Grudge Match From Hell : HD 6970 Vs GTX 560

[happy medium]

Yessir.

Why would it not be lol.

Ok just for the sake of asking , how many volts? Can it reach 1000 core?

 

[badb0y]

I didn't wanna mess around with the voltage to find the lowest stable one so I shot it at 1.3, 1000 core is not my max OC though I just used it for the sake of comparison :thumbsup: .

 

[Grooveriding]

Way to miss the point that was stated quite clearly in a bunch of posts. I own two EVGA 580s superclocked, want a photo?

Some of us care that faulty reasoning is being employed to try to cast a card in a positive light - that's it. Plenty of the Nvidia-friendly posts seem to highlight Nvidia cards and overclocking while failing to realize that AMD cards also overclock. Is that news to you, or something?

Well said. I own three expensive nv gpus, and it's not just for the heck of it.

It's the armchair fanboys who degenerate every thread into nv vs amd or create threads to that end using screwball logic and narrow examples for no clear purpose beyond what looks suspiciously like malicious obsession with trumpeting a dogma.

If they are least owned the hardware it would be something.

 

[happy medium]

I didn't wanna mess around with the voltage to find the lowest stable one so I shot it at 1.3, 1000 core is not my max OC though I just used it for the sake of comparison :thumbsup: .

Is there a program out there that you can use to see how much more wattage you use? I saw a review with a 6950 at 1010 core at 1.3v using 300 watts. I can't find it now, me and Castiel we talking about it.

I'm fairly sure a Gigabyte gtx560 at 1000 core uses 220 watts or mabe it was 230.

 

[badb0y]

Is there a program out there that you can use to see how much more wattage you use? I saw a review with a 6950 at 1010 core at 1.3v using 300 watts. I can't find it now, me and Castiel we talking about it.

If you find one let me know.

 

[Spike]

While I'm obviously on an AMD vid card kick at the moment (don't know how that happened) I don't see a lot wrong with this review. All the person did was show that you could overclock a cheaper card enough to reach the 6970 level of performance, at least at certain res/detail combinations. Does that make someone like me, who paid $380 for my 6970, feel bad? No, it does not. I wanted to pay for a card that could handle everything I have at 1920x1200 with most or all the details turned up and it turns out, I got that to my satisfaction.

Now, for someone with a smaller res monitor and more budget constraints suddenly this 560 looks like a good deal, and it is. If your willing to OC and possibly limit the life of the card and/or deal with more noise then more power to you. This way you get a great card for a lot less than the big boys (6950, 570, 6970, 580). The same thing could be said for the 6850/6870 cards. OC the crap out of them and they perform VERY well yet people don't get up in arms when someone does that and compares it to a higher priced card.

I used to always pay in the $200-250 range for my vid cards, oc the crap out of them, and be happy as a clam. As I've gotten a little older, and my income is a fair bit higher, I'm more willing to pay extra cash for a card that can perform the way I want at stock with the possibility of going higher with an OC later, to a greater or lesser extent. The way people are talking in this thread you would think they would be advocating that all 6950/570/6970/580 owners are dumb since similar (or even greater!) performance can be had by cheaper cards that are overclocked like mad.

Anyway, the 560 is shaping up to be a very good performer for the price. It's safe to say that anyone gaming at 1650 or 1080P have MANY great, and quite cheap, choices right now. I still can't believe how low the 460's and 5850's are. If I was running 1650 I would probably pick up one of those and hold it for a year until the next wave of cards is out.

 

[MentalIlness]

I noticed there are some still in denial that the gtx 460 ftw was a retail card.

 

[bryanW1995]

At that resolution, with these cards, I'd be using 8xAA or perhaps SSAA.

SSAA in most games, maybe even 4x in some. It's actually quite fun dragging out your older games when you get a new video card, seeing ridiculous eye candy on 16x10 is a blast!

Having said that, 16x10 is the absolute sweet spot for a 560. with that huge OC at that res it's probably >/= gtx 570 as well, right?

 

[bryanW1995]

@1920x1080 the gtx580 is 17% faster then a gtx570 on average.
Is 17% way faster? With video cards so close in performance today, that is dam close to way faster.

yeah, but that's why the 560 is so desirable at lower res, right? The higher core clock is more likely to have a huge impact than the weaker memory bandwidth. OC the 570 and 580 to the max at 1920x1080 and you'd probably see them much closer to parity. gtx 580, much like 69x0, only shines above the lesser cards at the very high res/AA levels that we all crave.

 

[cusideabelincoln]

SSAA in most games, maybe even 4x in some. It's actually quite fun dragging out your older games when you get a new video card, seeing ridiculous eye candy on 16x10 is a blast!

I love playing older games with high AA (so clear, I hate jaggies) and 60+ framerate. This is why I think you can't really have an overpowered card, as far as resolution goes. There's always something a card can't handle.

For a few more dollars (13$?) you can just buy a gigabyte SOC @ 1000 core garaunteed and it runs much more cool and quiet and uses much less power.

Stop spreading misinformation. Most data will point you to the opposite conclusion: The 6950 uses less power.

I'll point you to this post: http://forums.anandtech.com/showpost.php?p=31139526&postcount=244

According to TPU, clock for clock the 6950 looks like it's using less power than the 560. Not to mention 90% of the reviews will put a stock 6950 using less power than a 560.

 

[thilanliyan]

I can get it up to 1500 easy.

I just didn't because it doesn't have that big of an impact, BRB testing.

Hmm...I thought the memory not clocking high was the reason some people were killing their 6950s when unlocked. Err...maybe it was the memory timings.

The 6950s running at 1250mem are at the rated speed of the chips (5Gbps) aren't they? Also remember GDDR5 has some sort of error correction that underclocks itself even if you think you're running much higher.

 

[happy medium]

Stop spreading misinformation. Most data will point you to the opposite conclusion: The 6950 uses less power

Find me a factory overclocked 6950@ 1000 core that uses 195 watts and I'll stop spreading the so called mis-information

Here is a gtx560 @ 1000 core using 195 watts and 25 watts idle.
http://www.guru3d.com/article/gigabyte-gtx-560-ti-soc-review/7

 

[badb0y]

Hmm...I thought the memory not clocking high was the reason some people were killing their 6950s when unlocked. Err...maybe it was the memory timings.

The 6950s running at 1250mem are at the rated speed of the chips (5Gbps) aren't they? Also remember GDDR5 has some sort of error correction that underclocks itself even if you think you're running much higher.

It was the memory timings and the voltage that killed the memory in 6950s flashed to 6970s BIOS.

I went from P5696 to P5750~ just from OCing memory from 1250 to 1400.

@happy_medium, a 6950 will also perform better when it's clocked at 1000 core though right?

 

[Elfear]

Hmm...I thought the memory not clocking high was the reason some people were killing their 6950s when unlocked. Err...maybe it was the memory timings.

The 6950s running at 1250mem are at the rated speed of the chips (5Gbps) aren't they? Also remember GDDR5 has some sort of error correction that underclocks itself even if you think you're running much higher.

It was the timings I've read that have been giving people issues.

Both my 6950's hit 1500Mhz on the memory but now that you reminded me of the error correction, I'll have to run some tests and see if performance drops off. Has the error correction been confirmed with the 6000 series?

 

[cusideabelincoln]

Find me a factory overclocked 6950@ 1000 core that uses 195 watts and I'll stop spreading the so called mis-information

Here is a gtx560 @ 1000 core using 195 watts and 25 watts idle.
http://www.guru3d.com/article/gigaby...i-soc-review/7

You have to be joking. I just showed you a graph of both a 560 and 6950 and how much power they use per clockspeed per voltage increase.

Also it would be incredibly pointless to compare any results I find from another site to the ones you are looking at from Guru3D. Guru3D has a different testing methodology that they do not release to the public, as they state right there on that very page. Since different workloads stress video cards differently, it would be incredibly pointless to compare results of two different cards from one website to another.

Also that is Guru3D's estimated TDP. It isn't the actual measured value for the card itself. They are basing that number off of data they collected by measuring power consumption at the wall, and these numbers are affected by the efficiency curve of the power supply. Since the efficiency of the PSU will play heavily into the equation, I only find their summative chart at the end of that page useful for relative performance positioning and not for hard data. It would be better if their chart just tracked total system power consumption using identical systems. If they are using different system for their tests then that will change results. Another reason they advertise estimated is because they artifically "add" idle power draw, but they don't measure this idle power draw.

So to recap...

1. 195W is ESTIMATED - it's an educated guess.
2. They use a different testing methodology than other sites.

The only way to get a valid comparison is for Guru3D to test their own 6950 under the same conditions they test other cards. They actually do test the 6950, but it's at stock, and it uses less power than even the stock 560. So I'll reiterate... stop spreading misinformaton (or rather, uninformation? uknownformation?). Now it is your job to actually produce valid data instead of using logical fallacies. There isn't any data validating your power consumption claim, so how in the world can you claim to not stop spreading the FUD? Not only does your logic not making any sense here, but you asserting a very limited context for these cards to be in. So limited it ignores so many factors, which I'll describe below...

@happy_medium, a 6950 will also perform better when it's clocked at 1000 core though right?

And this is also a factor. A 6950 at 1000 MHz will outperform a 560 at 1000 MHz, considering a 6950 at 800 MHz already outperforms a 560 at 820 MHz. Thanks to the re-balancing of the architecture, the 6000 cards have good megahertz scaling. Also the memory bandwidth (and speeds) will affect both performance and power consumption.

So your statement ["Find me a factory overclocked 6950@ 1000 core that uses 195 watts"] really isn't that simple, happy. It doesn't account for memory bandwidth nor performance. And like I've already mentioned, it is not valid to compare the findings from one site to another in a manner as to determine how the products in question relate to one another. In other words, it's like comparing apples to oranges.

 

[happy medium]

You have to be joking. I just showed you a graph of both a 560 and 6950 and how much power they use per clockspeed per voltage increase.

Also it would be incredibly pointless to compare any results I find from another site to the ones you are looking at from Guru3D. Guru3D has a different testing methodology that they do not release to the public, as they state right there on that very page. Since different workloads stress video cards differently, it would be incredibly pointless to compare results of two different cards from one website to another.

Also that is Guru3D's estimated TDP. It isn't the actual measured value for the card itself. They are basing that number off of data they collected by measuring power consumption at the wall, and these numbers are affected by the efficiency curve of the power supply.

So to recap...

1. 195W is ESTIMATED - it's an educated guess.
2. They use a different testing methodology than other sites.

The only way to get a valid comparison is for Guru3D to test their own 6950 under the same conditions they test other cards. They actually do test the 6950, but it's at stock, and it uses less power than even the stock 560. So I'll reiterate... stop spreading misinformaton (or rather, uninformation? uknownformation?).

No , I stand by what I said.

TDP on guru's site for gtx560 @ 1000 core = 195
TDP on GURU's site for 6950 @ 800 core = 158

The 6950 will not only use 37 watts to get to 1000 core with voltage it needs to get there.
The gtx560 SOC can do 1044 core at factory STOCK VOLTAGE according to GURU's site.
Common sense tells you that the specially binned 13$ more gtx 560 SOC will use less power then a similarly clocked 6950 @ 1000 core.

 

[cusideabelincoln]

No , I stand by what I said.

You have absolutely no ground to stand on. That doesn't make you look good. And... stealth edits. Read them.

 

[thilanliyan]

No , I stand by what I said.

Just from the Guru3D link you gave, the stock 6950 is using ~40w less than the 560 SOC (and those are "estimates" by the way). Considering a 1GHz core 6950 is faster than a 1GHz 560, judging by power consumption is hard to do. You would have to equalize the performance, then check the power consumption which I don't think anybody has done.

Here's some more numbers:
http://www.tweaktown.com/reviews/3796/gigabyte_geforce_gtx_560_ti_1gb_soc_video_card/index18.html
The SOC and a 6950 at 970 use almost the same power, but at that speed the 6950 would probably be faster anyway.

Here's more:
http://www.overclockersclub.com/reviews/nvidia_asus_gtx560ti/16.htm
The 6950at around 900MHz uses less power than 2 560s at 1GHz core and performs about the same.

 

[happy medium]

Just from the Guru3D link you gave, the stock 6950 is using ~40w less than the 560 SOC (and those are "estimates" by the way). Considering a 1GHz core 6950 is faster than a 1GHz 560, judging by power consumption is hard to do. You would have to equalize the performance, then check the power consumption which I don't think anybody has done.

Here's some more numbers:
http://www.tweaktown.com/reviews/3796/gigabyte_geforce_gtx_560_ti_1gb_soc_video_card/index18.html
The SOC and a 6950 at 970 use almost the same power, but at that speed the 6950 would probably be faster anyway.

Thats a good link , now thats a 9 watt difference from 970 core on the 6950 to the gtx560's 1000 core. Add another notch of voltage and up the clocks 30 more on the core for the 6950 and wa-laa you have more then 9 watts EASY.

Thanks for making my point. So much for the "spreading false information" ha? I rest my case.

Apology accepted Casideabeulincoln

Edit: we were not talking about which card was faster. At least I wasn't. I really don't know. My arguement started at post #112.


This is an example of needless rhetoric which is inflammatory and leads to escalating unproductive discussion.

Keep your points technical, this "apology accepted" hyperbole is not adding value to your post nor to the thread.

Moderator Idontcare

 

[cusideabelincoln]

Thats a good link , now thats a 9 watt difference from 970 core on the 6950 to the gtx560's 1000 core. Add another notch of voltage and up the clocks 30 more on the core for the 6950 and wa-laa you have more then 9 watts EASY.

Thanks for making my point. I rest my case.

LOL, if you're point is purely conjecture and changing/ignoring goalposts. You're backpeddling on performance here...

Edit: we were not talking about which card was faster. At least I wasn't. I really don't know. My arguement started at post #112.

No, this argument started in the 560 reviews thread you stated a 560 had better performance per watt when overclocked.

My point was you can max both the 6950 and gtx560 's clocks and the gtx560 will come out ahead in every category, price ,performance, noise, temps.

That just isn't true. If the 560 ramps up clockspeed to beat the 6950 it will suffer in the power department. If the 560 can't clock high enough the 6950 will beat it in the performance department and could possibly beat it in the power department too.

Also we have to keep in mind the Gigabyte SOC chips are cherry-picked. They are the best chips and will have the best performance/watt characteristics. Other 560s aren't going to behave the same when overclocked as the Gigabyte SOC does. The Gigabyte chips are impressive, but other 560s won't be quite so. As such it is in the best interest to relegate this totally awesome power consumption at 1000 core to the Gigabyte SOC cards only. It doesn't apply to other 560s.

It's also impressive non-binned/random 6950s have such great performance/watt, too.

 

[bakalu]

This has to be a joke...OCed the GTX 560 so much yet left the 6970 practically stock?

While the comparison between a stock HD6970 and an overclocked GTX 560 does seem unfair,

2. The GTX560 is way overclocked compared to the 6970 which has a mere 10Mhz. (Oops!)
No doubt the 560 is a good chip, so there is no need for the results to be swayed like this.

Comparing a stock card to one that is overclocked 200mhz is pretty stupid.

At a res that no one buying a 6970 would play? No OC on the 6970. Cherry pick a couple benches. It's crap.

Sorry my english

Why is not fair. Compare performance. At stock clock

NVIDIA : GTX 560 Ti < GTX 570 < GTX 480 < GTX 580
AMD : HD 6870 < HD 5870 < HD 6950 < HD 6970

GTX 560 Ti @ 1015/1075 Vs HD 6970 @ 890/1375 : Fourth Vs First class

if you say GTX 560 Ti @ 1015/1075 GTX Vs HD 6970 @ 890/1375 is not fair.

What do you think about ?

HD 6870 @ 1000/1120 Vs GTX 580 default : Fourth Vs First class

fair or unfair

I just overclocked HD 6870 @ 1000 because of the HD 6870's CCC

I benchmark HD 6870 @ 1000/1120 at resolution 1680*1050
I benchmark GTX 580 at resolution 1920*1080

Fair, right ?

System benchmark with GTX 580:
CPU Core i7 Extreme 965 @ 3.6GHz
Mainboard ASUS Rampage II Gene
RAM 3x2GB DDR3
ASUS GeForce GTX 580 782/1564/1002 @ 772/1544/1002
Windows 7 Ultimate 64bit RTM
Driver NVIDIA 260.99 beta
PhysX_9.10.0514_SystemSoftware

System benchmark with 6870:
CPU Core i7 Extreme 980X @ 3.6GHz
Mainboard ASUS Rampage II Gene
RAM 3x2GB DDR3
ASUS Radeon HD 6870 DirectCU 915/1050 @ 1000/1120
Windows 7 Ultimate 64bit RTM
Driver AMD 11.1a hotfix
PhysX_9.10.0514_SystemSoftware

Fair, right ?

3DMark Vantage Performance

ASUS GeForce GTX 580 @ default

ASUS Radeon HD 6870 DirectCU 915/1050 @ 1000/1120

Benchmark DX11 H.A.W.X.2.

Settings of H.A.W.X.2

ASUS GeForce GTX 580 @ default & 1920*1080

ASUS Radeon HD 6870 DirectCU 915/1050 @ 1000/1120 & 1680*1050

Demo benchmark Dirt 2 DX11. AA=4X, Ultra Settings. POST PROCESS=HIGH

ASUS GeForce GTX 580 @ default & 1920*1080

ASUS Radeon HD 6870 DirectCU 915/1050 @ 1000/1120 & 1680*1050

 

[bakalu]

Benchmark game DX11 Lost Planet 2.

Settings of Lost Planet 2

Test A Results

ASUS GeForce GTX 580 @ default & 1920*1080

ASUS Radeon HD 6870 DirectCU 915/1050 @ 1000/1120 & 1680*1050

Test B Results

ASUS GeForce GTX 580 @ default & 1920*1080

ASUS Radeon HD 6870 DirectCU 915/1050 @ 1000/1120 & 1680*1050

Heaven Demo benchmark DX11. AA=4X, AF=16X, Tessellation=Extreme, Max Settings

ASUS GeForce GTX 580 @ default & 1920*1080

ASUS Radeon HD 6870 DirectCU 915/1050 @ 1000/1120 & 1680*1050

S.T.A.L.K.E.R Call of Pripyat DX11. AA=4X, Max Settings, Tessellation=ON

ASUS GeForce GTX 580 @ default & 1920*1080

ASUS Radeon HD 6870 DirectCU 915/1050 @ 1000/1120 & 1680*1050

 

[bakalu]

Benchmark Crysis Warhead. AA=4X, Enthusiast DX10, ambush map

ASUS GeForce GTX 580 @ default & 1920*1080

ASUS Radeon HD 6870 DirectCU 915/1050 @ 1000/1120 & 1680*1050

Benchmark Metro 2033. I play the first scene. Use Fraps. 1920*1080, AA=4X, AF=16X, Very High, DX11, Tessellation=ON, DOF=OFF, PhysX=ON

Settings of Metro 2033

I begin to benchmark here

And stop here

ASUS GeForce GTX 580 @ default & 1920*1080

2010-11-10 17

46 - metro2033
Frames: 7285 - Time: 221103ms - Avg: 32.948 - Min: 18 - Max: 69

ASUS Radeon HD 6870 DirectCU 915/1050 @ 1000/1120 & 1680*1050

2011-01-29 10:08:22 - metro2033
Frames: 7200 - Time: 228679ms - Avg: 31.485 - Min: 15 - Max: 65

comparing chart

 

[bakalu]

This has to be a joke...OCed the GTX 560 so much yet left the 6970 practically stock?

While the comparison between a stock HD6970 and an overclocked GTX 560 does seem unfair,

2. The GTX560 is way overclocked compared to the 6970 which has a mere 10Mhz. (Oops!)

No doubt the 560 is a good chip, so there is no need for the results to be swayed like this.

Comparing a stock card to one that is overclocked 200mhz is pretty stupid.

At a res that no one buying a 6970 would play? No OC on the 6970. Cherry pick a couple benches. It's crap.

3DMark 11 Performance

ASUS GeForce GTX 580 @ default

ASUS Radeon HD 6870 DirectCU 915/1050 @ 1000/1120

Call of Juarez DX10. AA=4X Super-sampling, Max Settings

ASUS GeForce GTX 580 @ default & 1920*1080

ASUS Radeon HD 6870 DirectCU 915/1050 @ 1000/1120 & 1680*1050

Resident Evil 5 DX10. AA=4X, Max Settings

ASUS GeForce GTX 580 @ default & 1920*1080

ASUS Radeon HD 6870 DirectCU 915/1050 @ 1000/1120 & 1680*1050

Why GTX 560 Ti @ 1015/1075 win HD 6970 @ 890/1375 (benchmark with the same resolution 1680*1050)

HD 6870 @ 1000/1120 (benchmark at 1680*1050) loser GTX 580 default (benchmark at 1920*1080)

Because Fermi architecture is wonderful, and Barts / Cayman architecture is too bad.

What more you can say ?

 

[happy medium]

Wow that was a great 1,2,3 and 4th post. hahahahaha
And welcome to the forums Bakalu.