[Guide] Dogie's Comprehensive Power Supply Analysis

[JLebowskiTheDude]

Its because when this thread is 9 months old, and the OP has been updated again and again, our random posts in the thread [especially 1st page] will remain old and out of date. I have the code recorded, thanks.

Okay.  I suppose that makes some sense.  

But that's true of all threads.  If I go read a major Pool thread, I'm going to find outdated posts.  Especially on the first page.

The specs of the powers supplies aren't going to change.  Prices will.  To me in a thread like this that "snapshot" data might actually be useful.  Like, 9 months ago this power supply was $175.  Now in the OP it's $100.

Your ratings will change too of course based on the pricing.  I don't think you have plans to update those on a very regular basis.  Maybe a high/low/average pricing would be a better approach.

For me personally I factor in cyclical pricing on technological purchases when making hardware choices.  Especially since sales are typically very cyclical.

[sniffinpoprocks]

Thanks SO MUCH for this post!! I think it really stands out next to your great setup guides. I'd like to caution others before recommending other power supplies be added to the list. A 'great' power supply in any other application doesn't make a great ASIC miner power supply. If it hasn't been tested stable for a few months on a Cube, Blade, Ant.. etc; please don't recommend it. I have gone through 2 other 'quality' power supplies that would do wonders in a PC, that have failed miserably within days at powering a single Cube. A 600w OCZ and a Cooler Master V1000 (1000w) both were not up to the task.
I just ordered 2 CXM750's to quench my thirst for power. Hope to have my Cubes up and rocking in a few days!!

If there was a tutorial section, I think you could fill it up alone.

And, as one of those taking up your time with PM's and you answering me in a very timely manner...
Thanks again!

[dogie]

Its because when this thread is 9 months old, and the OP has been updated again and again, our random posts in the thread [especially 1st page] will remain old and out of date. I have the code recorded, thanks.

Okay.  I suppose that makes some sense. 

But that's true of all threads.  If I go read a major Pool thread, I'm going to find outdated posts.  Especially on the first page.

The specs of the powers supplies aren't going to change.  Prices will.  To me in a thread like this that "snapshot" data might actually be useful.  Like, 9 months ago this power supply was $175.  Now in the OP it's $100.

Your ratings will change too of course based on the pricing.  I don't think you have plans to update those on a very regular basis.  Maybe a high/low/average pricing would be a better approach.

For me personally I factor in cyclical pricing on technological purchases when making hardware choices.  Especially since sales are typically very cyclical.

It will be updated pretty often (likely weekly), but sites like Amazon can change every 5 minutes if they're price matching competitors.

[dogie]

Thanks SO MUCH for this post!! I think it really stands out next to your great setup guides. I'd like to caution others before recommending other power supplies be added to the list. A 'great' power supply in any other application doesn't make a great ASIC miner power supply. If it hasn't been tested stable for a few months on a Cube, Blade, Ant.. etc; please don't recommend it. I have gone through 2 other 'quality' power supplies that would do wonders in a PC, that have failed miserably within days at powering a single Cube. A 600w OCZ and a Cooler Master V1000 (1000w) both were not up to the task.
I just ordered 2 CXM750's to quench my thirst for power. Hope to have my Cubes up and rocking in a few days!!

If there was a tutorial section, I think you could fill it up alone.

And, as one of those taking up your time with PM's and you answering me in a very timely manner...
Thanks again!

Thanks for your kind words Thats the problem with ASIC PSUs, providing 12-100A 24/7 is totally different than spike 12-100A. It takes a different beast to power an Antminer/cube.

[JLebowskiTheDude]

It will be updated pretty often (likely weekly), but sites like Amazon can change every 5 minutes if they're price matching competitors.

Yeah. Weekly at least here in the US can mean a wide variety of prices.  

That's why a high/low/average might be more meaningful.  Depends who the intended audience is I suppose?

[JLebowskiTheDude]

I'd like to caution others before recommending other power supplies be added to the list. A 'great' power supply in any other application doesn't make a great ASIC miner power supply. If it hasn't been tested stable for a few months on a Cube, Blade, Ant.. etc; please don't recommend it. I have gone through 2 other 'quality' power supplies that would do wonders in a PC, that have failed miserably within days at powering a single Cube. A 600w OCZ and a Cooler Master V1000 (1000w) both were not up to the task.
I just ordered 2 CXM750's to quench my thirst for power. Hope to have my Cubes up and rocking in a few days!!

I don't think that was directed specifically at me but in case it was by any chance all the power supplies I posted are confirmed to work with mining hardware.

The exception being the Seasonic SS-760XP2 which did NOT work with Cubes.  Even just a single, which was surprising but as I mentioned it's likely due to a protection issue rather than a lack of power issue.  The SS-760XP2 is a beast of a PC level supply.

[daddyfatsax]

Dogie, why did you not like the EVGA 1000W PSU? I ran 2 cubes and a BFL 60 single for 2 months on that thing and it was awesome. 2nd favorite is the Corsair AX 1200. Nice guide. Thanks!

[JLebowskiTheDude]

Dogie, why did you not like the EVGA 1000W PSU? I ran 2 cubes and a BFL 60 single for 2 months on that thing and it was awesome. 2nd favorite is the Corsair AX 1200. Nice guide. Thanks!

I don't think it's a question of "not liking" it.  It's a quality power supply.

It's the fact that the algorithm is taking price and perhaps other variables into account when computing the "dogie score".

My guess would be that the algorithm also weights all scores relative to one model.  That one being the Corsair TX750M currently.

[dogie]

Dogie, why did you not like the EVGA 1000W PSU? I ran 2 cubes and a BFL 60 single for 2 months on that thing and it was awesome. 2nd favorite is the Corsair AX 1200. Nice guide. Thanks!

I don't think it's a question of "not liking" it.  It's a quality power supply.

It's the fact that the algorithm is taking price and perhaps other variables into account when computing the "dogie score".

My guess would be that the algorithm also weights all scores relative to one model.  That one being the Corsair TX750M currently.

Its not relative, certain PSUs can go >100 if they're excellent and excellent value. Its rare that the much larger PSUs score well as their relative performance per $ per watt is low. Its almost always cheaper and more efficient and flexible to get two 2x ~750s.

[dogie]

Changelog 02/25/2014

• Added Seasonic range

[Unacceptable]

Very nice doggie!!! Great job!!! 

[JLebowskiTheDude]

Its not relative, certain PSUs can go >100 if they're excellent and excellent value. Its rare that the much larger PSUs score well as their relative performance per $ per watt is low. Its almost always cheaper and more efficient and flexible to get two 2x ~750s.

   Of course it's relative.   It might not be calculated that way, or intentional, or even intended bias.  But the TX750M scoring exactly a 1.0 relativity is what we call "not very statistically likely".  If it's a happy accident, I'm a monkey's uncle.

A human built the algorithm.  It's not infallible.   You have some obvious bias for Corsair.  There's nothing wrong their power supplies.  I own a ton of them.  They're great.  But there are equally decent and in some cases far better choices - especially economically speaking. Magically, they've been excluded from the list so far.  Unfortunately, the way the info is being presented, beginners might make the mistake of putting real stock in the "dogie score" and think it's a scalar difference between choices.  

FYI - some of your Corsair links are incorrect.  They simply lead to the wrong product.  I would have posted an update with correct links in a table as I did before but that wasn't very well received, so I'm erring on the side of caution this time around.

Lastly, I'll ask this again - who is the intended audience for this thread?  EU?  US?  Both?

Because the prices that are included with the Seasonic update are grossly incorrect for even the average price at major retailers like Newegg.  I'd recommend Hover Hound if you're not already using it.   Cases in point: Seasonic Platinum 760W was $100 (recently), is currently $169.99 currently, and is listed on this list as $199.99.  What's worse?  YOU SHOULDN'T BUY A SS-760XP2 TO USE WITH MINING HARDWARE.  Didn't someone already post in this thread to confirm at stock it WILL NOT WORK with mining hardware?  Oh right.  That was me!  

Back to pricing.  Since price is an inherent part of the algorithm, having inaccurate prices skews the data quite a bit.    Which is why I said perhaps a weighted average price or a high/low/average approach might be better.  I understand Amazon is usually a good common denominator for comparison shopping for a lot of things, but PC components are sometimes ridiculously overpriced there.

This is a good start towards a list, a very necessary thread here which I've glad you're working on, and a lot of the foundational work is there.  But it needs refinement.  I'd be happy to help, but last time I tried it wasn't all that well received.

[JLebowskiTheDude]

Back to pricing.  Since price is an inherent part of the algorithm, having inaccurate prices skews the data quite a bit.    Which is why I said perhaps a weighted average price or a high/low/average approach might be better. I understand Amazon is usually a good common denominator for comparison shopping for a lot of things, but PC components are sometimes ridiculously overpriced there.

It didn't make sense to link to hardware from all one vendor and at higher prices... until now.  I finally put two and two together.

OP contains Amazon referral links.

Didn't dawn on me at first to even check.  I believe in full disclosure so I am posting this information here.  

Also editing my posts in this thread back to their original info.  Because I'm actually here to help others without any promise of profit.

[Starlightbreaker]

Back to pricing.  Since price is an inherent part of the algorithm, having inaccurate prices skews the data quite a bit.    Which is why I said perhaps a weighted average price or a high/low/average approach might be better. I understand Amazon is usually a good common denominator for comparison shopping for a lot of things, but PC components are sometimes ridiculously overpriced there.

It didn't make sense to link to hardware from all one vendor and at higher prices... until now.  I finally put two and two together.

OP contains Amazon referral links.

Didn't dawn on me at first to even check.  I believe in full disclosure so I am posting this information here.  

Also editing my posts in this thread back to their original info.  Because I'm actually here to help others without any promise of profit.

well, it's kinda expected when someone's using url shortener.

i realized it since the first post.                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                         

[Unacceptable]

I wondered why he didn't just use Newegg 

[JLebowskiTheDude]

well, it's kinda expected when someone's using url shortener.

i realized it since the first post.

Not really.  From a readability and maintenance stand-point it would make sense to use it too.

Not to mention that most shorteners give clickthrough statistics which could be useful to see which supplies were more popular or at least getting more traffic.

I don't even have an issue with even using referral links. 

But it was done with no disclosure in a very clandestine manner.  And then to ask another user to REMOVE information from the thread because of the following bogus reason -

Its because when this thread is 9 months old, and the OP has been updated again and again, our random posts in the thread [especially 1st page] will remain old and out of date. I have the code recorded, thanks.

instead of the real reason - can't get income if someone buys one of those!

dogie also stated

I'll delete it out of mine once it stops being useful.

it was deleted within 12 hours.  none of that info was added to the OP.  I guess it stopped being "useful" already?   

[dogie]

 Of course it's relative.   It might not be calculated that way, or intentional, or even intended bias.  But the TX750M scoring exactly a 1.0 relativity is what we call "not very statistically likely".  If it's a happy accident, I'm a monkey's uncle.

1.001095833 if you really must know. And that's with a hell of a lot of scaling to get a range appropriate for humans to interpret.

A human built the algorithm.  It's not infallible.   You have some obvious bias for Corsair.  There's nothing wrong their power supplies.  I own a ton of them.  They're great.  But there are equally decent and in some cases far better choices - especially economically speaking. Magically, they've been excluded from the list so far.  Unfortunately, the way the info is being presented, beginners might make the mistake of putting real stock in the "dogie score" and think it's a scalar difference between choices.  

Corsair make the best PSUs with the best attributes for mining, at reasonable prices. The unbiased MATHS backs that up in some cases, why is that unexpected? The CXM range is already the best seller by FAR.

Lastly, I'll ask this again - who is the intended audience for this thread?  EU?  US?  Both?

US at this second.

Because the prices that are included with the Seasonic update are grossly incorrect for even the average price at major retailers like Newegg.  I'd recommend Hover Hound if you're not already using it.   Cases in point: Seasonic Platinum 760W was $100 (recently), is currently $169.99 currently, and is listed on this list as $199.99.  What's worse?  YOU SHOULDN'T BUY A SS-760XP2 TO USE WITH MINING HARDWARE.  Didn't someone already post in this thread to confirm at stock it WILL NOT WORK with mining hardware?  Oh right.  That was me!  

Back to pricing.  Since price is an inherent part of the algorithm, having inaccurate prices skews the data quite a bit.    Which is why I said perhaps a weighted average price or a high/low/average approach might be better.  I understand Amazon is usually a good common denominator for comparison shopping for a lot of things, but PC components are sometimes ridiculously overpriced there.

This is a good start towards a list, a very necessary thread here which I've glad you're working on, and a lot of the foundational work is there.  But it needs refinement.  I'd be happy to help, but last time I tried it wasn't all that well received.

Be patient, the guide went up three days ago and I've been in surgery for one one those, and not wanting to type much afterwards. Other sites will be added, but its very likely the 'cheapest' will change almost daily as they price match/stock price adjust all the time. Amazon links at least provide a 'here it is on the market, somewhere'. Its not easy to create dynamic code with static forum code, but it will get better.

[dogie]

Back to pricing.  Since price is an inherent part of the algorithm, having inaccurate prices skews the data quite a bit.    Which is why I said perhaps a weighted average price or a high/low/average approach might be better. I understand Amazon is usually a good common denominator for comparison shopping for a lot of things, but PC components are sometimes ridiculously overpriced there.

It didn't make sense to link to hardware from all one vendor and at higher prices... until now.  I finally put two and two together.

OP contains Amazon referral links.

Didn't dawn on me at first to even check.  I believe in full disclosure so I am posting this information here.  

Also editing my posts in this thread back to their original info.  Because I'm actually here to help others without any promise of profit.

Maybe because, as I said in my other post just now, its been up THREE days and I've been in surgery for one. More sites will be added soon, give me time. Yes, links have affiliate IDs in. I used to disclaimer them all the time, but it turned out no one actually cared. It helps contribute to what is otherwise a free resource, like google. And no, it doesn't even cover a tiny portion of time time.

The original reason still stands, I'm going to come back every week and update the OP, I highly doubt you are going to come back in 6 months and change your then grossly misrepresented links when they are out of date.

I wondered why he didn't just use Newegg  

Newegg is being added too, experimenting with whether to add both or the cheapest of that weak. Again, time. Be patient.

it was deleted within 12 hours.  none of that info was added to the OP.  I guess it stopped being "useful" already?  

It was added to the OP where useful, I have your code saved. But as exactly I was saying, the 'amazing deal' link you posted was $40 more expensive within 12 hours.

[JLebowskiTheDude]

Corsair make the best PSUs with the best attributes for mining, at reasonable prices. The unbiased MATHS backs that up in some cases, why is that unexpected? The CXM range is already the best seller by FAR.

Best is a subjective term.  The MATHS (maths? really?  ) also show better deals are out there with equally performing or even better hardware.

A $10 server-grade 90+ efficiency Z750P with a $40 board from sidehack will likely be the "best" at the moment.  I think the MATHS ( ) would show it beats out all Corsair models, even at their best sale prices.

It's not even on the list.

US at this second.

Then your pricing is extremely inaccurate.  I know.  I buy $1,000s of hardware per month.

Be patient, the guide went up three days ago and I've been in surgery for one one those, and not wanting to type much afterwards. Other sites will be added, but its very likely the 'cheapest' will change almost daily as they price match/stock price adjust all the time. Amazon links at least provide a 'here it is on the market, somewhere'. Its not easy to create dynamic code with static forum code, but it will get better.

I don't want you to doubt my appreciation.   I don't need this information but others do.  I thank you for that.

I don't doubt you have personal issues that may restrict your ability to keep this information organized or updated.  I offered to help.  Not only did you decline it, you removed the information for no other reason that it would conflict with you receiving a referral sale.

So I suppose what I do question is the manner in which this is being done at the moment.

[JLebowskiTheDude]

The original reason still stands, I'm going to come back every week and update the OP, I highly doubt you are going to come back in 6 months and change your then grossly misrepresented links when they are out of date.

Hey, thanks for the vote of confidence in a new member who is TRYING to help others.  We're you ever here on this forum doing that? I imagine you were.  How would it have felt if a member here said that to you?

It was added to the OP where useful, I have your code saved. But as exactly I was saying, the 'amazing deal' link you posted was $40 more expensive within 12 hours.

Again why I stated your pricing model is just not a good one.  A single point pricing model gives no indication of what might be available.  It just give single data point at whatever time you happened to update it.  Single point of data = single point of failure.  Can't see any kind of trending from ONE data point.  I don't care what kind of MATHS you use.

A weighted average or a high/low/average pricing model would give a far better indication.