GPU Value Comparisons

Overclocking doesn't impact the power draw, but overvolting does:
Unlike overclocking without overvolting, which saw fairly negligible increases in power consumption and thermal output, upping the GPU voltage, even by such a small amount as 60 mV, saw the power consumption rocket. When overclocked and overvolted to its maximum stable level, the MSI caused our test system to draw 343W from the wall; an increase of 20 per cent over the result using a standard GTX 560 Ti 1GB.
Click to expand...
Link
From a couple of other places I've also seen that volt tweaking it has a disproportional impact on draw.
 
Overclocking usually increases power draw linearly , while overvolting increases it exponentially.
 
Archer said:
Overclocking usually increases power draw linearly , while overvolting increases it exponentially.
Click to expand...

Yes, understood.

cerebus said:
Overclocking doesn't impact the power draw, but overvolting does:
Link
From a couple of other places I've also seen that volt tweaking it has a disproportional impact on draw.
Click to expand...

Of course it's going to consume more power.. I agreed to as much. The 460 does the same thing.

I'm not seeing this massive power consumption you are talking about (yes they used the word rocket, but you're ignoring the reality of the situation), in that very same review you linked to, it has the OC pwrcons at 343W vs the 570 at 330W (13W more Load), yet at Idle where it's going to sit at for most of it's life it's 135W vs the 570 at 150W (15W less at Idle).

And at those speeds it seems to be beating the 570 (in the link you provided) by a proportionate enough margin. So... I don't see a negative here?

And dude.. that MSI you linked to runs so cool n quiet, it's friggin sweet..
 
Last edited:
I'm just saying there is a drawback for overvolting the card - not a huge one, but it's there. Voltage tweaks always have a higher toll, but in this case it's higher than normal. To equal the 570 requires more volts than a 570 by itself; idle watts are neither here nor there.

And dude.. that MSI you linked to runs so cool n quiet, it's friggin sweet..
Click to expand...
I know right...I'm lus for one of these now :p
 
cerebus said:
I'm just saying there is a drawback for overvolting the card - not a huge one, but it's there. Voltage tweaks always have a higher toll, but in this case it's higher than normal. To equal the 570 requires more volts than a 570 by itself; idle watts are neither here nor there.


I know right...I'm lus for one of these now :p
Click to expand...

Totally!

However, I'd say Idle IS here AND there ;), since Idle consumption savings are as much if not slightly more than the marginal loss (over 570) at load (added bonus that it beats it). Plus the 560's get round 950 without voltage bumps, which lands you probably at or JUST below a stock 570.

But I'm beating the point to death now ;)
 
Last edited:
TopCa7 said:
Indeed, was also looking at that. Got a mate in USA atm and a cousin that lives there and is coming down next month. Seriously thinking about it :P

Yeah theres some popping up all over hay man.
Woot, landmark have them at around R2600-3000 depending on model and if its SOC etc

http://www.landmarkpc.co.za/store/advanced_search_result.php?keywords=gtx+560&x=0&y=0
Click to expand...

Ah I see Landmark raised the prices of the EVGA cards and placed the Club3D (R2647.27 ) at where the cheapest EVGA model was sitting. It was R2625.95 now R2725.95.

Good thing those reference cards are all exactly the same anyway!

6870/6950 1GB are now totally irrelevant.
 
Last edited:
Archer said:
That completely wrong. Finished updating, you'll see how close all three cards are... and you should also see that the 6950 is the one to go for (at current prices)
Click to expand...

Are you counting 1gb or 2gb 6950s? USD prices are around $300 for the latter, $270 for the former. GTX560 TI is 5% slower on average than 1gb 6950 and $20 cheaper, so it's parity or better than 6950 (not including the ability to flash a 6950 or the 560's redoubtable overclocking prowess). Local pricing for the 6950 1gb seems to be around r3000 while the 560 is overpriced at r2650, but I think it'll come down to r2500 before too long. But there is also the architecture and cooling to take into consideration, where 560 is hands down a better designed card.
 
Last edited:
I take the cheapest prices of all cards
So HD6950 2GB = R2850 in my charts, should actually drop this to R2800
GTX560Ti = R2650

Using that the 6950 is the better choice imo (more fps, while remaining better value for money)
 
Archer said:
I take the cheapest prices of all cards
So HD6950 2GB = R2850 in my charts, should actually drop this to R2800
GTX560Ti = R2650

Using that the 6950 is the better choice imo (more fps, while remaining better value for money)
Click to expand...

6950 2gb is r2749 at Woot. 560 is still overpriced locally though so I think prices will come down in the next couple of weeks as availability trickles through.
 
Archer: could you please include the prices as text on the Cost chart the next time you update it?

I've also posted that you can get the Inno3D GTX 470 for under R2400.
 
The other issue with the chart methodology is it doesn't take into account performance levels at different resolutions with AA and so forth - which is fine in a rough-and-ready sort of way, but for instance a r1800 card won't be expected to perform at 2560, and a r5k card will never be played under 1080 resolution, and preferably in 2560 8xaa cases. So architectural factors like bus width will cater for expected resolutions, and a much more expensive card when put into the context of its target market will appear to have a much higher value than if it's spread across lower settings where it appears disproportionately pricey.
 
Last edited:
I'll add in text prices when I have time
Not going to be able to do different resolutions/quality settings since not all sites do that, and I dont have a test bench :/ The results I take are usually 1920x1080 and at max settings. Thats what 99% of people aim for, so I feel its a good enough guide. If you want a higher/lower resolution you're going to have to extrapolate a little bit.
 
Have you thought of using 3dmark11 performance/extreme test as your yardstick as the settings are locked down for each?
I use their sites result search quite often http://3dmark.com/search. You could do a search using the same CPU with all the graphics cards in your graph.
You would have to filter out the OC'd cards/CPU's but you bound to find 100's of results of machines tested on stock settings with a common CPU.
There is also a handy histogram showing the most common scores for each arrangement.

Use it, don't use it.
 
doobyscoo said:
Have you thought of using 3dmark11 performance/extreme test as your yardstick as the settings are locked down for each?
I use their sites result search quite often http://3dmark.com/search. You could do a search using the same CPU with all the graphics cards in your graph.
You would have to filter out the OC'd cards/CPU's but you bound to find 100's of results of machines tested on stock settings with a common CPU.
There is also a handy histogram showing the most common scores for each arrangement.

Use it, don't use it.
Click to expand...

3DMark has never illustrated the variation of real-world gaming results accurately. You can have a card that gets more 3dmarks than another but is generally outperformed in actual games.

3DMark points = E-Peen inches.
 
It works fine to just spread gaming results at 1080. There's nothing really wrong with it. It's just worth pointing out that the upper and lower end cards aim for different gaming scenarios.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X