if you like yes, but before i ship i woul like o get the exact mh/s per card so no one freaks out at each other...will do this tonight.
You can expect 336 MHash at the stock core speed of 725. if OC is your thing, 800 will bring 371, 850 is 394, and 900 is 417. I'm sure with those numbers you can guess how much mhash you will get from other frequencies too. I ran an exhaustive max mhash per core speed per memory speed to get these numbers. You can look at Sheet 2 to see my data https://docs.google.com/spreadsheet/ccc?key=0AjXdY6gpvmJ4dGxqSmhrektQZ2FVVTNiNUE2RnM0U0E&hl=en_US#gid=1
|
|
|
Not trying to thread crap, but I have that very same sapphire 5870. It lacks a programmable voltage regulator. It does however have 3 voltages to select from: 1.05, 1.10, and 1.20, and it does seems to pick the next highest voltage if you request a lower voltage (0.95v = 1.05v, 1.049v = 1.05v, 1.05v = 1.05v, 1.051v = 1.10v). Just providing information to potential buyers.
Also interested in the 5970's. I offer $265 shipped to 54501, interested in 1 or 2 cards. eBay has been flooded with a lot of 5970's recently..
|
|
|
That is very interesting. I too use 1x width unpowered risers and went to check on my 3 rigs. My 4x5970 rig shows no signs of scorching yet. Neither does my 5x5850 rig. My 5x5870 rig though has slightly discolored 12v wires though (the same ones that are scorched on your picture). I touched it and the wires and the connector are fairly warm to the touch. Almost worrisome!
The power supply connected to the "warm" wires is a PC Power @ Cooling 1KW-SR (83% efficient, not 80plus certified). My other 2 rigs are using a hale90 750 watt and a rosewill lightning 1300 watt, both 80plus gold
|
|
|
I have no cpu bug. 5850, 5870, and 5970's using 11.12 driver, 2.1 sdk, windows 7 x64
|
|
|
Asking if orphan blocks are paid is pretty moot if the pool is paying with PPS anyways. Doesn't matter if a block is 100,000,000 shares and orphans. You were paid per share
|
|
|
So I took off the cooler and looked at the VRM heatsink. Tiny little thing... It looks like there's good contact between the VRMs, the long skinny thermal pad, and the long skinny heatsink. I reseated it, and I'm still getting VRM temps 110+C (tho I really didn't expect it to). I might look into replacing the pads with higher quality stuff. Will this do the trick?Same pads I used for my 5970. I even went the extra step of putting a tiny dot of shin-etsu x23-7783d thermal paste on each VRM chip. They now run significantly cooler than my other 3 "stock" 5970's, upwards of 10c. Bear in mind though that those thermal pads are a little tricky to work with. They're pretty delicate and the stiff plastic sheet that covers one side is difficult to take off without messing up the thermal pad to some degree. FWIW, I got both 1mm and 0.5mm pads to be safe, but you will likely only use 0.5.
|
|
|
That 750 watt PSU will be hardpressed to run 3 7970's. I have one OC'ed to 1075 core & use 370 watts.My system @ idle uses 150 watts (quad core CPU,8 gig RAM,3 HDD's,4 80mm fans). http://www.newegg.com/Product/Product.aspx?Item=N82E16814129264So the 7970 is using 220 watts mining (which @ 610 mh/s @68c w/fan@ 65% is fantastic ).220x3=660+150 for system=810 watts. Good luck Bear in mind that's after inefficiencies. Assuming 90% efficiency, 810 watts at the wall is something like 730 watts which is within the capabilities of the power supply.
|
|
|
is there any way to set the gpu/mem clock and the voltage via some flag/command? cant find a way to overclock under win7 without tools..
Not with phoenix. Phoenix is strictly a miner. No eyecandy or extras. Just a miner.
|
|
|
Have you try re-assembling your heatsinks? It's possible you have bad contact between the VRM and whatever is cooling it. I shaved off something like 15C from both of my VRM's on my one 5970 by using some thermal pads I got from frozencpu (the fuji extreme 11 watts/mk one)
|
|
|
Hi. I have a severe bug. I was using 2.7.1 no problems. I tried 2.7.3 and got this error. restarting cgminer doesn't fix it. it prevents me from mining entirely. I am using a single 5870 on 12.1 driver, 2.1 sdk (platform 1 specified in conf file), dynamic intensity, phatk, worksize 128, and vectors = 2. I reverted back to 2.7.1 and have no problems. windows 7 x64 ultimate
|
|
|
875 on 0.95? That sounds almost impossible! My 5870 got 925 @ 1.00V, and 725 @ 0.95.
I highly doubt it was at 1.00v. It was likely at 1.063v, which is the next voltage step in many 5870 bios's.
|
|
|
I know this is an old thread, but I used the same exact thermal pads in the same exact configuration (0.5mm on everything except the PCI-E bridge chip and the big blocky chips next the the hot VRM's) and my results are pretty much spot on with what you said. The pads are amazing, though a little tricky to work with.
What you said about using thick thermal paste I believe is wrong though. You want the heatsink and the hot gpu core to be as close as possible. If it's clogged with a lot of low thermal conductive thermal paste (compared to solid copper), the heat won't dissipate as good. If you apply your paste good enough, using "thin" paste will work just as well; just don't shift the heatsink around when sandwiching it all together. Having excessive paste ooze out the sides of the GPU core is fine as long as there is great contact between the entire GPU die and the copper heatsink that is on top of it. If none oozes out, you could be left with gaps between the heatsink and core, which means less surface area to allow for the heat to transfer into the heatsink.
|
|
|
Bamt still only shows the 2 gpu's on the x16's, the third card runs its fans but is just phantom to the system.
If you're sure you shorted the correct pins and they are actually shroting (not loose or disconnected), then it's probably a limitation of your motherboard. Like I said, my MSI Z77A-G45 motherboard turns off a 1x slot when the adjacent 1x slot is populated. It's possible your board may do the same (go from 16x/1x/0x mode to 8x/0x/8x mode)
|
|
|
You shouldn't need to go to the hardware store. Find some junk molex/sata power adapters and use the wires on those. Also my BTC address in my sig is valid.
|
|
|
You likely need to short the PCI-E presence pins. It's all motherboard dependant. Some motherboard don't need presence shorting at all. Some need only 1 or 2 slots shorted, Some need all of them shorted. You will only need to short the slots that have the detection problems though. My EVGA 790i Ultra SLI motherboard needs all the slots shorted to detect cards on 1x risers. My Asrock 890GX Extreme4 only needs the 1x and the middle 16x slots shorted. My MSI Z77A-G45 has 7 PCI-E slots. Only 6 can work at the same time; 1 pci-e slot turns off entirely if the adjacent 1x slot is populated, and the 2 middle slots require presence shorting as well. What I use for a wire is find an old molex to SATA power adapter (or any other adapter with similar wire thickness and you don't plan on using it), and cut off about a 4-5 inch long piece of wire. Strip off about a 1/4 inch on each end, and use your finger to bend 8 strands of wire off to the side; cut off the rest as close to the shielding as possible. Twist em together and insert each end into the proper holes on the PCI-E slot before you plug in your riser. When plugged it, they should have some resistance on being pulled out, though likely won't be much. You should be able to insert almost all of the bare wire into each hole if you cut it appropriately short.
|
|
|
The highest performing low profile card will be a Radeon HD 5670: 775MHz with 400 Stream Processors Compare to a 5770 - at 850MHz with 800 Stream processors, a 5770 gets 220Mhash/s, so expect around 100Mhash/s from a 5670. Which is currently about .04 BTC a day. This card doesn't require a PCIe power connector, but it's power rating is 65 watts, so it may or may not shut down the Dell's power supply once you fire up a Bitcoin miner after adding this card. Actually, theres a better one. 6670. 480 SPU's @ 800MHz, single slot lower profile. http://www.newegg.com/Product/Product.aspx?Item=N82E16814131469
|
|
|
I went ahead and disassembled the 5970 again and took note of how deep the hotspots penetrated the 1mm thermal pads that I put on before. I also decided to be a hypocrite and used a tiny dot of Shin-Etsu x23-7783D thermal paste on each hotspot to ensure good contact.
RED left a medium impression, but it also has a lot of surface area and was probably the main source of bending of the heatsinks/PCB. I chose to use 0.5 mm pads and dots of Shin-Etsu. The same applied to the backside. ORANGE, YELLOW, and DARK GREEN all left VERY deep impressions. I chose 0.5mm and dots of Shin-Etsu. LIGHT GREEN and PINK barely left an impression. I chose 1mm and dots of Shin-Etsu. LIGHT BLUE left a mild impression, less than RED. I chose 1mm and a dot of Shin-Etsu.
I have re-assembled it all and let it mine for about 20 minutes to get a good "average" temperature. For comparison's sake, the 5970 below it was running 58.5C and 62C, and the 5970 I needed to fix the thermal pads on was running 62.5C and 64.5C. Since I have fixed it, the 5970 below it is running 58C and 61.5C, and the 5970 I fixed is now running 55C and 58C, which is an improvement of 7C and 6C respectively. The VRM's are also running MUCH cooler compared to my other cards. They all run 1v. The "normal" 5970 is running 56C and 72C. The "fixed" 5970 is running 45C and 59C. Very cool! But then again, I also have to partially blame my highly thermally conductive pads that I chose.
|
|
|
Stop Typing Like This. It Is Extremely Annoying And You Will Probably Get Less Help Because Of It.
|
|
|
I was really hoping not to mix up thermal paste and thermal pads on the same hotspots. Whenever I did that in the past, it just got ugly.
|
|
|
|