Sorry to be replying so late, holidays plus a ski trip have mostly kept me away from the forum. I'm going to add my answers in the quote in blue.
I'll break this up to make it easier to follow. I'll leave the discussion alone after this post.
Fans were not powered externally, and I believe that to be disingenuous since no one will use a second power supply to power the main power supply's fan(s) in the real world. Most server PSU fans I've come across are 5-10w. These PSUs definitely use more juice with their fans. Don't think it's enough to bring efficiency down by any meaningful amount, much less 4.3%. Where did you get that figure?
The 2880W PSU uses 3 60x38mm fans that are rated for 1.68A each. At the PSU's typical 12.2V output, that would be maximum of 20.5W each, for a total of 61.5W. @ your 50% load testing of 1440W, that could account for a margin of error up to 4.27% if they were at max load. I don't think following the testing procedures for 80+ should be considered disingenuous if you are going to use their outcome/ratings in your assessment of the PSU. If anything, I think it's disingenuous for you to use their rating label if you are going to ignore their protocol whether you agree with it or not. I'm not here to debate the real-world meaningfulness of Ecova's rating system, that is for the end user to decide for themselves just as EPA ratings on new vehicles (and not real-world fuel economy) are up to the consumer to interpret. What I wouldn't do though is tell the dealer that the EPA rating on my new car is wrong based on my own fuel economy, if I'm not willing to replicate the EPA's testing procedure myself.
I do agree that the cooling setup used by the 2880W PSU's is not very efficient, it relies on small diameter high RPM fans to push air through a long narrow tunnel filled with components. Most ATX PSU's use very large single ~0.6A fan (~7W) to cool them, which is why they wouldn't have had near the impact in your testing in the past. Thankfully MarkAZ has nearly finished his acrylic solution to replace the loud stock fans with a single 120mm fan which will reduce noise and wasted power from the 3x screamers.
These PSUs do use a very large amount of power for cooling vs typical PSUs. That said, what matters to the end user is efficiency at the wall. The fans being powered needs to be included in that efficiency figure when miners calculate power requirements. Of all the PSUs I've tested, none of them were measurably affected by their fan's power consumption vs the ecova testing results. I know this case is special considering the 3 screamers, but it is what it is.I've already described my testing methods to you. 20-25c ambient, Watts Up? Pro .net AC meter, $300 AC/DC amp clamp to measure DC amperage and multiply it by voltage measured at the load end of wires to account for vdroop. Some simple math then gives us efficiency numbers. IF you don't want to accept my methodology, I'm very open to suggestions to improve my process. It has proven to be very accurate when used with PSUs that have results posted by ecova, even when taking the fan into consideration. Furthermore, a 1500GH/s SP20E has consistently used less power at the wall using an actual platinum PSU vs the 2880W-er by a large enough amount. I do consider that a little unfair since 1 SP20E won't hit 50% load on the 2880w, but on the other hand the other PSU was close to 100% load which lowers it's efficiency.
Where do you put the DC amp clamp to measure the DC load by the miners, do you have to take 1 strand of a PCIe cable or can you group them all in together? I'm honestly curious about that, seems interesting as a way to compare miner efficiency vs manufacturers published specs. And to see if for example the SP20e's DC Watts are accurate from the GUI.
SP20 @ 1500 GH/s is only 900W DC IIRC? So your 87% efficiency was measured at that load or at true 50% load? This is the stuff I'm mainly interested in about your testing method (and I apologize if you've already posted about it) all I read was that you tested 87% efficiency at 50% load, which would have included fan draw of up to 60.5W of that. However, if it was the SP20E @ 1500GH/s, load would be a fair amount less than 50%. Also, depending on what the fans were drawing and the DC load, the fans could make up more than the 4.27% figure of the energy consumption if it was less the 1440W load I thought it was. I'm very interested if you managed to test it at different loads to see if it was consistent? The real-world impact of this is that fans in general would make up a significantly lower portion of the wasted power (reflected as inefficiency in your test) at 100% load than they would at 20% load. This would also skew the results when comparing the efficiency of a PSU @ 100% load to a larger PSU @ 30% load, which is what I interpreted you as saying.
If you are open to suggestions, what I would love to see is a straight AC draw comparison vs a known similar 80+ platinum PSU at 50% and 100% DC load. I have 2x 2980W PSU's coming and would be happy to lend one to you, along with another 2880W PSU for redundancy to test and show your results here. If you have enough mining power to load it up that is. Since it's impossible for anyone to make a definitive conclusion without actually following Ecova's guidelines, I think a comparison vs a near identical and known tested 80+ platinum PSU like the 2980W would be as close as one could get.
The amp clamp goes as close to the miner's connectors as possible with all 12v strands bundled together.
My SP20E testing goes back at least 6 months, so I can't remember exactly and as I said I've lost my documentation of the results unfortunately. I do plan on testing another SP20E soon enough, though, to compare 2 smaller platinum server PSUs I'm using. I'll get back to you with SP20E figures then. I'd be happy to test your PSUs, but after taking a quick glance through the thread before replying I think Phil has that covered. What you call overnight is actually a 5 to 6 year difference in engineering and manufacturing, a new (better) OEM, not to mention the adoption of PSU efficiency measures to reduce global power consumption of data centers and consumer IT and a shift to ever more efficient PSUs to meet those goals. The 2880w came before all of that. Don't see why you have such difficulty believing that it's possible. You'll have to forgive me for not taking your word that it's platinum when there is absolutely no proof of that anywhere to be found. I've only seen that it's platinum being parroted on this forum. I hope you'll do your own tests and post the results. Until then, the burden of proof is on you to prove it is platinum and not bronze.
What you're saying about the timeline makes sense and could absolutely be right, it just didn't seem to me like much was changed through the years with the 2880W PSU's. And if it had, that it wouldn't have happened all at once with a new model. All I am interested in is the truth of this. And what you've
showntold me is not exactly compelling, just that you tested the PSU @ 50% (or possibly 30% load if it was a single SP20, which i'm hoping you can clarify) and claim Bronze-level efficiency. But you didn't follow 80+'s guidelines to determine that, and are now arguing that those guidelines are disingenuous...
The 2880w has nothing to do with the 2980w. The only thing they have in common are their exteriors and interfacing connector. IBM commissioned Delta to make an updated, slightly more powerful, and significantly more efficient power supply than Astec was providing them with in regards to the 2880w. Changes made to the 2880w PSU between the first revision and the last have absolutely no bearing on what has been done with the 2980w, especially since it's not the same company manufacturing them.I hope you aren't taking this as a knock on you or j4bber. It is anything but. You guys sell what I consider to be the best server equipment for mining, and I know that you both believed the PSUs to be platinum rated. I know you wouldn't have sold them as platinum power supplies without believing that they were, and no one in their right mind will think that.
Edit: With the exception of S4, what I've seen is a constant load from miners on the power supply. Never seen it move more than 0.1A while mining on S7, for example.
I realize I may be coming off as abrasive, and for that I apologize. The truth is I'm not taking this personally, I'm just learning now about how they actually rate PSU's, and learned how much wasted energy the fan packs on the 2880's could be burning (which makes me hopeful to test out and offer MarkAZ's alternate acrylic panel to mount a single 120mm fan instead). Now what I would like to know is if these PSU's would actually meet the ratings that I was under the impression they met, and I believe the best way to do that (if you are willing) would be to have you compare the 2880W to the near identical and tested platinum 2980W PSU, to take your methods/equipment out of the equation (even if the method/equipment isn't perfect, it should be consistent enough to show a difference as significant as this). If they do not match, I will be more than happy to change the wording in my threads. The last thing I or J4bberwock are trying to do is to deceive anybody. I started selling the product because I believed it was the best solution on the market, and still believe that today.
I have a bunch of server PSU's I could lend you for your PSU efficiency thread you mentioned too, DPS 2000's, DPS 750's, 800 GBA's, etc.
I'm absolutely willing to test both the 2880w vs the 2980w for you if you want. I'm 99% sure the 2980W-er will pull about 6-7% less than the 2880W-er at the wall, which is the only thing that counts IMHO. I have the means to load both PSUs anywhere from 20%-100%. Using a single 120mm fan will definitely help the 2880w PSU to consume less at the wall, but the same is true of the 2980w.
Thanks for the offer on the PSUs, but I already have quite a few units to test which might take a while depending on my work load. I'm not looking to become the next jonnyguru or anything, nor do I have the necessary skills, equipment, or knowledge to do so. I may consider taking requests for reviews of specific units that I don't have on hand eventually, but the unit would have to be provided by the requester if I don't have it on hand. Doubt that will happen, though.