OFFICIAL KEPLER "GTX680" Reviews

[gwolfman]

It has a link to a full tweaktown review and PDF in case it goes down.

 

[blastingcap]

Hexus.net

Yes that's the one!

It has a link to a full tweaktown review and PDF in case it goes down.

Thanks but I wish it were a legit site, not TT, which doesn't even test BF3, Skyrim, etc. NV skew? They had a terrible conclusion back in the day as well for the GTS 450--so obviously biased for NV that I swore I would never take them seriously again. The comments section on that article in their forums were absolutely livid, too. http://forums.tweaktown.com/video-c...e-bias-wards-nvidia-latest-gts450-review.html

Edit: TT's power chart seems to indicate that power draw trails the 7970, especially in idle. Is that a typo?! Unexpected given the -1 GB VRAM difference.

Edit 2: Hmm, it seems that the GTX 680 gets slower and slower relative to the 7970 when you either a) go up in resolution, or b) add AA or AF. And since most people are running at higher resolutions with AA and AF, that's not good news.

In Metro for instance it is merely on par with the 7970 at 1080p and actually a tiny bit slower at 1600p, assuming these slides are real. http://www.tweaktown.com/reviews/46...reference_card_video_card_review/index11.html The same pattern holds for other games too, e.g., Dirt 3. 26% faster at 1200p shrinks to about 20.x% at 1600p.

TT is also either incompetent or hiding something when they have the high AA/AF tests on only a few select games: http://www.tweaktown.com/reviews/46...reference_card_video_card_review/index14.html I mean wtf, show us BF3 with high AA and AF!

So apparently once you add in AA/AF the 7970 doesn't look as bad. It even beats the 680 on some resolution and AA/AF combinations: http://www.tweaktown.com/reviews/46...reference_card_video_card_review/index14.html (63 vs 53 fps for AvP! It was only 96 v 91 before. So the 7970's lead went from 5.5% to 19%... and it may be worse as you go up to 1600p instead of the 1200p they tested at) Meanwhile at 1200p, the 680's lead in Mafia II shrinks to 84/81 = 3.7% when you add AA/AF. Without AA/AF it's 135/118 = 14.4%

But this is TT we're talking about, home of the NV skew (historically, regardless of what is going on now), so I would prefer to wait for [H] or some other more reputable site's review.

 

[gwolfman]

Did you check this one out:
http://forums.anandtech.com/showthread.php?p=33177271 ???

 

[Nintendesert]

Where's BF3 on that review.


And Skyrim?

 

[gwolfman]

Where's BF3 on that review.


And Skyrim?

Nope and nope. :*(

 

[3DVagabond]

Why do we always have to go through this BS. [H] tends to be a pretty polarizing site, for some reason. Everytime someone says they like their reviews someone has to say they are total sh1t. It's a very good review site. It offers a unique perspective of the equipment. TPU is usefull too, but W1zzard uses heaps of outdated games and, usually, outdated drivers. He's usually not afraid though to pump some volts into the cards and see how far they will go. Very useful power numbers though. We could go on about the different sites. In the end, if one truly was difinitive, then, "there would only be one".

This gets even more tiring than nVidia vs. AMD.

 

[Grooveriding]

Crummy review. Poor choice of games, and most importantly, no overclocking.

 

[Silverforce11]

TT?

Hawx2 - check
LP2 - check
JC2 - check
Mafia2 - check
Metro? - check
Farcry2? - check

I think the only "recent" TWIMTBP game they are missing is Batman. Why not include Batman? FAIL.

wait... FARCRY 2???????!! For a moment there i thought we're in 2012..

 

[Imouto]

Crummy review. Poor choice of games, and most importantly, no overclocking.

I'm with you here. OC wasn't in THG's leaked benchs either. WTF?

 

[gwolfman]

Any idea when this is supposed to go on sale? I was expecting midnight Pacific time and I'm getting sleepy.

 

[Silverforce11]

Wait... why is it using ~60W more than a 7970?

I thought this thing was meant to be lower TDP... unless.. no.. could it be? NV is bullshitting TDP again?!

Gtx680 = Idle: 206, Load: 456
7970 = Idle: 127, Load: 389

There's something wrong, it can't be using ~80W more on idle..

Edit: It smells to me like NV went balls to wall and really did clock to 1006mhz from their "original" ~700 default causing a massive increase in power use. There goes their TURBO boost feature, getting another 50mhz only, it would have been awesome if it was 700 - > 1050mhz turbo.

 

[blastingcap]

I'm bracing myself for disappointment if the TT charts are accurate. It seems that the 680's edge over the 7970 shrinks as you climb in resolution, or if you add AA/AF. It looks to be ~10% faster at 1200p without AA/AF but stumbles at 1200p w/ AA/AF, let alone 1600p w/ AA/AF. It may be that the 256 bit memory bus is coming back to haunt it. And then there's the power draw which seems unimpressive. 206 vs 127 watts idle? 456 vs 389 load? Even if the PSU weren't that efficient, that's still a major difference in the wrong direction. Seriously? Is TT wrong or is THG is wrong? They are complete opposites as far as perf-per-watt.

 

[Imouto]

Edit: It smells to me like NV went balls to wall and really did clock to 1006mhz from their "original" ~700 default causing a massive increase in power use. There goes their TURBO boost feature, getting another 50mhz only, it would have been awesome if it was 700 - > 1050mhz turbo.

It stinks actually.

 

[gwolfman]

I hope TT's wrong, esp since they rate it at 300wTDP, right? Btw, what the max theoretical TDP for 2x 6-pin PCI-e power?

 

[MisterMac]

This is NOT good for my eu market prices gpu's.


Where's the HAM NVIDIA, WHERES THE HAM.

7950/7870 will hopefully drop prices :/

 

[blastingcap]

I hope TT's wrong, esp since they rate it at 300wTDP, right? Btw, what the max theoretical TDP for 2x 6-pin PCI-e power?

225 watts but TT is showing system watts and not the card's watts.

 

[gwolfman]

Why isn't it for sale yet? :*(

 

[amenx]

I dont think Nvidia would have touted the power savings abilities of Kep if the results were like the TT review. I'll bet its an anomaly that does not show up in other reviews.

 

[Silverforce11]

I dont think Nvidia would have touted the power savings abilities of Kep if the results were like the TT review. I'll bet its an anomaly that does not show up in other reviews.

I hope so, because that's freaken ridiculous for such a small die on 28nm, its sucking nearly as much power as a gtx580 with worse idle.

Bring on the OC vs OC performance!

 

[notty22]

Heres the conclusion from TT.
And I've never heard the memory bus spec being mentioned so much until AMD finally went bigger there. Now it seems it's the panacea for hi res performance. lol

Final Thoughts


From our introduction it's obvious our personal feelings towards NVIDIA. With that said, though, it wouldn't take away from us looking at a product objectively and the GTX 680 is just an awesome video card. The performance really is quite amazing and you can see that at times it has no issues competing with the dual GPU Radeon HD 6990 video card from AMD. While that might be a card based on the last generation core from AMD, it's still considered the fastest on the market.

Outside of the performance, though, one of the biggest issues for AMD is going to be the price of the card. So often we've seen AMD just behind NVIDIA for performance, but come in at a more aggressive price point. It's always been this kind of argument where if you want to save some money, but still get excellent performance you buy the AMD offering. If you just want the fastest, you buy the NVIDIA offering.

Again due to the fact that NVIDIA sent us nothing we don't know officially the price the card will launch at. Saying that partners have told us around the $549 mark and a leak from Newegg sees closer to the $500 mark depending on the brand. Considering the performance here today, this is a bad thing for AMD's Radeon HD 7970 at $549.

 

[Silverforce11]

"The performance really is quite amazing and you can see that at times it has no issues competing with the dual GPU Radeon HD 6990 video card from AMD."

It helps when 3/4 of the bench suite are TWIMTBP games. I can see how they would conclude as such..

 

[MisterMac]

It's totally bandwith starved from my viewpoint.
That's the reason it fails @ high resolutions.

It'd make sense to release this for 300 usd with a 705 mhz clock and that bandwith bus. - high oc headroom, but ultimately starved and youd have to get a 670/680 gk110/110 to really push.

That's so obviously the original plan but now that amd ball dropped, nvidia decided "Hey, if they can screw over people, so can we! YAY!".


They just oc'ed the [thing] and laugh @ amd until the big brother comes.

[Love] you nvidia :/

[Love] you AMD too :/


I'm with team Blue from now on. -_-

You're on vacation from now on. We've made it clear time and time again that you can't use profanity here
-ViRGE

 

[Ieat]

"The performance really is quite amazing and you can see that at times it has no issues competing with the dual GPU Radeon HD 6990 video card from AMD."

It helps when 3/4 of the bench suite are TWIMTBP games. I can see how they would conclude as such..

One review and you skimmed through the info faster then flys on s#%@ to attack nvidia and discredit the reviewing website. I guess mission accomplished. lol Next time you might want to actually read the first page of the review to find out they are actually at odds with nvidia before putting the tin foil hat on. sigh...

 

[Silverforce11]

One review and you skimmed through the info faster then flys on s#%@ to attack nvidia and discredit the reviewing website. I guess mission accomplished. lol Next time you might want to actually read the first page of the review to find out they are actually at odds with nvidia before putting the tin foil hat on. sigh...

So you read their review?

What's your conclusion, do you agree with their conclusion?

Do you think it's good that an "unbias" reviewer uses 3/4 of their benches products designed by TWIMTBP? What would you say if reviewers started to use mostly AMD Evolved games only? You would conclude as any reasonable person would, there's obvious bias.

 

[blastingcap]

One review and you skimmed through the info faster then flys on s#%@ to attack nvidia and discredit the reviewing website. I guess mission accomplished. lol Next time you might want to actually read the first page of the review to find out they are actually at odds with nvidia before putting the tin foil hat on. sigh...

Regardless of whatever they are claiming now, TT is historically known as pro-NV. See, e.g., their GTS 450 review: http://www.xtremesystems.org/forums...e-GTS-450-Cyclone-1GB-Review-Fishy-conclusion

And TT toes the line when it comes to using NV-approved benches.