PSUs for 6800 vs 6800xt?

[terminalinfinity]

So hopefully I'll be able to snag one of the 6800s or 6800xts at launch. What's confusing me is the power supply requirement.  If it has a 300 watt power draw, lets say 400 watts with power spikes and overclocking, and your otherwise total system power draw with CPU is about 120-150 watts, leading to a ~550 watt max system draw, why the 750 watt PSU requirement?

 

The 6800 reference design has a 650 watt PSU recommendation, the 6800xt reference has a 750 watt PSU recommendation

 

The Sapphire 6800xt has a 800 watt PSU recommendation!

 

So my main question out of all of this is, should my Silverstone ET700-MG be enough to run a 6800xt?  Or should I temper my expectations to a 6800?

https://www.eteknix.com/silverstone-et700-mg-power-supply-review/7/

I should note that in this test my PSU model had over 500 watts of headroom, safely providing up to 1281 watts of power.

[Chua]

I was just researching this today for my planned build with a 6800!

 

I found these two seasonic psu's for a pretty reasonable price with good reviews. The gx-650 has more ports and is fully modular. The gm-650 is a bit cheaper with fewer ports and is only partly modular.

https://www.newegg.com/seasonic-focus-plus-650-gold-ssr-650fx-650w/p/N82E16817151186?Description=seasonic focus gx-650&cm_re=seasonic_focus gx-650-_-17-151-186-_-Product

https://www.newegg.com/seasonic-focus-650-gold-ssr-650fm-650w/p/N82E16817151202?Description=gm 650&cm_re=gm_650-_-17-151-202-_-Product&quicklink=true

hope this helps

[mariushm]

They consume less than nvidia cards, around 250-300 watts.

So add the power consumption of the cpu and hard drives and other things to 250-350w and you got your wattage.

If you combine it with a quad core Ryzen that sips power, you could do with a 500w psu. If you have a 8-12 core processor and several hard drives, a 650-750w psu may be needed.

 

They recommend higher wattage power supplies so that people won't combine them with shitty power supplies, like no-name chinese 500-750w that in reality can barely do 2/3 of that.

[Letgomyleghoe]

4 minutes ago, mariushm said:

They recommend higher wattage power supplies so that people won't combine them with shitty power supplies, like no-name chinese 500-750w that in reality can barely do 2/3 of that.

its not that they cant do the rated watts its that the rails cant, alot of the problems people are having are psu's with multiple rails, and the rail for pci power can only output ~300-400w so they crash

[terminalinfinity]

1 minute ago, Letgomyleghoe said:

its not that they cant do the rated watts its that the rails cant, alot of the problems people are having are psu's with multiple rails, and the rail for pci power can only output ~300-400w so they crash

Thanks.  That was what I was looking for.  So a single rail PSU means that single rail can do the rated wattage?  Sorry not good with the technical stuff.

[Chiyawa]

27 minutes ago, terminalinfinity said:

Thanks.  That was what I was looking for.  So a single rail PSU means that single rail can do the rated wattage?  Sorry not good with the technical stuff.

Yes.

[OrionFOTL]

8 hours ago, Letgomyleghoe said:

its not that they cant do the rated watts its that the rails cant,

You're saying there are no bad PSUs that can't output their rated power, and that there are only bad PSUs with bad multiple rail config. That's not true.

What you're describing is a very valid issue, and certainly relevant to this topic, but cheap power supplies with completely fake, inflated power ratings (like a PSU with output diodes rated for 250W total, yet the sticker on the PSU claiming it can do 500W) is a far more widespread problem than PSUs which are capable of outputting their full rated power, except having poorly designed OCP limits on their 12V rails. Let's not forget the threat of power supplies rated at 2x higher than what their internal components can do.

8 hours ago, terminalinfinity said:

So a single rail PSU means that single rail can do the rated wattage? 

No, it simply means there is one 12V rail - there are no specific implications about rated wattage beyond that. Good PSUs with mutliple 12V rails can also output their full power solely on 12V. They just have safety power limits on certain sets of connectors, for example: the ATX connector and one PCIe connector are limited to 420W max, another two PCIe connectors are limited to 420W, and all the SATA and peripheral cables are limited to 300W max from 12V.

Some PSUs with multiple 12V rails have those limits set too low for the top end cards with their huge power spikes. In some cases, those huge power spikes are enough to trip one of those safety power limits on certain sets of cables, even if the power spikes don't reach the total combined rated power of the PSU.

Single 12V rail power supplies don't have any safety limits on sets of cables, so there is nothing to trip for the spikes. (Except the total power limit.)

 

9 hours ago, terminalinfinity said:

So my main question out of all of this is, should my Silverstone ET700-MG be enough to run a 6800xt?  Or should I temper my expectations to a 6800?

Unfortunately we can't really tell if your PSU is good for that or not, for the simple reason that there are no reviews of it. The Eteknix page you linked to can't be called a review, because none of the tests are performed correctly.

The voltage regulation test is the most hilarious one, and also the worst offender - their way of calculating it makes every PSU enthusiast scratch their head. It's like they only had a vague idea that voltage regulation is related to measuring output voltages and arriving at a nice percentage score, like other reviewers do, but they didn't know what it actually is or how it is calculated, so they began throwing random formulas at it, until they found one that generates nice looking % scores. Unfortunately, their measure of "average deviance" is completely useless and absolutely meaningless in terms of the PSU's performance. And indeed that's not the only thing wrong with those articles.

The reviews on that website are one of the worst I've seen in terms of "being terrible, yet, at first sight believable to outsiders". In my experience I notice mainly two categories of PSU reviewing sites: ones which are pretty blatantly bad (the ones who just measure voltages using HWInfo on a real PC running a benchmark), and sites which do solid, correct tests in several most important categories but can't afford the equipment to do a full suite like Tom's Hardware. Eteknix sits in a bizarre category of their own, since they post numbers and tables that look like a real review (some familiar numbers with percents, some oscilloscope screenshots, some graphs), but the way of getting and interpreting them is completely made up. They do have some measuring equipment - a load tester and an oscilloscope, they just seemingly don't know how to use them to actually get useful results. Reading commentary around the tested PSU's internal photos and the testing methodology page makes it obvious they're written by a person who just doesn't know much about testing power supplies, and power supplies in general.

 

I'm sorry for the rant, I realize it's not particularly useful to you. To the point: despite the review data being useless, we can tell from the internal photos of your PSU that it's a well-built high end one, and it shouldn't have troubles powering anything up to its rated power. There's no reason to preemptively worry about it having issues powering a 6800XT. If it can't handle the card (like some PSUs with RTX3080) it will shut down, but there's a much higher chance it will simply work fine.

It's also good to wait for reviews of the cards, so we can see how much power they really consume and how big transient spikes they cause on PSUs.

 

 

[terminalinfinity]

26 minutes ago, OrionFOTL said:

You're saying there are no bad PSUs that can't output their rated power, and that there are only bad PSUs with bad multiple rail config. That's not true.

What you're describing is a very valid issue, and certainly relevant to this topic, but cheap power supplies with completely fake, inflated power ratings (like a PSU with output diodes rated for 250W total, yet the sticker on the PSU claiming it can do 500W) is a far more widespread problem than PSUs which are capable of outputting their full rated power, except having poorly designed OCP limits on their 12V rails. Let's not forget the threat of power supplies rated at 2x higher than what their internal components can do.

No, it simply means there is one 12V rail - there are no specific implications about rated wattage beyond that. Good PSUs with mutliple 12V rails can also output their full power solely on 12V. They just have safety power limits on certain sets of connectors, for example: the ATX connector and one PCIe connector are limited to 420W max, another two PCIe connectors are limited to 420W, and all the SATA and peripheral cables are limited to 300W max from 12V.

Some PSUs with multiple 12V rails have those limits set too low for the top end cards with their huge power spikes. In some cases, those huge power spikes are enough to trip one of those safety power limits on certain sets of cables, even if the power spikes don't reach the total combined rated power of the PSU.

Single 12V rail power supplies don't have any safety limits on sets of cables, so there is nothing to trip for the spikes. (Except the total power limit.)

 

Unfortunately we can't really tell if your PSU is good for that or not, for the simple reason that there are no reviews of it. The Eteknix page you linked to can't be called a review, because none of the tests are performed correctly.

The voltage regulation test is the most hilarious one, and also the worst offender - their way of calculating it makes every PSU enthusiast scratch their head. It's like they only had a vague idea that voltage regulation is related to measuring output voltages and arriving at a nice percentage score, like other reviewers do, but they didn't know what it actually is or how it is calculated, so they began throwing random formulas at it, until they found one that generates nice looking % scores. Unfortunately, their measure of "average deviance" is completely useless and absolutely meaningless in terms of the PSU's performance. And indeed that's not the only thing wrong with those articles.

The reviews on that website are one of the worst I've seen in terms of "being terrible, yet, at first sight believable to outsiders". In my experience I notice mainly two categories of PSU reviewing sites: ones which are pretty blatantly bad (the ones who just measure voltages using HWInfo on a real PC running a benchmark), and sites which do solid, correct tests in several most important categories but can't afford the equipment to do a full suite like Tom's Hardware. Eteknix sits in a bizarre category of their own, since they post numbers and tables that look like a real review (some familiar numbers with percents, some oscilloscope screenshots, some graphs), but the way of getting and interpreting them is completely made up. They do have some measuring equipment - a load tester and an oscilloscope, they just seemingly don't know how to use them to actually get useful results. Reading commentary around the tested PSU's internal photos and the testing methodology page makes it obvious they're written by a person who just doesn't know much about testing power supplies, and power supplies in general.

 

I'm sorry for the rant, I realize it's not particularly useful to you. To the point: despite the review data being useless, we can tell from the internal photos of your PSU that it's a well-built high end one, and it shouldn't have troubles powering anything up to its rated power. There's no reason to preemptively worry about it having issues powering a 6800XT. If it can't handle the card (like some PSUs with RTX3080) it will shut down, but there's a much higher chance it will simply work fine.

It's also good to wait for reviews of the cards, so we can see how much power they really consume and how big transient spikes they cause on PSUs.

 

 

Wow thanks for the insight.  Won't use that site again.  There's a third category of review sites: Ones that just look at the manufacturer's data sheet and then give their "review", without ever having done any actual hands on work with what their "reviewing".  Though this certainly isn't limited to PSUs.  Right now I'm looking for an AIO for a small form factor case and I'm running into the same issue: most of these "Best 120mm AIOs" have done no actual testing, they just looked at a feature list and decided X is better than Y.  And this makes up the vast majority of PSU reviews I encounter.  "We like this PSU because price to performance, modularity" blah blah but no actual testing.  So I knew to look for actual tests using external equipment, thanks in large part to Jonnyguru (I didn't understand the equipment he was using but I could tell this it was something your average joe couldn't do at home).  I never dreamed somebody would go through the effort and the cost of actually getting the equipment to test PSUs just to completely fudge up the tests.  And I don't have the knowledge to know if the tests are legit or not.