Maybe i missed it, but what about security features like OCP and so on? From the "power specifications" on page 1 i would asume the PSU to be a multi-rail device, of course. But is it truly a two-rail design or is it split even more?
I'm curious why your full load test pulled so much current on the 3.3/5V rails. You've got them both over 90% individually; but only have the 12v rail at 78% load. Any realistic load near the full output would be skewed much more heavily toward maxing out 12V. On the mobo, USB ports are the only significant user of +5V; and I don't think there're any major users of 3.3V at all. (PCIe cards are allowed to draw upto 10W of it; but AFAIK the only ones that did were transitional models that combined an existing legacy-PCI design with a bridge chip.)
More troubling is that you're drawing a total of 193W of 3.3/5V power but the PSU is only rated to give 150W combined on those rails. While it obviously didn't cause anything to fail or go out of spec; but going nearly 30% out of spec is troubling. The only justification for doing so deliberately that I can think of is if your tester couldn't go above 110A on the 12V rail; although in that case I think it should've been called out explicitly in the article.
Actually, I have a much better justification for that.
The only official testing procedures that exist, which are given in the methodology article, result to these minimum loads. If you had checked this article, you would see that my tester can go up to 240A on the 12V alone anyway.
I cannot lower the load to the 3.3V/5V buses without inventing my own testing methodology, in which case I would be rendering all comparisons between reviews useless and misleading. This is a general problem with very high output PSUs, the specifications go down the drain and fail to meet even basic certification standards, all in the name of massive power output. Perhaps I will consider "circumventing" this issue in my future reviews by forcing a divider once the load exceeds 1000W.
Is there any hardware which could even use this amount of power?
Let's say some OC'd 8 core Intel i7 or dual CPU 18 core xeons, 4x crossfired 7990s under full mining load with a 8 drive raid 5 15k RPM SAS setup, let's add some hardcore water cooling and every device in the house charging off USB.
Even without overclocking even a 3 GPU system could exceed 1kw. There've been cards with rated TDPs of at least 300W; IIRC seeing 375 for something but that might've been a 3rd party dual GPU card not an official design since IIRC the PCIe spec tops at 300W.
4x300W cards is 1200W, add an overclocked CPU fast ram and all the other odds and ends in a case and you could hit >1400W easily enough. (According to CPUz, my 4790k hit 145-150W at 4.7ghz with 32gb of DDR3-2400 ram under some prime95 FMA stress testing loads.)
You can't crossfire 4 7990s, as they are dual chip already, so 2 7990s is quad crossfire, and would use less power than 4 7970s.
I used to have an overclocked i7 920 (130W TDP, pushed to 4GHz so probably ~180W), 12GB ram, a few hard drives, 7970+7990 trifire, and absolute max I'd use (corsair link on an AX1500i) was ~880W.
Now, with 4970k and 980SLi, I struggle to hit ~475W.
The Powercolor Devil13 (2x R9 290X) can pull loads in excess of 450W (and clear excess of the ATX specifications), so a crossfire setup with two of those and an AMD FX-9590 can exceed 1100W on CPU & GPU alone. Add Memory and a bunch of Harddrives, and maybe a dozen electronic gadgets on the USB ports, and you can use the 1500kW in extreme load cases.
Not saying you should, an Intel 2011-3 system with two Titan X will give you more gaming power in barely over 600W peak power consumption.
One overclocked 290X hit 400 watts in Tom's testing. It was only a very brief spike, but those spikes should be factored in. So, three of those plus a 9590 and you're talking 1465 watts.
I would appreciate a different direction from the psu makers. With today's low power systems idling at 15-30W it would make more sense to focus on low load efficiency and smaller total power output, while maintaining the same quality components found in higher powered models.
People always overbuild their PSU's. They go for watts over quality. It's a consumer misunderstanding of power requirements and efficiency.
Most OEM systems come with ~200-watt PSU's these days. Even high-end workstations like HP Z and Dell Precisions have 280-watt PSU's.
My Xeon workstation with a GTX970 has a 400-watt PSU and my Overclocked i7 SLI gaming PC has a 550-watt PSU, and both have headroom (peak draw from the wall for each under OCCT\furmark is 363-watts and 509-watts.)
I'm not saying a product like this has no reason to exist, but let's just say it's for "the 1%."
The most recent 80+ spec, Titantium does focus on low load efficiency by requiring 90% efficiency at a 10% load. For all but the most severe cases of mis-matched system power and PSU size this will bring idle loads for systems with dGPUs into the high efficiency operating range.
If OEMs actually refresh their low power models the same should be possible for IGP only boxes; but the negligible number of 80+ units of any size in that range make me suspect we'll be waiting a while for that to happen. Accommodating Haswell's deeper sleep states might push more of them to refresh their designs over the next year or two; but only 9/23 haswell compatable <400W models on newegg are gold certified (none are platinum). However the additional 432 non-haswell certified PSUs in that power bracket make it clear that this is going to be a very slow process. TBH for those systems actual amounts of power saved are small enough that it's only at very high power prices that a high efficiency PSU will pay for itself over a reasonable time horizon. Unless forced by regulators, I wouldn't hold my breath for any significant movement in the small PSU segment.
Fools and their money hey... there are some people who purchase merely for the wattage number on the PSU. Says nothing of the quality and even less of the actual requirement. I wonder if this was a hangover from the days where a cheap 300W POS, er PSU would struggle to output 200W without exploding. I peaked 350 watts with a pretty hefty system (GTX 680, overclocked i5-3570K, 2 HDDs). That's 250 watts of headroom from my "low end" 600W PSU. Things might be different in AMD world but Intel/NVIDIA just keep getting more and more PPW.
I didn't read the full article, so I may be restating, but for all the people asking what you need 1500W for, the efficiency of these usually peaks at about 50 - 70% of rated power if I remember right. While even a 750 W pull is a pretty hefty system, there might be a slight market for the eco-conscious tech enthusiast...
Okay, so maybe the audience is still only about 4 people.
You obviously didn't read much if any of the article: You're way off.
This is an 80+ gold unit which means that it needs to be >88% efficiency from 20-100% of full load (@220v, >87% at 110). It managed >90, and peaked at 92% efficient at a 50% load.
Sub 70% efficient power supplies are designs from a decade or more ago; and are long gone from the market (possibly excepting some of the sketchiest no-name models); bargain basement 80+ units are widely available now.
Lastly, the power rating is output power, not input. A hypothetical 50% efficient 1500W PSU would still output 1500W of power; it would however draw 3000W at the intake.
reininop: "the efficiency of these usually peaks at about 50 - 70% of rated power" DanNeely: "You obviously didn't read much if any of the article: You're way off..[snip].. and peaked at 92% efficient at a 50% load."
I'd say reininop was spot on. They were stating that efficiency is at its best at 50%-70% load, not that the PSU is 50%-70% efficient.
You know, I typically read the intro and the final words and then if I find it interesting I go back and read the rest. In this case, I read some of the comments first as well. I have since went back and saw your power efficiency graph. I felt the comment was meant to be taking as a bit of a joke as I was pointing out a user case that almost certainly doesn't exist. For some reason, you seemed to take personal offense to it.
As for the rest of you comment, I am well aware of what the 80+ gold means. I never said anything about the conversion efficiency, only the peak rate of efficiency as a function of the output power. Hence, when I said tech enthusiast, I was attempting to imply someone who actually has components that require 750 W to operate. I assume the confusion arose from my use of the word "pull" when discussing the actual system power requirements and you assumed I meant the outlet power.
The problem for the eco-conscious tech enthusiast is that the efficiency of this power supply drops significantly when power consumption falls below 300 watts. So it may be efficient during heavy gaming, but not so much when you take a break to post comments on anandtech.com.
On the other hand, at idle load, an improvement in efficiency can make a difference of only a few watts, so it's understandable that the focus of 80plus has been on efficiency at higher loads where a little improvement can make a big difference. That was the low-hanging fruit. With that taken care of, they should now be improving the low end.
I'd just like to point out that anything above a 1440W PSU (assuming that's what they're pulling off the "mains") should really be on a 20A breaker/circuit. I know that 20A circuits are very common in the US - but they're far less common in Canada.
The reason for this is that in both Canada and the US (and Europe too?) the normal household breakers are only rated to sustain 80% of their rated capacity. Ergo a 15A breaker will only hold 12A and a 20A breaker will hold 16A. Respectively those are good for 1440W and 1920W.
It's good to see the C19 plug being used here as that should force people to use a 20A T-slot plug (15/20A Hybrid) that should be common in the US. It seems very few people are aware of this but it's an important consideration - especially if you plan on using this PSU to its full potential.
Consumer-grade 1500W PSU in year 2015? Buying one should be a freaking criminal offense -- together with multi-GPU setups and such. Really, people, wake the frack up!
Criminal offense? How about asking people who just dropped 800 or more dollars on monitors (which you can still buy today) to replace them because the manufacturer isn't going to bother updating its firmware to support Freesync?
Throwing this out there, I've owned a SilverStone Zeus 850W PSU for more than 8 years now... Switched 4 PC configurations on it. There were periods of more than 2 years without cleaning it. About an year ago I measured it extensively, I literally had to bend down in an efford to hear the fan, and all I have to say about my (first) experience with the brand is rounded up in one word - immaculate.
"Albeit crude, the assembly quality certainly has room for improvement but it does not cause reliability concerns. "
It might not cause concerns for you, but it sure does for ME.
I think we need a new set of standards to judge PSUs by. Long gone are the days when Tomshardware and others demonstrated that most PSUs designed for desktop computers were simply bad and it didn't take much digging to find out the truth of the matter. Now brand name companies and manufacturers up and down the line understand how to pass tests in reviews and build their PSUs to do so. Good news, we are less likely to buy one that fries our build in its lifetime.
Bad news, PSUs like this one slip in to the category of "acceptable" along with far better built devices. Name brands like Silverstone hide the fact that they farm their business out to the lowest bidder and some of their PSU lines are far from good or reliable. A bit of sloppy work here or there might not seem all that important until you think about how time and use magnify those weaknesses and can easily cause them to become real problems. And with a PSU, "real problems" can translate to a fried mainboard, CPU or other components that get taken out when it self-destructs.
Assembly quality is where engineering meets the real world, and a PSU is only as good as its weakest solder joint.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
32 Comments
Back to Article
malkolm - Tuesday, April 7, 2015 - link
Maybe i missed it, but what about security features like OCP and so on?From the "power specifications" on page 1 i would asume the PSU to be a multi-rail device, of course. But is it truly a two-rail design or is it split even more?
DanNeely - Tuesday, April 7, 2015 - link
I'm curious why your full load test pulled so much current on the 3.3/5V rails. You've got them both over 90% individually; but only have the 12v rail at 78% load. Any realistic load near the full output would be skewed much more heavily toward maxing out 12V. On the mobo, USB ports are the only significant user of +5V; and I don't think there're any major users of 3.3V at all. (PCIe cards are allowed to draw upto 10W of it; but AFAIK the only ones that did were transitional models that combined an existing legacy-PCI design with a bridge chip.)More troubling is that you're drawing a total of 193W of 3.3/5V power but the PSU is only rated to give 150W combined on those rails. While it obviously didn't cause anything to fail or go out of spec; but going nearly 30% out of spec is troubling. The only justification for doing so deliberately that I can think of is if your tester couldn't go above 110A on the 12V rail; although in that case I think it should've been called out explicitly in the article.
E.Fyll - Thursday, April 16, 2015 - link
Actually, I have a much better justification for that.The only official testing procedures that exist, which are given in the methodology article, result to these minimum loads. If you had checked this article, you would see that my tester can go up to 240A on the 12V alone anyway.
I cannot lower the load to the 3.3V/5V buses without inventing my own testing methodology, in which case I would be rendering all comparisons between reviews useless and misleading. This is a general problem with very high output PSUs, the specifications go down the drain and fail to meet even basic certification standards, all in the name of massive power output. Perhaps I will consider "circumventing" this issue in my future reviews by forcing a divider once the load exceeds 1000W.
Pissedoffyouth - Tuesday, April 7, 2015 - link
Is there any hardware which could even use this amount of power?Let's say some OC'd 8 core Intel i7 or dual CPU 18 core xeons, 4x crossfired 7990s under full mining load with a 8 drive raid 5 15k RPM SAS setup, let's add some hardcore water cooling and every device in the house charging off USB.
Would this even push a kilowatt?
MobiusPizza - Tuesday, April 7, 2015 - link
The only case you are even anywhere close to using >1kW is with quad SLI/Crossfire and top end GPUs.DanNeely - Tuesday, April 7, 2015 - link
Even without overclocking even a 3 GPU system could exceed 1kw. There've been cards with rated TDPs of at least 300W; IIRC seeing 375 for something but that might've been a 3rd party dual GPU card not an official design since IIRC the PCIe spec tops at 300W.4x300W cards is 1200W, add an overclocked CPU fast ram and all the other odds and ends in a case and you could hit >1400W easily enough. (According to CPUz, my 4790k hit 145-150W at 4.7ghz with 32gb of DDR3-2400 ram under some prime95 FMA stress testing loads.)
3DVagabond - Saturday, April 11, 2015 - link
Try and get a quad sli/crossfire setup to load all 4 gpu's 100%. Not going to happen.hammer256 - Tuesday, April 7, 2015 - link
4x GTX690, a -E platform, and lots of fans would do the trick ;) Not for gaming though.hammer256 - Tuesday, April 7, 2015 - link
The PSU used for that system is a LEPA G1600, a bit cheaper than this one.rtho782 - Wednesday, April 8, 2015 - link
You can't crossfire 4 7990s, as they are dual chip already, so 2 7990s is quad crossfire, and would use less power than 4 7970s.I used to have an overclocked i7 920 (130W TDP, pushed to 4GHz so probably ~180W), 12GB ram, a few hard drives, 7970+7990 trifire, and absolute max I'd use (corsair link on an AX1500i) was ~880W.
Now, with 4970k and 980SLi, I struggle to hit ~475W.
ShieTar - Thursday, April 9, 2015 - link
The Powercolor Devil13 (2x R9 290X) can pull loads in excess of 450W (and clear excess of the ATX specifications), so a crossfire setup with two of those and an AMD FX-9590 can exceed 1100W on CPU & GPU alone. Add Memory and a bunch of Harddrives, and maybe a dozen electronic gadgets on the USB ports, and you can use the 1500kW in extreme load cases.Not saying you should, an Intel 2011-3 system with two Titan X will give you more gaming power in barely over 600W peak power consumption.
Oxford Guy - Thursday, April 9, 2015 - link
One overclocked 290X hit 400 watts in Tom's testing. It was only a very brief spike, but those spikes should be factored in. So, three of those plus a 9590 and you're talking 1465 watts.jabber - Tuesday, April 7, 2015 - link
Handy review for the three or four that might need this.sweeper765 - Tuesday, April 7, 2015 - link
I would appreciate a different direction from the psu makers.With today's low power systems idling at 15-30W it would make more sense to focus on low load efficiency and smaller total power output, while maintaining the same quality components found in higher powered models.
darkfalz - Tuesday, April 7, 2015 - link
That describes most PSUs actually sold. The one in this review is strictly for e-peen who devote far too much of their income to their PCs.Samus - Tuesday, April 7, 2015 - link
People always overbuild their PSU's. They go for watts over quality. It's a consumer misunderstanding of power requirements and efficiency.Most OEM systems come with ~200-watt PSU's these days. Even high-end workstations like HP Z and Dell Precisions have 280-watt PSU's.
My Xeon workstation with a GTX970 has a 400-watt PSU and my Overclocked i7 SLI gaming PC has a 550-watt PSU, and both have headroom (peak draw from the wall for each under OCCT\furmark is 363-watts and 509-watts.)
I'm not saying a product like this has no reason to exist, but let's just say it's for "the 1%."
DanNeely - Wednesday, April 8, 2015 - link
The most recent 80+ spec, Titantium does focus on low load efficiency by requiring 90% efficiency at a 10% load. For all but the most severe cases of mis-matched system power and PSU size this will bring idle loads for systems with dGPUs into the high efficiency operating range.If OEMs actually refresh their low power models the same should be possible for IGP only boxes; but the negligible number of 80+ units of any size in that range make me suspect we'll be waiting a while for that to happen. Accommodating Haswell's deeper sleep states might push more of them to refresh their designs over the next year or two; but only 9/23 haswell compatable <400W models on newegg are gold certified (none are platinum). However the additional 432 non-haswell certified PSUs in that power bracket make it clear that this is going to be a very slow process. TBH for those systems actual amounts of power saved are small enough that it's only at very high power prices that a high efficiency PSU will pay for itself over a reasonable time horizon. Unless forced by regulators, I wouldn't hold my breath for any significant movement in the small PSU segment.
darkfalz - Tuesday, April 7, 2015 - link
Fools and their money hey... there are some people who purchase merely for the wattage number on the PSU. Says nothing of the quality and even less of the actual requirement. I wonder if this was a hangover from the days where a cheap 300W POS, er PSU would struggle to output 200W without exploding. I peaked 350 watts with a pretty hefty system (GTX 680, overclocked i5-3570K, 2 HDDs). That's 250 watts of headroom from my "low end" 600W PSU. Things might be different in AMD world but Intel/NVIDIA just keep getting more and more PPW.reininop - Tuesday, April 7, 2015 - link
I didn't read the full article, so I may be restating, but for all the people asking what you need 1500W for, the efficiency of these usually peaks at about 50 - 70% of rated power if I remember right. While even a 750 W pull is a pretty hefty system, there might be a slight market for the eco-conscious tech enthusiast...Okay, so maybe the audience is still only about 4 people.
DanNeely - Tuesday, April 7, 2015 - link
You obviously didn't read much if any of the article: You're way off.This is an 80+ gold unit which means that it needs to be >88% efficiency from 20-100% of full load (@220v, >87% at 110). It managed >90, and peaked at 92% efficient at a 50% load.
Sub 70% efficient power supplies are designs from a decade or more ago; and are long gone from the market (possibly excepting some of the sketchiest no-name models); bargain basement 80+ units are widely available now.
Lastly, the power rating is output power, not input. A hypothetical 50% efficient 1500W PSU would still output 1500W of power; it would however draw 3000W at the intake.
Gigaplex - Tuesday, April 7, 2015 - link
reininop: "the efficiency of these usually peaks at about 50 - 70% of rated power"DanNeely: "You obviously didn't read much if any of the article: You're way off..[snip].. and peaked at 92% efficient at a 50% load."
I'd say reininop was spot on. They were stating that efficiency is at its best at 50%-70% load, not that the PSU is 50%-70% efficient.
reininop - Wednesday, April 8, 2015 - link
You know, I typically read the intro and the final words and then if I find it interesting I go back and read the rest. In this case, I read some of the comments first as well. I have since went back and saw your power efficiency graph. I felt the comment was meant to be taking as a bit of a joke as I was pointing out a user case that almost certainly doesn't exist. For some reason, you seemed to take personal offense to it.As for the rest of you comment, I am well aware of what the 80+ gold means. I never said anything about the conversion efficiency, only the peak rate of efficiency as a function of the output power. Hence, when I said tech enthusiast, I was attempting to imply someone who actually has components that require 750 W to operate. I assume the confusion arose from my use of the word "pull" when discussing the actual system power requirements and you assumed I meant the outlet power.
KAlmquist - Wednesday, April 8, 2015 - link
The problem for the eco-conscious tech enthusiast is that the efficiency of this power supply drops significantly when power consumption falls below 300 watts. So it may be efficient during heavy gaming, but not so much when you take a break to post comments on anandtech.com.wbwb - Wednesday, April 8, 2015 - link
On the other hand, at idle load, an improvement in efficiency can make a difference of only a few watts, so it's understandable that the focus of 80plus has been on efficiency at higher loads where a little improvement can make a big difference. That was the low-hanging fruit. With that taken care of, they should now be improving the low end.Poik - Tuesday, April 7, 2015 - link
I'd just like to point out that anything above a 1440W PSU (assuming that's what they're pulling off the "mains") should really be on a 20A breaker/circuit. I know that 20A circuits are very common in the US - but they're far less common in Canada.The reason for this is that in both Canada and the US (and Europe too?) the normal household breakers are only rated to sustain 80% of their rated capacity. Ergo a 15A breaker will only hold 12A and a 20A breaker will hold 16A. Respectively those are good for 1440W and 1920W.
It's good to see the C19 plug being used here as that should force people to use a 20A T-slot plug (15/20A Hybrid) that should be common in the US. It seems very few people are aware of this but it's an important consideration - especially if you plan on using this PSU to its full potential.
Gigaplex - Tuesday, April 7, 2015 - link
Or, like the bulk of the rest of the world, it should run at ~240V.Pissedoffyouth - Wednesday, April 8, 2015 - link
Hmm really puts in perspective how lucky we are to have 240v in rest of worldex_User - Wednesday, April 8, 2015 - link
Consumer-grade 1500W PSU in year 2015? Buying one should be a freaking criminal offense -- together with multi-GPU setups and such. Really, people, wake the frack up!Oxford Guy - Thursday, April 9, 2015 - link
Criminal offense? How about asking people who just dropped 800 or more dollars on monitors (which you can still buy today) to replace them because the manufacturer isn't going to bother updating its firmware to support Freesync?ZeliaS - Wednesday, April 8, 2015 - link
Throwing this out there, I've owned a SilverStone Zeus 850W PSU for more than 8 years now... Switched 4 PC configurations on it. There were periods of more than 2 years without cleaning it. About an year ago I measured it extensively, I literally had to bend down in an efford to hear the fan, and all I have to say about my (first) experience with the brand is rounded up in one word - immaculate.Sabresiberian - Wednesday, April 8, 2015 - link
"Albeit crude, the assembly quality certainly has room for improvement but it does not cause reliability concerns. "It might not cause concerns for you, but it sure does for ME.
I think we need a new set of standards to judge PSUs by. Long gone are the days when Tomshardware and others demonstrated that most PSUs designed for desktop computers were simply bad and it didn't take much digging to find out the truth of the matter. Now brand name companies and manufacturers up and down the line understand how to pass tests in reviews and build their PSUs to do so. Good news, we are less likely to buy one that fries our build in its lifetime.
Bad news, PSUs like this one slip in to the category of "acceptable" along with far better built devices. Name brands like Silverstone hide the fact that they farm their business out to the lowest bidder and some of their PSU lines are far from good or reliable. A bit of sloppy work here or there might not seem all that important until you think about how time and use magnify those weaknesses and can easily cause them to become real problems. And with a PSU, "real problems" can translate to a fried mainboard, CPU or other components that get taken out when it self-destructs.
Assembly quality is where engineering meets the real world, and a PSU is only as good as its weakest solder joint.
Oxford Guy - Thursday, April 9, 2015 - link
35 dB at 250 watts, 38 dB at 550 watts, 39 dB at 650 watts...