[CLOSED] ASICMiner Prisma 1.4th/s - 1.47 BTC - page 14.

GrapeApe

sr. member

Activity: 476

Merit: 250

Quote from: PlanetCrypto on November 10, 2014, 10:36:58 AM

So the hypothesis is that with the extra board count attached to the BE (versus a single 24x Tube) causes it overheat and freak out.
I just took it out of case and put a 40mm fan over it.

Am and have been a big fan of the 1300's, they're just a little pricey compared to a 1800W-2500W server supply.
We have 5 or 6 1300's here in our environment, they whirr quietly and do their job.

98% of expected @ what clock rate?

At 240 Sorry your still having problems. It's warm here today I have the window A/C going in the mining room but still no resets. Still at 98% all boards alive.

LordPaco

full member

Activity: 120

Merit: 100

The hot board has been unplugged as soon as I discovered it with the exception of testing after inspecting/reseating (with no burn marks or damage), and the inital burn in before I knew I had problems. The other problem boards are not seeing this same high temp but I have not tested every chip to confirm there isn't a hot one somewhere.

Out of 12 hashing boards (3 prisma), 8 are working without issue, 3 work ~20-60% of the time and not for long and have resetting issues, 1 obviously has temp issues.

judypug1956

sr. member

Activity: 355

Merit: 276

Quote from: LordPaco on November 10, 2014, 01:25:56 PM

I am pointing the IR probe to the same transistor from tube to tube. This particular transistor is the worst case. I inspected/reseated the hot problem board already and I see nothing abnormal with it.

Another important point: The high temp will be reached even in the case of the prisma just being on, and not hooked up to BE controller and actually hashing.

I cannot confirm that all the problem hashing boards exhibit this kind of temp, some more research to be done today mixed in with some real work Tongue

take them offline now!

exact amount of boards bad = 4

exact amount of boards GOOD = ?

LordPaco

full member

Activity: 120

Merit: 100

I am pointing the IR probe to the same transistor from tube to tube. This particular transistor is the worst case. I inspected/reseated the hot problem board already and I see nothing abnormal with it.

Another important point: The high temp will be reached even in the case of the prisma just being on, and not hooked up to BE controller and actually hashing.

I cannot confirm that all the problem hashing boards exhibit this kind of temp, some more research to be done today mixed in with some real work Tongue

wasikidding

sr. member

Activity: 268

Merit: 250

Well could it be a thermal pad and heatsink issue the heat difference is crazy

LordPaco

full member

Activity: 120

Merit: 100

So this last batch of prisma's seem to have some problems. The first 4x I recieved from Crazy Guy do work, albeit with a dead hashing board that is totally my fault so I'll ignore that.

I dont think it is related to BE controller because I can swap two controllers and the 4x stack that works, works the same way no matter what controller I use, where the problem prisma array has problems no matter what.

It will reboot ever 10-120minutes. Power downs constantly on the same 4 out of the 12 hashing boards, after reseating the same boards, same thing. Switched order of the chain, and the problem boards move to where I put them.

Here are some odd pictures for you. I pointed my IR thermometer at the transistors at a problem board, and a working board.

Hoping the usb controller will help me isolate the problem more specifically, but my IR probe is pointing me to actual hardware issues.

Problem board (107.9C temp):

Working board: (50.5C temp)

I know hard to see it is reporting in Celsius - What is going on here?

wasikidding

sr. member

Activity: 268

Merit: 250

Quote from: mrpark on November 10, 2014, 10:59:11 AM

Can any of the Prisma gurus tell me if this output is expected for normal operations? I have been running the Primsa a few days now. But it shows "Running time" up only 20 minutes? Is my Prisma rebooting ever 20 minutes? Should I be concerned about this?

Im assuming you have hit the refresh on your browser or the wakeup button

wasikidding

sr. member

Activity: 268

Merit: 250

I had all those issues been running now for 15 hours no issues the tube port on ck solo pool had been my best the BE controller is the root of most issues all day yesterday would drop to like 300 gh/s while hashing at slush's etc mmpool and slush and ck all saw improvment from handles out and relocation of miner and controller

mrpark

sr. member

Activity: 364

Merit: 250

Can any of the Prisma gurus tell me if this output is expected for normal operations? I have been running the Primsa a few days now. But it shows "Running time" up only 20 minutes? Is my Prisma rebooting ever 20 minutes? Should I be concerned about this?

PlanetCrypto

sr. member

Activity: 462

Merit: 250

Cover off of the BE with 40mm fan blowing 70 degree air on it, declocked to 230.
This time it made it about 1 hour and 10 minutes.
This is beginning to piss me off.

PlanetCrypto

sr. member

Activity: 462

Merit: 250

Quote from: GrapeApe on November 10, 2014, 10:14:36 AM

@PlanetCrypto
Everything you are describing is what I was experiencing even the fake hashing. I find that the BEController is very sensitive and it is the culprit here. I moved the controller to a COLD SPOT and problems disappeared for me. Put it somewhere cool. The EVGA 1300 G2 is just fine with the cable configuration talked about several pages back it's how I'm powering all 3 of my Prismas.

I've been hashing now at 98% of expected for over 36 hours no problems now whatsoever. HWe is 0.33%

So the hypothesis is that with the extra board count attached to the BE (versus a single 24x Tube) causes it overheat and freak out.
I just took it out of case and put a 40mm fan over it.

Am and have been a big fan of the 1300's, they're just a little pricey compared to a 1800W-2500W server supply.
We have 5 or 6 1300's here in our environment, they whirr quietly and do their job.

98% of expected @ what clock rate?

GrapeApe

sr. member

Activity: 476

Merit: 250

@PlanetCrypto
Everything you are describing is what I was experiencing even the fake hashing. I find that the BEController is very sensitive and it is the culprit here. I moved the controller to a COLD SPOT and problems disappeared for me. Put it somewhere cool. The EVGA 1300 G2 is just fine with the cable configuration talked about several pages back it's how I'm powering all 3 of my Prismas.

I've been hashing now at 98% of expected for over 36 hours no problems now whatsoever. HWe is 0.33%

PlanetCrypto

sr. member

Activity: 462

Merit: 250

Quote from: philipma1957 on November 10, 2014, 09:46:47 AM

Quote from: PlanetCrypto on November 10, 2014, 07:51:01 AM

Quote from: PlanetCrypto on November 10, 2014, 07:49:44 AM

Have been arguing with 2x Prisma's connected to a BE.
Can't wait for those USB-UART adapter(s) to show up and get rid of this rinky-dink BE controller.

Usually 2 reboots (hard or soft) are required for the BE to recognize all the boards.
Changing ASIC Section WakeUp period from the stock 3600 to 7200 helped immensely.

Clocking @ 250, handles in, powered by DPS-1600BB's, 8 PCIe plugs per Prisma, Real Perf Hash Rate = 2885 GH/s, Hashing to BAN-Hash rate is ~2750 GH/s
All the screws on the hash boards were loose, some more than others, some so loose they were about to fall out.

Initially powered them with 2x EVGA 1300's feed with 120V and 4 PCIe cables per Prisma, Kill-A-Watt indicated @ 240 they were drawing 1250W each, @ 270 they were drawing 1350W each.
The PCIe cables were slightly more than warm to the touch, regardless.

Hence the shift to DPS-1600BB's which can supply 1800W each and 8 PCIe's per Prisma.

Data Center Temp is 70F-80F.

The BE also seems to get confused after an hour or so if worker difficulty is greater than 1024 and stops hashing.
Hash rate indicated on the machine is normal, but pool reports 0.
Worker diff currently locked to 512. Testing.

Clocking @ 250, handles in, powered by DPS-1600BB's, 8 PCIe plugs per Prisma, Real Perf Hash Rate = 2885 GH/s, Hashing to BAN-Hash rate is ~2750 GH/s
All the screws on the hash boards were loose, some more than others, some so loose they were about to fall out.

Initially powered them with 2x EVGA 1300's feed with 120V and 4 PCIe cables per Prisma, Kill-A-Watt indicated @ 240 they were drawing 1250W each, @ 270 they were drawing 1350W each.
The PCIe cables were slightly more than warm to the touch, regardless.

Hence the shift to DPS-1600BB's which can supply 1800W each and 8 PCIe's per Prisma.

Data Center Temp is 70F-80F.

The BE also seems to get confused after an hour or so if worker difficulty is greater than 1024 and stops hashing.
Hash rate indicated on the machine is normal, but pool reports 0.
Worker diff currently locked to 512. Testing.

drop the clock. 250 is too high try 230 . I found the short tubes were heat sensitive and preferred underclocking handles out and 2 fans not one.

also try hashing them at mmpool.org

my short tubes ran really well at that pool.

Roger that.
Declocked to 230, now let's see if it runs more than an hour.
Do you have a spec/model # of a suitable/matching fan?

Have measured the chip temps with IR and while there is a difference between the front chips and the back chips the delta T isn't that great.
Tj is probably within operating specs.

PlanetCrypto

sr. member

Activity: 462

Merit: 250

Quote from: TracerX on November 10, 2014, 09:45:18 AM

Quote from: PlanetCrypto on November 10, 2014, 07:51:01 AM

Clocking @ 250, handles in, powered by DPS-1600BB's, 8 PCIe plugs per Prisma, Real Perf Hash Rate = 2885 GH/s, Hashing to BAN-Hash rate is ~2750 GH/s
All the screws on the hash boards were loose, some more than others, some so loose they were about to fall out.

Initially powered them with 2x EVGA 1300's feed with 120V and 4 PCIe cables per Prisma, Kill-A-Watt indicated @ 240 they were drawing 1250W each, @ 270 they were drawing 1350W each.
The PCIe cables were slightly more than warm to the touch, regardless.

Hence the shift to DPS-1600BB's which can supply 1800W each and 8 PCIe's per Prisma.

Data Center Temp is 70F-80F.

The BE also seems to get confused after an hour or so if worker difficulty is greater than 1024 and stops hashing.
Hash rate indicated on the machine is normal, but pool reports 0.
Worker diff currently locked to 512. Testing.

Thanks for the Kill-A-Watt data, I appreciate it!

In our experience the EVGA 1300's pickup a few percent improvement in efficiency when run off of 240V.
Running off of 120V they are ~90%-92%, and typically ~92%-95% off of 240V.

Was trying to make the point that OC'ing a Prisma with an EVGA 1300 feed by 120V is a recipe for burning up a $200 power supply.
They are a beefy PC supply, but would imagine they will be short lived if used to power an OC'ed Prisma for any length of time (i.e. days).
A warm environment would exasperate this situation.

philipma1957

legendary

Activity: 4256

Merit: 8551

'The right to privacy matters'

Quote from: PlanetCrypto on November 10, 2014, 07:51:01 AM

Quote from: PlanetCrypto on November 10, 2014, 07:49:44 AM

Have been arguing with 2x Prisma's connected to a BE.
Can't wait for those USB-UART adapter(s) to show up and get rid of this rinky-dink BE controller.

Usually 2 reboots (hard or soft) are required for the BE to recognize all the boards.
Changing ASIC Section WakeUp period from the stock 3600 to 7200 helped immensely.

Clocking @ 250, handles in, powered by DPS-1600BB's, 8 PCIe plugs per Prisma, Real Perf Hash Rate = 2885 GH/s, Hashing to BAN-Hash rate is ~2750 GH/s
All the screws on the hash boards were loose, some more than others, some so loose they were about to fall out.

Initially powered them with 2x EVGA 1300's feed with 120V and 4 PCIe cables per Prisma, Kill-A-Watt indicated @ 240 they were drawing 1250W each, @ 270 they were drawing 1350W each.
The PCIe cables were slightly more than warm to the touch, regardless.

Hence the shift to DPS-1600BB's which can supply 1800W each and 8 PCIe's per Prisma.

Data Center Temp is 70F-80F.

The BE also seems to get confused after an hour or so if worker difficulty is greater than 1024 and stops hashing.
Hash rate indicated on the machine is normal, but pool reports 0.
Worker diff currently locked to 512. Testing.

Clocking @ 250, handles in, powered by DPS-1600BB's, 8 PCIe plugs per Prisma, Real Perf Hash Rate = 2885 GH/s, Hashing to BAN-Hash rate is ~2750 GH/s
All the screws on the hash boards were loose, some more than others, some so loose they were about to fall out.

Initially powered them with 2x EVGA 1300's feed with 120V and 4 PCIe cables per Prisma, Kill-A-Watt indicated @ 240 they were drawing 1250W each, @ 270 they were drawing 1350W each.
The PCIe cables were slightly more than warm to the touch, regardless.

Hence the shift to DPS-1600BB's which can supply 1800W each and 8 PCIe's per Prisma.

Data Center Temp is 70F-80F.

The BE also seems to get confused after an hour or so if worker difficulty is greater than 1024 and stops hashing.
Hash rate indicated on the machine is normal, but pool reports 0.
Worker diff currently locked to 512. Testing.

drop the clock. 250 is too high try 230 . I found the short tubes were heat sensitive and preferred underclocking handles out and 2 fans not one.

also try hashing them at mmpool.org

my short tubes ran really well at that pool.

PlanetCrypto

sr. member

Activity: 462

Merit: 250

When this happens the hash rate listed in the BE is normal (It thinks it's hashing).
A reset or 2 causes it to start hashing @ the pool again.
All of our other of our miners act normally when this happens (So it's NOT a networking issue).
The is a "known good" BE as it was used flawlessly on a 24x tube for months.

I'm scratching my head on this one.
But re-booting every hour is flat unacceptable.

TracerX

hero member

Activity: 918

Merit: 1002

Quote from: PlanetCrypto on November 10, 2014, 07:51:01 AM

Clocking @ 250, handles in, powered by DPS-1600BB's, 8 PCIe plugs per Prisma, Real Perf Hash Rate = 2885 GH/s, Hashing to BAN-Hash rate is ~2750 GH/s
All the screws on the hash boards were loose, some more than others, some so loose they were about to fall out.

Initially powered them with 2x EVGA 1300's feed with 120V and 4 PCIe cables per Prisma, Kill-A-Watt indicated @ 240 they were drawing 1250W each, @ 270 they were drawing 1350W each.
The PCIe cables were slightly more than warm to the touch, regardless.

Hence the shift to DPS-1600BB's which can supply 1800W each and 8 PCIe's per Prisma.

Data Center Temp is 70F-80F.

The BE also seems to get confused after an hour or so if worker difficulty is greater than 1024 and stops hashing.
Hash rate indicated on the machine is normal, but pool reports 0.
Worker diff currently locked to 512. Testing.

Thanks for the Kill-A-Watt data, I appreciate it!

PlanetCrypto

sr. member

Activity: 462

Merit: 250

Quote from: PlanetCrypto on November 10, 2014, 07:49:44 AM

Have been arguing with 2x Prisma's connected to a BE.
Can't wait for those USB-UART adapter(s) to show up and get rid of this rinky-dink BE controller.

Usually 2 reboots (hard or soft) are required for the BE to recognize all the boards.
Changing ASIC Section WakeUp period from the stock 3600 to 7200 helped immensely.

Clocking @ 250, handles in, powered by DPS-1600BB's, 8 PCIe plugs per Prisma, Real Perf Hash Rate = 2885 GH/s, Hashing to BAN-Hash rate is ~2750 GH/s
All the screws on the hash boards were loose, some more than others, some so loose they were about to fall out.

Initially powered them with 2x EVGA 1300's feed with 120V and 4 PCIe cables per Prisma, Kill-A-Watt indicated @ 240 they were drawing 1250W each, @ 270 they were drawing 1350W each.
The PCIe cables were slightly more than warm to the touch, regardless.

Hence the shift to DPS-1600BB's which can supply 1800W each and 8 PCIe's per Prisma.

Data Center Temp is 70F-80F.

The BE also seems to get confused after an hour or so if worker difficulty is greater than 1024 and stops hashing.
Hash rate indicated on the machine is normal, but pool reports 0.
Worker diff currently locked to 512. Testing.

Clocking @ 250, handles in, powered by DPS-1600BB's, 8 PCIe plugs per Prisma, Real Perf Hash Rate = 2885 GH/s, Hashing to BAN-Hash rate is ~2750 GH/s
All the screws on the hash boards were loose, some more than others, some so loose they were about to fall out.

Initially powered them with 2x EVGA 1300's feed with 120V and 4 PCIe cables per Prisma, Kill-A-Watt indicated @ 240 they were drawing 1250W each, @ 270 they were drawing 1350W each.
The PCIe cables were slightly more than warm to the touch, regardless.

Hence the shift to DPS-1600BB's which can supply 1800W each and 8 PCIe's per Prisma.

Data Center Temp is 70F-80F.

The BE also seems to get confused after an hour or so if worker difficulty is greater than 1024 and stops hashing.
Hash rate indicated on the machine is normal, but pool reports 0.
Worker diff currently locked to 512. Testing.

PlanetCrypto

sr. member

Activity: 462

Merit: 250

Have been arguing with 2x Prisma's connected to a BE.
Can't wait for those USB-UART adapter(s) to show up and get rid of this rinky-dink BE controller.

Usually 2 reboots (hard or soft) are required for the BE to recognize all the boards.
Changing ASIC Section WakeUp period from the stock 3600 to 7200 helped immensely.

Clocking @ 250, handles in, powered by DPS-1600BB's, 8 PCIe plugs per Prisma, Real Perf Hash Rate = 2885 GH/s, Hashing to BAN-Hash rate is ~2750 GH/s
All the screws on the hash boards were loose, some more than others, some so loose they were about to fall out.

Initially powered them with 2x EVGA 1300's feed with 120V and 4 PCIe cables per Prisma, Kill-A-Watt indicated @ 240 they were drawing 1250W each, @ 270 they were drawing 1350W each.
The PCIe cables were slightly more than warm to the touch, regardless.

Hence the shift to DPS-1600BB's which can supply 1800W each and 8 PCIe's per Prisma.

Data Center Temp is 70F-80F.

The BE also seems to get confused after an hour or so if worker difficulty is greater than 1024 and stops hashing.
Hash rate indicated on the machine is normal, but pool reports 0.
Worker diff currently locked to 512. Testing.

philipma1957

legendary

Activity: 4256

Merit: 8551

'The right to privacy matters'

my long tube unit is on the way. will run it :

1) handles out
2) add this fan-- http://www.amazon.com/gp/product/B00A460TK6/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1
3) use a rasp pi-- with crazy's custom software
4)will post a lot of photos
5) use this psu-- http://www.amazon.com/EVGA-Supernova-Platinum-1600-Watt-220-P2-1600-X1/dp/B00NJG61JQ/ref=sr_1_2?s=electronics&ie=UTF8&qid=1415622444&sr=1-2&keywords=evga+1600

oh BTW coins are now above 370 usd

Topic: [CLOSED] ASICMiner Prisma 1.4th/s - 1.47 BTC - page 14. (Read 49921 times)