Cold Test Results

For the testing of PSUs, we are using high precision electronic loads with a maximum power draw of 2700 Watts, a Rigol DS5042M 40 MHz oscilloscope, an Extech 380803 power analyzer, two high precision UNI-T UT-325 digital thermometers, an Extech HD600 SPL meter, a self-designed hotbox and various other bits and parts. For a thorough explanation of our testing methodology and more details on our equipment, please refer to our How We Test PSUs - 2014 Pipeline post.

As expected, the performance of the 650W and the 750W versions of the Onyx is very similar, with the performance curves only slightly leaning to the right of the power axis for the higher wattage model. Both units have exactly the same nominal load range (20%-100%) efficiency, which is 88%. The energy conversion efficiency is excellent for 80Plus Bronze certified units, with remarkably high efficiency readings at lower and higher loads, but the peak efficiency of the units was not high enough to warrant them a better efficiency rating.

The internal operating temperatures of the Riotoro Onyx PSUs are very reasonable,  considering the power output and efficiency rating of the units. Their thermal control appears to be very simplistic, with the fans adjusting their speed according to the internal temperature of the units. The fan of the 650W model displays a “ladder” behavior, increasing its speed in steps in relation to the load, where the fan’s speed in the 750W model increases slightly more linearly. This is of no practical consequence to the users though as the sound pressure levels are about the same for an equal power load. 

External and Internal Design Hot Test Results: Hot Box at 45-50ºC


View All Comments

  • Cellar Door - Wednesday, March 29, 2017 - link

    Nobody cares that you had a PSU fail - there are plenty of new companies that make great products and there are plenty of old companies that make bad products.

    Aaannnndd nobody cares what you are going to do.
  • t4murphy - Sunday, March 26, 2017 - link

    Same here. Been using my 750 silencer for 9 yrs. Couple SLI builds and MoBo no. 5 still going strong Reply
  • BrokenCrayons - Friday, March 24, 2017 - link

    Aw, be nice to them. They get a lot of unwarranted abuse from people already. Reply
  • bigboxes - Friday, March 24, 2017 - link

    Yeah, I bought a FSP (Fortran) 700 watt power supply in 2006 for my first dual core build. I could've swore that AnandTech had recommended the damn thing. Can't find a link to support that. Guess my memory is going. I tried out FSP as alternative when Antech was going through their QC issues. That PSU lasted maybe a year and then died in a puff of smoke. Was so pissed. Lot of money down the drain. I had bought a Seasonic PSU for my file server and liked it enough that I went to Seasonic for my main rig on my next build. I spend the money now for reliability as my boxes remain up 24/7. No more unknown cheap knock off brands for boxes. Reply
  • kn00tcn - Wednesday, March 29, 2017 - link

    fsp is known to be one of the highest quality makers together with seasonic, both of them are suppliers to other brands like corsair

    you people with your ignorance & single model failure stories that somehow represent an entire brand...
  • lefizz - Monday, March 27, 2017 - link

    I bought a Tagan when they first came out probably based on the review. 18 months laters in took out my motherboard and cpu. Probably £400 worth of damage, luckily i was next to it when it failed or else i could have started a fire.

    On a side not i was working in an studio in London in the early 2000s and a few of the Dell workstations went up in smoke in spectacular fashion from damaged USB ports on the front. On one occasion one literally filled the studio with smoke in the 5 seconds it took for me to turn around and turn the machine off. It was a big studio, like a smoke bomb going off.
  • rocky12345 - Thursday, March 30, 2017 - link

    I. had the same 700 watt Tagan was a great PSU. 2007-2014 I got my money's out of it. Heck it powered a Q6600@3.8Ghz Geforce 9800 GT ,Radeon HD 4870x2,Georce 580 & 680 GTX cards all overclocked over the years. What finally got it was a spider crawled into it and zap a flash of light and system went off.

    I took the cover off that's where I seen fried spider lol. I cleaned it up plugged it into a junker system and it worked fine. It worked two more years after the short out. Actually it still works I just upgraded to a newer unit. So either you got a bad unit or it was always running hot as the fans where small on those units. I was bummed when I found out they were gone.
  • DanNeely - Thursday, March 23, 2017 - link

    BFG lasted from 2002 to 2010,when bought your card (2008 or later) they were a well established and highly recommended company at time; not an unknown upstart. And if their lifetime warranty was extravagant EVGA and XFX were both offering IIRC 5 years at the time. Both of the latter reduced theirs because of, and IIRC BFG was broken by, the rise of GPU compute and Crypto Coin mining resulting in much higher lifetime load times and consequently much higher than budgeted for failure rates. Reply
  • DanNeely - Thursday, March 23, 2017 - link

    The really long warranties were also semi-supported for by nVidia insisting on card makers taking large numbers of low end GPUs to get full access to the premium parts. This stuck them with more budget cards than they could easily sell, a number of them were used up providing warranty replacements for similar performing older high end cards. eg my GTX6800 was replaced with a GT8500 when 3 or 4 years old. Reply
  • feelingshorter - Thursday, March 23, 2017 - link

    I owned a BFG 6600GT and that dates back to 2005 and then bought a 8800GT as well, which was before compute or CUDA cores. Wasn't until the 8800GTX that CUDA cores even came on the market. I don't even remember any notable use for compute back then, even wikipedia says Bitcoin was founded in 2009 and earliest reviews for the 8800GTX was 2006. Feels like BFG failed due to the warranties because GPUs start to fail en mass right after 2 years more so than over usage of compute for mining. Reply

Log in

Don't have an account? Sign up now