• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

The first "proper" Kepler news Fri 17th Feb?

Unreal 3 engine definitely looks incredible. I'm also unbelievably intrigued by the fact that before it was using 3 580s, and it's now using a single 6XX...

Personally I'll wait until mid-summer anyway before considering a GPU upgrade, and I'll only get one if I can pick up something particularly quiet for an acceptable performance boost over my 480.

kd
 
The code has clearly been optimised since then. Any one who thinks that the 680 is as fast as three 580s is clearly high.

Also it just dawned on me that 3Dmark 11 uses Physx. No wonder they're showing scores for that and not any games.
 
CAT-THE-FIFTH said:
Here is the explanation for the 40% magic BIOS:

http://www.nvnews.net/vbulletin/showpost.php?p=2535096&postcount=100

So know the GTX670TI is 5% behind the HD7970. When flashed to the GTX680 BIOS it was 40% faster than an HD7970.

This means there is a 45% performance difference between the GTX670TI and GTX680,supposedly.
Click to expand...

02.gif
 
So my new improved list about the full fat GK104:
1.)40% Faster than an HD7970
2.)Was 10% before but all of a sudden gained another 30%
3.)3DMark11 score 15% higher than HD7970
4,)Skimmed GK104 has 3DMark11 score 5% lower than HD7970
5.)Consumes less power than an HD7970
6.)Has a smaller die
7.)Has 2GB or 4GB of VRAM
8.)Can use system RAM
9.)Shorter PCB than a GTX570
10.)Is only $299
11.)Has backlit GEFORCE GTX logo on the top of the card
12.)Is the PhysX MASTER
 
Last edited:
CAT-THE-FIFTH said:
This means there is a 45% performance difference between the GTX670TI and GTX680,supposedly.
Click to expand...

I would be extremely surprised if that is the case.

For one, I can't see any GK104 card being that far ahead of the 7970.

For another, if (as they claim, and would I doubt) the 670 and 680 have the same number of active SP, then this boost would need to come entirely from clockspeed. There is simply no way that one part is going to be clocked over 40% higher than the other... this would imply (for example) 670 = 700Mhz, 680 = 1000Mhz+.


I'm straight-up calling BS on this one.
 
Duff-Man said:
I would be extremely surprised if that is the case.

For one, I can't see any GK104 card being that far ahead of the 7970.

For another, if (as they claim, and would I doubt) the 670 and 680 have the same number of active SP, then this boost would need to come entirely from clockspeed. There is simply no way that one part is going to be clocked over 40% higher than the other... this would imply (for example) 670 = 700Mhz, 680 = 1000Mhz+.


I'm straight-up calling BS on this one.
Click to expand...

I agree with Duff-man.

I wouldn't rely on that post with those 'scores' as absolute truth about gtx670/680. For all we know those could be madeup numbers
 
I actually read the link again.

"The same site posted the first pic of GTX 670Ti card, the admin played with the card confirmed that GTX 670Ti at reference clock is 5% behind the performance of Radeon HD 7970 and admin flashed the bios update gave GTX 670Ti 40% performance boost over Radeon HD 7970, it used the same 1536 CUDA cores as GTX 680."

Sorry,I messed up!! Its the GT670TI which is 40% faster than an HD7970 3GB with a new BIOS.

WTF??
 
Last edited:
I would be astounded if the 680 was 40% faster than the 7970. However, if true I would happily buy one and pay £500-600 for it.

Right now I'm just interested to see if the 680 beats the 7970 and if so by how much. If it's only by about 5-10%, it doesn't have much overclocking headroom and it costs a lot more then it's not going to really appeal to me.

I think I'll wait until the Kepler announcement. Is it still due on the 12th? If the reviews are the same day - or due soon after - then I'll check it out and grab one if it offers good value for money. If this is going to drag on or Kepler under-performs I'll probably just pick up a 7970, overclock it and grab a second when they come down in price.
 
CAT-THE-FIFTH said:
I actually read the link again.

"The same site posted the first pic of GTX 670Ti card, the admin played with the card confirmed that GTX 670Ti at reference clock is 5% behind the performance of Radeon HD 7970 and admin flashed the bios update gave GTX 670Ti 40% performance boost over Radeon HD 7970, it used the same 1536 CUDA cores as GTX 680."

Sorry,I messed up!! Its the GT670TI which is 40% faster than an HD7970 3GB with a new BIOS.

WTF??
Click to expand...

WTF indeed. LOL.
 
So my new improved list about the GK104:
1.)Semi-skimmed GTX670TI was 5% slower than HD7970
2.)Magic BIOS flash gives 40% faster speed than HD7970 for GTX670TI
3.)GTX680 was 10% faster than HD7970
4.)Speed of GTX680 with Magic BIOS unknown
5.)3DMark11 score 15% higher than HD7970
4,)Semi-skimmed GK104 has 3DMark11 score 5% lower than HD7970
5.)Consumes less power than an HD7970 and is a 195W card
6.)Has a smaller die
7.)Has 2GB or 4GB of VRAM
8.)Can use system RAM
9.)Shorter PCB than a GTX570
10.)Is only $299
11.)Has backlit GEFORCE GTX logo on the top of the card
12.)Is the PhysX MASTER
13.)Has single card surround
14.)Is the UE4 MASTER
15.)Has the same memory bandwidth as the GTX580
16.)Does some other stuff that I might go WTF at
 
Last edited:
There were some reports earlier this week that the 670/680's only had cores running at ~700MHz. Now imagine if the "magic bios" increases these very low clocks by ~40% to 950-1000MHz (something most 28nm cards should be capable of).

A magic bios is possible if indeed NVidia sent out smoke and mirror beta cards. I am not saying this is fact or even likely, just possible.
 
CAT-THE-FIFTH said:
So my new improved list about the GK104:
1.)Semi-skimmed GTX670TI was 5% slower than HD7970
2.)Magic BIOS flash gives 40% faster speed than HD7970 for GTX670TI
3.)GTX680 was 10% faster than HD7970
4.)Speed of GTX680 with Magic BIOS unknown
5.)3DMark11 score 15% higher than HD7970
4,)Semi-skimmed GK104 has 3DMark11 score 5% lower than HD7970
5.)Consumes less power than an HD7970 and is a 195W card
6.)Has a smaller die
7.)Has 2GB or 4GB of VRAM
8.)Can use system RAM
9.)Shorter PCB than a GTX570
10.)Is only $299
11.)Has backlit GEFORCE GTX logo on the top of the card
12.)Is the PhysX MASTER
13.)Has single card surround
14.)Is the UE4 MASTER
15.)Has the same memory bandwidth as the GTX580
16.)Does some other stuff that I might go WTF at
Click to expand...

**** YEAH!
 
This is getting kind of like the 6900 series speculation. A few hours before reviews hit they were going to be 2048 shader monsters which were 80% faster than GTX480's. A few hours later a lot of people were disappointed by cards that were not much faster than the 6800 series.

I don't expect GK104 to be noticeably faster than a 7970. Hell, I don't care if it is a bit slower just so long as it offers a better VFM proposition than the overpriced 7900's.
 
You must log in or register to reply here.
Back
Top Bottom