pazor
Legendary
Offline
Activity: 966
Merit: 1000
|
|
May 01, 2012, 08:27:14 PM |
|
So it looks like the bottom heatsink and fan is definitely helping with the speed/thermals. I am going to give it a test.
Does anybody know if the heatsink has to be suspended in air? I am putting these babies on top of a wooden rack. I was thinking of supporting the middle with another screw since I have noticed the board (quads) to flex a little bit. Now I am thinking of covering the bottom with a piece of thermal pad that is supported by a heatsink (I have ~20 old 1U socket 478 Cu heatsinks). I assume that as long as the pad/case is non-conductive/capacitive I should be alright. But please let me know if this wouldn't work before I experiment. (@pazor: Your setup is definitely more elegant. But I have a number of boards and the cost with the screws/added heatsinks is a consideration.)
FPGA pad (fasten to the board via the 4 standoff screws, maybe the Fujipoly thermal pad, frozencpu dot com) heatsink (same height as the 4 raised standoffs) Wood
The alternative is to use 4 passive VGA coolers attached to the bottom of the LX150. Probably safer and cleaner.
try to use the 4 passive vga coolers and take a fan to blow constantly from the bottom. i think that should make sense. if i change my cooling setup with the two titan heatsink/fan combination i will report the results. good luck!
|
treuhand-Dienst gewünscht? - frag per PM an BTC 174X17nR7vEQBQo4GXKRGMGaTmB49Gf1yT
|
|
|
ummas
|
|
May 01, 2012, 08:34:40 PM |
|
I bought 4 heatsinks to mount in fromthe bottom through thermalpad. Best bacplate ever. I`ll try to watercool it and report the effect.
|
|
|
|
CA Coins
Donator
Sr. Member
Offline
Activity: 305
Merit: 250
|
|
May 02, 2012, 06:21:20 AM |
|
Most likely I will go with the passive VGA memory coolers, but I am surprised I haven't got any "you're crazy" comments for my other idea.
|
|
|
|
gr0bi42
|
|
May 02, 2012, 08:08:30 AM |
|
Good thermal pads can be used. (But not the ones delivered with the cooler. They are quite poor). Thermal grease is better because the layer should always be thinner than a pad. I recommend low viscosity thermal grease (if you are from Germany you can use item no. "LEITPASTE 20GR" from http://www.reichelt.de) and the "blob in the center" method. It is simple, quick and save. I use this method for testing new batches and never had any problems. I've tried Arctic MX 2. It's better than Artic Silver 5, but I could not experience such a huge step like Chefnet: or better cooling, my single went up from 204 to 220 with artic mx-2 under the cooler
... usually I get 4-8 Mhz more. 002-0: ztex_ufm1_15y1-04A347274B-1: f=200.00MHz, errorRate=0.00%, maxErrorRate=0.58%, hashRate=200.0MH/s, submitted 12 new nonces, luckFactor=0.99 002-0: ztex_ufm1_15y1-04A347274B-2: f=204.00MHz, errorRate=0.00%, maxErrorRate=1.52%, hashRate=204.0MH/s, submitted 17 new nonces, luckFactor=0.97 002-0: ztex_ufm1_15y1-04A347274B-3: f=196.00MHz, errorRate=0.00%, maxErrorRate=0.81%, hashRate=196.0MH/s, submitted 10 new nonces, luckFactor=1.03 002-0: ztex_ufm1_15y1-04A347274B-4: f=200.00MHz, errorRate=0.00%, maxErrorRate=1.18%, hashRate=200.0MH/s, submitted 19 new nonces, luckFactor=1.02 --- 002-0: ztex_ufm1_15y1-04A347274B-2: f=208.00MHz, errorRate=0.00%, maxErrorRate=0.50%, hashRate=208.0MH/s, submitted 9 new nonces, luckFactor=0.96 002-0: ztex_ufm1_15y1-04A347274B-3: f=200.00MHz, errorRate=0.00%, maxErrorRate=0.62%, hashRate=200.0MH/s, submitted 11 new nonces, luckFactor=0.98 002-0: ztex_ufm1_15y1-04A347274B-4: f=204.00MHz, errorRate=0.00%, maxErrorRate=0.99%, hashRate=204.0MH/s, submitted 15 new nonces, luckFactor=0.96 002-0: ztex_ufm1_15y1-04A3476278-1: f=208.00MHz, errorRate=0.65%, maxErrorRate=3.22%, hashRate=206.6MH/s, submitted 18 new nonces, luckFactor=0.96 @Chefnet: How do you apply the grease? A blob in the middle? How much? Moving the heatsink around, or just pushing it down? The one problematic board still sucks. Hashrate went up: 002-0: ztex_ufm1_15y1-04A3476278-1: f=200.00MHz, errorRate=0.87%, maxErrorRate=1.00%, hashRate=198.3MH/s, submitted 14 new nonces, luckFactor=0.82 002-0: ztex_ufm1_15y1-04A3476278-2: f=196.00MHz, errorRate=0.23%, maxErrorRate=0.55%, hashRate=195.5MH/s, submitted 16 new nonces, luckFactor=0.88 002-0: ztex_ufm1_15y1-04A3476278-3: f=196.00MHz, errorRate=0.00%, maxErrorRate=0.81%, hashRate=196.0MH/s, submitted 13 new nonces, luckFactor=0.83 002-0: ztex_ufm1_15y1-04A3476278-4: f=192.00MHz, errorRate=0.11%, maxErrorRate=0.50%, hashRate=191.8MH/s, submitted 10 new nonces, luckFactor=0.95 --- 002-0: ztex_ufm1_15y1-04A3476278-1: f=208.00MHz, errorRate=0.81%, maxErrorRate=3.22%, hashRate=206.3MH/s, submitted 16 new nonces, luckFactor=0.96 002-0: ztex_ufm1_15y1-04A3476278-2: f=204.00MHz, errorRate=0.26%, maxErrorRate=1.20%, hashRate=203.5MH/s, submitted 17 new nonces, luckFactor=1.06 002-1: ztex_ufm1_15y1-04A3476278-3: f=204.00MHz, errorRate=0.04%, maxErrorRate=3.01%, hashRate=203.9MH/s, submitted 15 new nonces, luckFactor=0.98 002-1: ztex_ufm1_15y1-04A3476278-4: f=196.00MHz, errorRate=0.00%, maxErrorRate=0.56%, hashRate=196.0MH/s, submitted 10 new nonces, luckFactor=1.06
...but No.4 still is below 200 Mhz and I'm getting FPGA-shutdowns (5 times since yesterday): 002-0: ztex_ufm1_15y1-04A3476278-1: Error: Hash rate drop of 5.8% detect. This may be caused by overheating. FPGA is shut down to prevent damage. 51.99384114688102: Device disabled since 2012-05-02T03:05:59 002-0: ztex_ufm1_15y1-04A3476278-2: Error: Hash rate drop of 5.4% detect. This may be caused by overheating. FPGA is shut down to prevent damage. 51.772818289686384: Device disabled since 2012-05-02T02:57:22
|
Donations are welcome: 1Btf3BqUegfe5iFdWsgfBf1Ew3YsAvsrLT
|
|
|
Chefnet
|
|
May 02, 2012, 08:11:44 AM |
|
i have allocated it around the chip not only in the middle. why it was such a big step I don't know.
|
|
|
|
ztex (OP)
Donator
Sr. Member
Offline
Activity: 367
Merit: 250
ZTEX FPGA Boards
|
|
May 02, 2012, 09:00:27 AM |
|
without bottom heatsink and fan (fan is recommended) between 212MH/s and 216MH/s with error rates abou 0.3-0.6%
with bottom heatsink and fan now 220MH/s with 0.00% error rate
The bottom side of the PCB only gets hot if the heat transfer on the top side of the chip is insufficient. If heat sink is installed properly the temperature on the bottom side is about 10°C over ambient temperature. Cooling it down by 5°C would cool down the chip by about 2°C (due to the bad thermal conductivity of the PCB). This has no effect. Use low viscostity thermal grease (e.g. the one delivered with the cooler) and the (big) "blob in the center" method: https://bitcointalk.org/index.php?topic=49180.msg875211#msg875211 . This is simple, quick and save.
|
|
|
|
BR0KK
|
|
May 02, 2012, 10:31:19 AM |
|
My quad was not working properly at first. I reapplied new thermal compound and the heatsinks. Then it worked perfect My new single, that I programmed with the wron bitstream, works verry well to. It runs 220 MHz with the stock cooler and stock thermal compound!;)
|
|
|
|
Turbor
Legendary
Offline
Activity: 1022
Merit: 1000
BitMinter
|
|
May 02, 2012, 08:16:50 PM |
|
I was playing around with two 120mm fans. Let's make it short: It was disappointing ! No matter from where you blow at the board you will not gain more than 4 MHz. Looks like things are really maxed out with this heatsink.
|
|
|
|
simon66
|
|
May 02, 2012, 10:53:52 PM |
|
I was playing around with two 120mm fans. Let's make it short: It was disappointing ! No matter from where you blow at the board you will not gain more than 4 MHz. Looks like things are really maxed out with this heatsink.
Please post some pics lol. Of your board that is.
|
|
|
|
Turbor
Legendary
Offline
Activity: 1022
Merit: 1000
BitMinter
|
|
May 03, 2012, 10:30:13 AM |
|
Please post some pics lol. Of your board that is.
I'm facing some nasty DCOD problems. Yesterday evening the boards ran 2x 212, 1x 208 and 1x 204. When I woke up board nr3 is down. Overheat protection. An other one is down to 200 MHz. I removed nr3 and placed it again with the same result. But this time all the other chips follow the downclock. 10 seconds and you have a dead board. I switched power supply with the same result. Going to re-upload the firmware this evening and screwing down the heatsinks. I dislike the weakness of the PCB it looks like it's bending a bit. Not sure if that is part of the problem. I'm pretty pissed but I know this from my 1.15d board. I'm ready for some clamp action.
|
|
|
|
gr0bi42
|
|
May 03, 2012, 11:36:13 AM |
|
Please post some pics lol. Of your board that is.
I'm facing some nasty DCOD problems. Yesterday evening the boards ran 2x 212, 1x 208 and 1x 204. When I woke up board nr3 is down. Overheat protection. Same problem here. My top-board (228, 232, 235, 224) was down in the morning due to Overheat protection. But I think it's not a cooling issue, rather than a power issue. The board was running super stable for days. Yesterday I changed the setup to another room. PSU is a Cooler Master Gold 700W and the setup consumes 400W at the power plug (CPU, 2x 5870, 3x Ztex Quad). I power the Quads from one PCIe power cable. This cable has 1x PCIe-6/8 and 1x PCIe-6 connector. PCIe cable to Quad's is wired in this way (the little o is the pin of the PCIe connector): o-12v Q1 o-GND Q1 PCIe-6 (1) o- o-GND Q1 o-12v Q2 o-GND Q2
o- o-GND Q2 PCIe-6 (2) o- o-GND Q3 o-12v Q3 o-GND Q3
The Quad which came down over night is Q3 and is connected to the second PCIe-6 plug of the power cable, which in turn extends from the first PCIe-6/8 plug. Maybe this is the problem (GND wires)?
|
Donations are welcome: 1Btf3BqUegfe5iFdWsgfBf1Ew3YsAvsrLT
|
|
|
ztex (OP)
Donator
Sr. Member
Offline
Activity: 367
Merit: 250
ZTEX FPGA Boards
|
|
May 03, 2012, 04:20:46 PM |
|
If there are problems with particular boards, please send me the complete logs (from starting the board till occurrence of the problem) per email ( ). I cannot help If I do not know what is going on.
|
|
|
|
Turbor
Legendary
Offline
Activity: 1022
Merit: 1000
BitMinter
|
|
May 03, 2012, 08:26:06 PM Last edit: May 04, 2012, 04:40:43 AM by Turbor |
|
It's only one chip from the board that doesn't work on my side. Looks like i got the ztex trio Edit: Ztex offered me to send the board in
|
|
|
|
nelisky
Legendary
Offline
Activity: 1540
Merit: 1002
|
|
May 04, 2012, 08:13:54 AM |
|
I have a test branch with support for ZTEX 1.15y quad boards. Anyone willing to give it a try please grab the source at https://github.com/nelisky/cgminer/tree/ztex-120417If you are unable to compile it yourself I might be able to provide binaries on some platforms.
|
|
|
|
Kuma
Member
Offline
Activity: 107
Merit: 10
|
|
May 04, 2012, 08:57:44 AM |
|
Great job Nelisky, I'll try to test it today and I hope Iwon't burn my Ztex board .
|
|
|
|
BR0KK
|
|
May 04, 2012, 09:28:44 AM |
|
I can tell u that cgminer works charming with all my boards (running since yesterday. Ill post a pic)
|
|
|
|
pazor
Legendary
Offline
Activity: 966
Merit: 1000
|
|
May 04, 2012, 09:32:19 AM |
|
I can tell u that cgminer works charming with all my boards (running since yesterday. Ill post a pic)
any advance ? what the differenc to btcminer of ztex ?
|
treuhand-Dienst gewünscht? - frag per PM an BTC 174X17nR7vEQBQo4GXKRGMGaTmB49Gf1yT
|
|
|
BR0KK
|
|
May 04, 2012, 09:38:37 AM |
|
For example multi platform mining? (so u don't need Java and u can plug your boards into the GPU miners that u already hav in use) Better GUI (I mean the visible screen with all the info like stales, has rate, etc.) Multipool option End of thinking capability :p Go read cgminer thread for all the benefits
|
|
|
|
gr0bi42
|
|
May 04, 2012, 10:38:04 AM |
|
I have a test branch with support for ZTEX 1.15y quad boards. Anyone willing to give it a try please grab the source at https://github.com/nelisky/cgminer/tree/ztex-120417If you are unable to compile it yourself I might be able to provide binaries on some platforms. Compiled... and it's running on 2 Quads. Great job!!!
|
Donations are welcome: 1Btf3BqUegfe5iFdWsgfBf1Ew3YsAvsrLT
|
|
|
gr0bi42
|
|
May 04, 2012, 11:11:53 AM |
|
I have a test branch with support for ZTEX 1.15y quad boards. Anyone willing to give it a try please grab the source at https://github.com/nelisky/cgminer/tree/ztex-120417If you are unable to compile it yourself I might be able to provide binaries on some platforms. Compiled... and it's running on 2 Quads. Great job!!! [2012-05-04 12:50:41] ZTEX 003:005-04A3466370: Frequency change from 212.00 to 216.00 Mhz [2012-05-04 12:50:41] ZTEX 003:005-04A3466370: Frequency change from 212.00 to 216.00 Mhz [2012-05-04 12:50:43] Accepted 710365f7.ca63213a ZTX 1 pool 0 [2012-05-04 12:50:43] Accepted 6b11ae9f.70effa52 ZTX 6 pool 0 [2012-05-04 12:50:51] ZTEX 001:014-04A346BE51: Frequency change from 212.00 to 216.00 Mhz [2012-05-04 12:50:52] ZTEX 001:014-04A346BE51: Frequency change from 212.00 to 216.00 Mhz [2012-05-04 12:50:52] ZTEX 003:005-04A3466370: Frequency change from 216.00 to 212.00 Mhz [2012-05-04 12:50:52] ZTEX 001:014-04A346BE51: Frequency change from 212.00 to 216.00 Mhz [2012-05-04 12:50:52] ZTEX 001:014-04A346BE51: Frequency change from 212.00 to 216.00 Mhz [2012-05-04 12:50:54] Accepted b7f4faac.c6f9861c ZTX 1 pool 0 [2012-05-04 12:50:55] Accepted 1af6ea76.aa90d4f9 ZTX 6 pool 0 [2012-05-04 12:50:58] Accepted 47e2c5b7.ef88efee ZTX 3 pool 0 Missing a minor info about the fpga number on the quad. For instance: [2012-05-04 12:50:41] ZTEX 003:005-04A3466370: Frequency change from 212.00 to 216.00 Mhz does not say which of the 4 quads you are talking about. Put a suffix "-1" ... "-4" at the end of the serial number. In this way: [2012-05-04 12:50:41] ZTEX 003:005-04A3466370-1: Frequency change from 212.00 to 216.00 Mhz Hope this is possible.
|
Donations are welcome: 1Btf3BqUegfe5iFdWsgfBf1Ew3YsAvsrLT
|
|
|
|