How fast are your 5770 overclocks?

[Dark4ng3l]

I just got my MSI hawk 5770 and of course I had to see how fast it could go. Running furmark for 20 minutes at stock speeds (the core is overclocked to 875 out of the box) the card topped out at about 72c. Pretty good I don't think thermals are going to be a huge problem overclocking this card.

So I loaded up MSI afterburner and after some fiddling I managed to find the limit for the memory is pretty much 1400 MHz. At 1400 MHz I had no problems with the card but anything higher and it would crash after 5 seconds in furmark.

The stock voltage on my card is higher than the reference boards(1.2v) and I can get the core to about 950 on that. At 1.225 I can get 975 but it takes a full 1.3v to reach 1 GHz. I haven't tried to push the voltage to the max supported 1.35v or to push the core past 1GHz at all.

So I guess my current top is pretty much 1GHz core and 1400 memory. Just to be safe i'm running it at 975/1350.

I am kind of disappointed in my card because some people have been getting up to 1.1GHz on the core and close to 1500 memory. I know I still have a bit of headroom on the core if I go to 1.35v( maybe I can get it into the 1035 range or something) But even stock it's pretty damn fast compared to the 8500GT that I used as a temporary solution. Now I can actually use a sniper rifle in BF:BC2 without it being an exercise in futility. I'm just wondering how this measures up with what other people have been getting on their cards.

 

[happy medium]

I don't have a 5770 ,but my 5750 will run furmark and bench at 915,1375 at stock voltage.
I wish I could overvolt it, I bet she would do 1000 core.

Most 5770's I've seen are doing 950 core 1400 memory at stock voltage.
I don't think the memory is overvolted when you increase the voltage.

 

[Attic]

With no voltage increase on my Sapphire V2 (Egg Cooler) 5770 I can bench as high as 950/1400, but in games I'm only stable at 920/1350. I say only, but i'm still happy with the O/C.

Recently I was having issues with BFBC2 and I lowered to 900/1300 and had no more issues with the game. So that's where it's at now 24/7.

Performance i'm happy with, and the card is very very quiet with the egg cooler, silent from 8ft. It's in a HTPC so that's important. Only noise I have is from a couple 80mm fans behind my CPU that I had to use when my low RPM ones started making a racket.

 

[saratoga172]

I got mine at stock voltage from 850 -> 950 with ATI overdrive. I think mem went to about 1345ish. Didn't really notice much of an fps increase or anything and I can already max out pretty much everything at 1080p so I just left it at stock.

I'll oc it in a year or two when I really need the power... Or revert back to my 19" 1440x900 monitor. Either way I'm a happy camper.

 

[poohbear]

i overclocked mine to 920ish and the memory to about 1400, but really it provided a 2-5% increase in FPS, which is pretty pathetic. Just leaving it at stock as overclocking these guys doesnt really yield much of a performance increase. I've read that maybe its the 128bit?

 

[dug777]

i overclocked mine to 920ish and the memory to about 1400, but really it provided a 2-5% increase in FPS, which is pretty pathetic. Just leaving it at stock as overclocking these guys doesnt really yield much of a performance increase. I've read that maybe its the 128bit?

It certainly makes the argument that the 5770s are memory bandwidth limited look a little odd if going from 1200MHz quad pumped to 1400MHz quad pumped (a 16.6% increase) has nothing like that impact on framerates

As I understood it, memory bandwidth is entirely a function of the bus width and the memory speed. Of itself, I can't quite see how bus width would create a limiting factor in the face of increasing memory speed, unless the GPU memory controller areas simply can't handle more than a certain bandwidth...

Remember that from a stock of 850 an overlock of 920 is only an 8% increase to start with as well, so if the 5770 is more core limited (as BFG's testing suggests), then you would be expecting at best close to an 8% improvement in FPS under ideal conditions (the 5% you mention would be reasonable). That isn't much: (32.4 fps vs 30 fps, or 54 vs 50fps, 75.6fps vs 70fps).



EDIT: FYI xbitlabs tests an oc'd 5770, and the results generally seem to line up with that sort of an increase, their conclusion stating that The average performance growth of the Radeon HD 5770 overclocked to 940MHz GPU and 1445 (5870) MHz memory frequencies is about 12%, reaching 14% in some applications...just a little better than the core increase of around 11% (noting they use an i7 testbed to your x3).

http://www.xbitlabs.com/articles/video/display/radeon-hd5770-hd5750_17.html#sect0

 

[v8envy]

Realize that the 5770 does rudimentary checksumming of memory and re-tries on error. Raising the memory clock can actually result in a loss of performance. This is one overclock where you have to check not only for stability but also performance increase.

 

[hans030390]

Realize that the 5770 does rudimentary checksumming of memory and re-tries on error. Raising the memory clock can actually result in a loss of performance. This is one overclock where you have to check not only for stability but also performance increase.

This is very true. When overclocked, keep note of the performance increase. Next, raise your voltage. Test again. If your results increase, your card is actually NOT stable at that OC and needs to have its voltage raised.

Eventually you'll reach a point where raising the voltage no longer increases performance, assuming the card is having these memory check errors. This is when the card is considered stable and has reached its maximum performance given a certain OC.

 

[dug777]

This is very true. When overclocked, keep note of the performance increase. Next, raise your voltage. Test again. If your results increase, your card is actually NOT stable at that OC and needs to have its voltage raised.

Eventually you'll reach a point where raising the voltage no longer increases performance, assuming the card is having these memory check errors. This is when the card is considered stable and has reached its maximum performance given a certain OC.

It would be interesting to see people compare their fps in at least one benchmark (at a sensibly GPU limited setting of course) for stock, just the GPU oc, just the mem oc, and then both.

You should see straight away whether you are getting a performance decrease running a stock GPU, and it always casts interesting insights into the balance of a card...

 

[hans030390]

It would be interesting to see people compare their fps in at least one benchmark (at a sensibly GPU limited setting of course) for stock, just the GPU oc, just the mem oc, and then both.

You should see straight away whether you are getting a performance decrease running a stock GPU, and it always casts interesting insights into the balance of a card...

I thought that someone already did that on here...

 

[dug777]

I thought that someone already did that on here...

They did (BFG, bless his heart!)

It would be nice to see a little more rigour from the 'casual' overclocker as well however, especially given how quickly and easily you could run that set of benchmarks, and it would let you know what was actually delivering 'value' in terms of your overclock.

 

[hans030390]

They did (BFG, bless his heart!)

It would be nice to see a little more rigour from the 'casual' overclocker as well however, especially given how quickly and easily you could run that set of benchmarks, and it would let you know what was actually delivering 'value' in terms of your overclock.

You name a benchmark you want done, and I'll do it.

 

[Dark4ng3l]

Well after some more extensive tweaking I lowered the memory to 1360 (I had a memory related failure after about 7-8 hours of playing with the card at 1375 memory) But my card still hits a wall at 1GHz pretty much. I can't even get it stable at 1010 at 1.35v and I need 1.3v to make it work at 1GHz.

But it's really weird how the card hits a wall at about 970(1.2v) to 1000(1.3v) to about 1007(1.35v). I set fan control on manual and put some more aggressive settings on it so the card tops out at about 71-72c @ 1GHz(1.3v) then I upped the clock speed and it was more stable at 1010 with 1.3v than I was able to get it with 1.35v and auto fan settings.

The fans on my Hawk are pretty good. Up to 70% you can't hear it at all over the other fans. 75% you can hear it but it's bearable. 80% is annoying, 90% dustbuster GeForce FX5800 ultra level and 100% is really over the top noisy. With the fans on auto even at 77-78 the fans would stay under 70%. When I forced them to go to 75% when the temp hits 65 I managed to lower the load temps by about 6-7c.

Another factor is possibly my really crappy PSU could be failing to provide good voltage under load and this would limit it's overclocking capability. Either that or I have a below average card. I'll probably be getting a new case/psu some time this year so I can do additional testing then but right now i"m fairly content at 1GHz/1360 memory (though I would have loved > 1400 memory as the cards biggest problem is bandwith)

 

[jtisgeek]

what are you guys doing for the overclock error that puts the core at 157 when not in use. I use 2 24 inch lcd's and they go crazy when it under clocks?

Just got my 5770 back and wanting to play with it.

 

[solofly]

935/1300 rock solid!

 

[Attic]

what are you guys doing for the overclock error that puts the core at 157 when not in use. I use 2 24 inch lcd's and they go crazy when it under clocks?

Just got my 5770 back and wanting to play with it.

Do you have this issue if you don't overclock? For single monitor 157 is a standard underclock, and works fine.

On my 4830, if I overclock it I cannot use two monitors due to issues, if I leave it at stock i can run two monitors just fine.

Not overclocking isn't the dream solution, but I think that should work.

 

[jtisgeek]

Do you have this issue if you don't overclock? For single monitor 157 is a standard underclock, and works fine.

On my 4830, if I overclock it I cannot use two monitors due to issues, if I leave it at stock i can run two monitors just fine.

Not overclocking isn't the dream solution, but I think that should work.

No issues when I over clock just that the 157 gives me flickering on my 3 screens. This is still a error in the drivers my card runs at 400 core when not overclocked overclocking it still messes up the default clocks.

There is some work around from the older drivers but makes overclocking a pain .

 

[Dark4ng3l]

Have you tried using MSI afterburner? My card never goes below 400MHz when I overclock it. And you can easily set it back to default speeds and then to your overclocked seting only when you need it using the program.

 

[jtisgeek]

Yeah does it with afterburner also. Yeah I know I can just change it when I go to play games but it's a pain.

 

[Dark4ng3l]

Just set up a profile at stock and an overclocked profile then go in options and set hotkeys for them ( I use ctrl+alt+numpad1, etc) and you can switch between them in 1 second from inside your game or whatever without even touching the program.

 

[jtisgeek]

Just set up a profile at stock and an overclocked profile then go in options and set hotkeys for them ( I use ctrl+alt+numpad1, etc) and you can switch between them in 1 second from inside your game or whatever without even touching the program.

Good idea I'll look in to it.