock on the reference 680 will easily match an overclocked 7970, while consuming less power and saving $ in the process.
I think HD7970 + high overclock is a great card, but sold at a price that only makes sense as a result of GTX680 shortages. Furthermore, in light of $450+ HD7950s, GTX680's value becomes even more clear.
Last edited:
Zanovar
Diamond Member
- Jan 21, 2011
- 3,446
- 232
- 106
Subyman
Moderator <br> VC&G Forum
- Mar 18, 2005
- 7,876
- 32
- 86
People, you are debating terminology. Overclock has always meant "increasing the clocks higher than they are capable of in stock form" in my book. If this is the case, then the nvidia is not "overclocking" itself. It is varying the clocks within an engineered range. To overclock would be to push it farther than it is possible for the card to clock in stock form.
It is impossible for a stock card to "overclock" itself since it was engineered to clock at those speeds, hence it isn't "overclocked" by definition. This is precisely why nvidia calls it a boost and not an overclock.
It is impossible for a stock card to "overclock" itself since it was engineered to clock at those speeds, hence it isn't "overclocked" by definition. This is precisely why nvidia calls it a boost and not an overclock.
Dont even know why this is news. We already know this, 7970s at 1.2ghz or above will beat a "stock" gtx680. Heck it will equal an OC gtx680. If you can get 7970s to 1.3ghz it will be a bit faster OC gtx680.. but its using over 100W more than the gtx680.
That last part is why its not attractive and gtx680 is still the better product.
That last part is why its not attractive and gtx680 is still the better product.
I'd likely buy a 7970, if I were in the market for a top end card, but if I were a gambler, I'd bet that card's going to be $700. Maybe more. It is an awesome looking card though. :thumbsup:
That coming from you is funny.
What about overclocking itself?
Quite the opposite
Hypocrite much? http://forums.anandtech.com/showthread.php?t=2238013
Edit: Not having a shot at you - just saying
OK, let me hit you with some hard facts, The GTX is $30 cheaper in the USA whilst having a poor PCB and is short of 1GB VRAM (wait & see how much the 4GB is LOL)
In U.K & Europe the hd7970 is $50 cheaper and is widely available, if you look at it from a global perspective, you will realize that the 7970 is the better buy as its cheaper and most importantly a better built card (extra VRAM to boot)
Last edited:
Keysplayr
Elite Member
- Jan 16, 2003
- 21,219
- 56
- 91
That coming from you is funny.
What about overclocking itself?
Quite the opposite
Hypocrite much? http://forums.anandtech.com/showthread.php?t=2238013
Edit: Not having a shot at you - just saying
OK, let me hit you with some hard facts, The GTX is $30 cheaper in the USA whilst having a poor PCB and is short of 1GB VRAM (wait & see how much the 4GB is LOL)
In U.K & Europe the hd7970 is $50 cheaper and is widely available, if you look at it from a global perspective, you will realize that the 7970 is the better buy as its cheaper and most importantly a better built card (extra VRAM to boot)
Here are some hard facts for you to my friend.
I see Gigabyte and HIS now offering 7970s for 529.00 now. All others are 549.00 or higher. Those are the only two vendors I see that reflect your claim of the GTX being 30.00 cheaper in the USA. Any reference 7970 should be an additional 50.00 cheaper than that and that is being generous.
Next, I've noticed many fans are trying to perpetuate the "poor quality component/PCB" rumor and maybe it'll stick routine for the GTX680. Where is your evidence, man? Show us why the GTX680 PCB is "poor". I don't see anyone having any trouble with them.
Next, I've also noticed the GPGPU naysayers of yesterday now embrace GPGPU as a very, very important feature in a GPU. Haven't you?
Next. And finally. If you're overclocking the non reference 7970 to it's absolute max on air, then so should a non reference GTX680 be overclocked to it's absolute max on air and THEN compared. None of this "Look how a massively overclocked non reference 7970 crushes a stock reference GTX680" nonsense. See the logic, or don't. I don't care. Remember how ballistic fans went when Nvidia released overclocked GTX460's to be benched against stock competitors cards? Ba-lis-tic! But dems da facts.
Last edited:
Average ref 7970 OC on air is 1.2ghz.
1.2ghz 7970 is faster than most gtx680 OC already or matching the ones going at 1.3ghz. It's not slower, its just using heaps more power for a marginal perf difference thats not noticeable. +/- 5% is not going to be visible.
1.2ghz 7970 is faster than most gtx680 OC already or matching the ones going at 1.3ghz. It's not slower, its just using heaps more power for a marginal perf difference thats not noticeable. +/- 5% is not going to be visible.
Last edited:
I can go over 1.2GHz, but the noise... egads, the noise...
I think I was spoiled by my Sapphire Vapor-X 5870's and Sapphire Toxic 4870 before it. I didn't even know my video cards had fans. My 7970 is very acceptable up to 1125Mhz and gets a little loud but still 'ok' with some voltage and 1.2GHz. But, beyond around that point, I don't care if this thing could hit 2GHz, no way I would do anything to make any more noise than it already does.
Keysplayr
Elite Member
- Jan 16, 2003
- 21,219
- 56
- 91
You should show us. Got some reference links to show a 1.2GHz 7970 matching a 1.3GHz GTX680? Or a 1.2GHz 7970 faster than an o/c'd GTX680?
I know full well you mean to say that out of the box, you consider the GTX680 to be overclocked, but we have to keep it real. Out of the box is out of the box as per Intel turbo'd CPU's are.
So, what you really said above is, 1.2ghz 7970 is faster than most gtx680 stock already.... etc. etc.
Keep in mind that you'll have to eat these words when AMD's next gen debuts with it's own version of GPU boost. Oh yes, it will happen.
Have you even seen that PCB? Its power delivery? The whole PCB just screams cost cutting. It has to be heavily modded just to OC it on LN2 by gluing additional power delivery, you can do that with stock 7970. It's just barely adequate for air cooling. Maybe for you it's fine just as the PCB of GTX590 but other people have higher standards. It's a fact that its PCB is poor it is not up for a discussion. What's up for a discussion is if it really matters. For air cooling I don't think it does but it still leaves a bad taste in my mouth considering it's a 500$ card.
Last edited:
There's already plenty of links to show OC vs OC reviews and benches (max OC 7970 beat a +135% gtx680), you can dig them out of the OLD review thread yourself. This is beating a dead horse.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 24K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter