Mr. K6's 6950 -> 7970 Overclocking Review

[MrK6]

I have a vested interest since I'm running dual overclocked 6950's.

What voltages are you running, and did you have any stability problems?

How long were your benchmarks (for instance, in skyrim?)

Once I'm through with the number crunching (which is my favorite, so I'm doing it first ), I'll throw up a detailed rundown of my system/specs and the benchmarks, including settings, design, etc. I have screenies of all the settings and data that I'll eventually like to throw up, but it will take a bit. To answer your questions, my 6950 runs at 1.3V core, perfectly stable. It'll actually do 1030MHz at that vCore, but these are my 24/7 clocks that I tested at and I insist they be very stable. The custom benchmarks are Fraps runs lasting 60 seconds, these are Skyrim, the Witcher 2, and BF3.

 

[n7]

Thanx for sharing.

I'm a bit surprised that the overclocked 6950 improves less @ 2560x1600 than it does @ 1080p.

 

[MrK6]

Thanx for sharing.

I'm a bit surprised that the overclocked 6950 improves less @ 2560x1600 than it does @ 1080p.

Glad to, and I was surprised by that as well. But I think that indicates that there's something else inhibiting performance at that resolution beyond simple speed. Taking a quick look at benches like Crysis: Warhead and Crysis 2, I'm betting I hit a vRAM wall. The 3GB 7970 will help shed some light on this.

 

[Troll Trolling]

Have you tested how much performance the shader unlocking only gains?
And how much the energy consumption changes?

 

[MrK6]

Have you tested how much performance the shader unlocking only gains?
And how much the energy consumption changes?

Roughly 3-5%, depending on the application. I tested this in my 5850 -> 6950 review (in my sig). The energy consumption change from unlocking is undetectable/within the margin of error.

 

[BallaTheFeared]

Where did you get the Official Benchmark for The Witcher 2?

 

[MrK6]

Where did you get the Official Benchmark for The Witcher 2?

There is none. I thought I remember seeing a program similar to Adrenaline/Frost for the Crysis series, but I couldn't find it. I used my own custom benchmark instead, but this has not been reflected in the first post (although it is in the spreadsheet).

 

[BallaTheFeared]

I can't even describe the level of disappointment you've brought me.

I had such high expectations, the only way it could go was down though


Need benchmark tool noaw!!

 

[MrK6]

I can't even describe the level of disappointment you've brought me.

I had such high expectations, the only way it could go was down though


Need benchmark tool noaw!!

Lol, sorry. The cut scene I use to benchmark is right in the beginning of the game where you first talk to Foltest. I think I'll throw up videos of each of my custom benchmarks so people can exactly see the scenes I'm using.

 

[MrK6]

OK, I added some commentary on the results in the second post. More to come as I finish the analysis. One thing to point out is a changed the criteria of the performance/W indices to only include gaming benchmarks. I think 3DMark11 and Unigine Heaven put an unrealistic load on the system in that they only load the GPU's, so the numbers are artificially low.

 

[Stuka87]

6950 OC Results and Commentary:

BF3 was another game I was disappointed with. I played multiplayer for about 3 weeks and never even cracked up the single player, and I'm glad I didn't waste my time. This game is consolized crap and I can't believe I wasted money pre-ordering it (at least I got Karkand for free, if I ever play it). Everything from the "Press Start" main menu to unskippable intro movies and no quicksave screams that this was a console port and DICE was talking out of both sides of their mouths. They're not getting another pre-order from me, and definitely not $60.[/ranting]

Ok, I am confused here. What "Press Start" button and intro movies are you talking about? When I play, I go to server, click the server I want, and go straight into the game. I don't even see an intro logo of any sort. Unless of course you are talking about the single player. And even then I don't recall any press start screen. But who has ever bought a BF game for the single player mode? The game is designed for multi-player. And personally, I don't find the multi-player to be consoleized at all.

 

[badb0y]

I think he messed up between Crysis 2 and BF3

 

[MrK6]

Ok, I am confused here. What "Press Start" button and intro movies are you talking about? When I play, I go to server, click the server I want, and go straight into the game. I don't even see an intro logo of any sort. Unless of course you are talking about the single player. And even then I don't recall any press start screen. But who has ever bought a BF game for the single player mode? The game is designed for multi-player. And personally, I don't find the multi-player to be consoleized at all.

You must have missed all the hype and marketing that focused solely on the single player. In fact, at the two expos I went to, they showed and talked about nothing but the single player. Hell, PAX East had 20 minutes of single player, that's it. And the multiplayer is a step backward from what BF2 and 1942 to were. Is it as bad as CoD? No, but how low do you want to set that bar.

I think he messed up between Crysis 2 and BF3

Nah, I knew Crysis 2 was a lost cause, but BF3 surprised me.

In either case, I'd like to redirect the thread back on topic; I won't mention the games' content again.

 

[MrK6]

I'm hopefully going to get a minute today to detail the benchmarking suites as well as post some videos of runs through my custom benchmarks. Keep checking the spreadsheet for updates and I'll restrict major updates to reserved posts.

 

[MrK6]

After a near miss and half a heart attack, I'll have an XFX 7970 at my doorstep tomorrow. I'll try to get done as much as I can, at least stock clock tests, a quick air cooling overclock, and I'll start assembling my loop. We'll see :thumbsup:

 

[Madcatatlas]

Reminds me of keysplars 480 review,! Looking forward to seeing the final 'product'

 

[Granseth]

Glad to, and I was surprised by that as well. But I think that indicates that there's something else inhibiting performance at that resolution beyond simple speed. Taking a quick look at benches like Crysis: Warhead and Crysis 2, I'm betting I hit a vRAM wall. The 3GB 7970 will help shed some light on this.

If you think you're hitting a vRAM wall you could run a test on the 6950 where you just test if there is a linear (or close to) change in performance when you change the memory frequency.

 

[MrK6]

Reminds me of keysplars 480 review,! Looking forward to seeing the final 'product'

I hope to deliver :thumbsup:

If you think you're hitting a vRAM wall you could run a test on the 6950 where you just test if there is a linear (or close to) change in performance when you change the memory frequency.

That's assuming that the access speeds can overcome a capacity insufficiency, which I don't believe is how it works (someone correct me if I'm wrong).

I got her installed and benching @ stock on air,

Yes, I know my tubing looks disgusting. The flash makes it look worse than it really is and I'll be changing it out tomorrow. The reason for tomorrow is my new XSPC RayStorm for my CPU has not arrived yet, so I won't be taking apart my loop yet. That's actually a good thing I think, since I'll be using the same loop for my 7970 tests, which will keep things very consistent. I'll also be able to run some quick tests again to see how well the new blocks (and clean out) up the performance of my loop. I'll report back with some initial air cooling data for those interested, and then I'll immediately throw it on water. I think it's only fair for the card to be on water for both the stock and OC tests in order to maintain consistency.

For first impressions - the card is the same size as the 6950, which is easily manageable in a midtower. The fan itself is amazing, AMD outdid themselves here. At idle, it stays at 20% and is silent. The XFX card only comes with a DVI->HDMI adapter, so if you're an Eyefinity user, the Sapphire model is the way to go for now. Also, I have to put in a plug for NCIXUS - awesome company. This XFX arrived at my door in Boston, MA from CA in less than 24 hours. Their service, especially customer service, is top notch. If you're looking for an alternative to Newegg or Amazon, especially for big ticket items, they're the best. More to come .

 

[Cookie Monster]

Itching to hear back about the cooling capabilities of this card and the sound tone of the fan during load.

Thanks for the sharing!

 

[Termie]

Congrats. Can't wait to see your detailed review.

One thing I noticed with NCIX over the holidays, since I was listing a lot of their deals in Hot Deals, was that they'd go out of stock really quickly on discounted items, or more accurately, go out of stock at that price. I thought that was a bit deceptive, as in some cases they clearly were continuing to sell the product at a higher price than their specials list indicated. Glad things worked out for you, though...then again, this wasn't a discounted item.

Have fun benching!

 

[MrK6]

Some data on Air cooling for those curious -

Stock Air, No OC
Idle: 42C, 20% fan
Highest Load in Crysis 2 Bench: 83C, 49% fan
Idle Power: 90W (from wall)
Peak Power Load in Crysis 2 Bench: 412W
3DMark11 P Score at stock: 7861 (51.4% higher than stock 6950, 24.2% higher than 6950 OC)
Crysis 2 2560x1600 4xAA: 37.2FPS (79% higher than stock 6950, 57.0% higher than 6950 OC)

Wow, just wow. This card is a MONSTER in every metric. The load fan noise is excellent, it's just a ton of wooshing air. I'm spoiled by my near silent H2O setup, but as far as reference designs goes, this is the best yet. I think part of it may have to do with AMD really opening up the back of the card to allow better airflow (restriction = noise). The power consumption is ungodly, but so is the performance. I'm excited to see what the analysis concludes after I finish the benches and crunch some numbers. I'm going to push an air OC now and report back shortly, then it's H2O time :twisted:.

 

[MrK6]

Itching to hear back about the cooling capabilities of this card and the sound tone of the fan during load.

Thanks for the sharing!

Will do, and I'll throw up a quick video of the card while benching so you can hear it for yourselves (although my camera is a point-and-shoot, so there's your caveat ).

Congrats. Can't wait to see your detailed review.

One thing I noticed with NCIX over the holidays, since I was listing a lot of their deals in Hot Deals, was that they'd go out of stock really quickly on discounted items, or more accurately, go out of stock at that price. I thought that was a bit deceptive, as in some cases they clearly were continuing to sell the product at a higher price than their specials list indicated. Glad things worked out for you, though...then again, this wasn't a discounted item.

Have fun benching!

Yes, they have a very dynamic stocking system which can burn you if you don't pay attention. Basically, if something goes out of stock, they'll raise the price for back orders. However, their customer service is excellent and if you call them up they'll re-credit you with no issue if you do get caught in a price hike. This happened on my Acer 3820TG and I got all $48 back, including the extra in the conversion of Canadian to USD.

 

[Arzachel]

Some data on Air cooling for those curious -

Stock Air, No OC
Idle: 42C, 20% fan
Highest Load in Crysis 2 Bench: 83C, 49% fan
Idle Power: 90W (from wall)
Peak Power Load in Crysis 2 Bench: 412W
3DMark11 P Score at stock: 7861 (51.4% higher than stock 6950, 24.2% higher than 6950 OC)
Crysis 2 2560x1600 4xAA: 37.2FPS (79% higher than stock 6950, 57.0% higher than 6950 OC)

Wow, just wow. This card is a MONSTER in every metric. The load fan noise is excellent, it's just a ton of wooshing air. I'm spoiled by my near silent H2O setup, but as far as reference designs goes, this is the best yet. I think part of it may have to do with AMD really opening up the back of the card to allow better airflow (restriction = noise). The power consumption is ungodly, but so is the performance. I'm excited to see what the analysis concludes after I finish the benches and crunch some numbers. I'm going to push an air OC now and report back shortly, then it's H2O time :twisted:.

That power draw is from the wall, right? If not

 

[MrK6]

That power draw is from the wall, right? If not

Lol, yeah, from the wall. Factor it by .8 since I'm on a 4 1/2 year old Corsair 620HX. However, seeing how overclocking goes, I might be getting a new PSU sooner than later .

 

[Termie]

Lol, yeah, from the wall. Factor it by .8 since I'm on a 4 1/2 year old Corsair 620HX. However, seeing how overclocking goes, I might be getting a new PSU sooner than later .

You're also pushing a huge overclock on that 2500k, which must require some serious volts. That being said, even HardOCP's huge 1.25Ghz overclock on their 7970 only increased power draw by 100w, pulling a total of 462w on a 2600k@4.8: http://www.hardocp.com/article/2012/01/09/amd_radeon_hd_7970_overclocking_performance_review/6. Your stock power draw is about 40w higher, though.

Figuring you hit around the same OC on your 7970, you'd be at 412w x 0.8 + 100w = 430w load on the PSU, which is still only 70% of your rated capacity. I'm not an expert on split-rail PSUs, but that's where you might run into trouble. You have three 12v rails with 18a each: http://www.newegg.com/Product/Produc...82E16817139002