TechyInAZ :
turkey3_scratch :
If it's an overclocked GPU then I say 450W PSU. If it is reference then 200W CPU 150W GPU. Should be fine.
If a 450W PSU will work, can you show all of us proof from the web?
I really didn't want this thread to get derailed, I was just trying to show a point to the OP that he needs to specify his model number. I did this by telling him a lower-wattage PSU could easily handle his hardware when his might not. I'm just trying to make him realize that is all, somehow it always turns into this type of discussion then. But anyway, if we insist.
Well here is the RX 480
http://www.tomshardware.com/reviews/amd-radeon-rx-480-polaris-10,4616-9.html okay 164W in Metro Last Light 4K, I was 15W off.
So if he overclocks his FX 8350 to 4.8Ghz, and the FX 9590 is 5Ghz and a 220W CPU, then it can be implied the FX 8350 average power should be below 220W. In addition, that is under heavy load. Games that use two or four cores should not push the CPU very far past 125W even with the overclock, since only a few cores will be doing the heavy lifting.
This Techpowerup review measures the power on the EPS12V cable. So this included the CPU's power and the motherboard VRM loss power. Under load, the FX 8350 + motherboard VRMs was only 95W.
https://www.techpowerup.com/reviews/AMD/FX-8350_Piledriver_Review/4.html This is more representative of gaming load compared to worst case scenario. Intel's latest CPUs are typically 50W for an I5 when gaming or 70W for an I7. Since gaming doesn't usually load hardware to the max.
Anyway, that same TPU article measures full system under load with the FX 8350. Add on the 165W and now you are at 305W. With the overclock of the CPU, expect power to then increase to probably 350W when gaming. Expect the highest average to be around 400W (i.e. Furmark + some insane CPU load). I would not expect power to go above 400W. But now you have to think about transient spikes. Which is hard since few sites use an oscilliscope.
Toms did use one for the RX 480. Metro Last Light 4K. Most spikes are around 235W for the RX 480 but a few hit 300W occasionally. So then take into account FX 8350 power spikes, if they so just happen to coincide with the spikes of the RX 480 under some ridiculously heavy load, the FX8350 spikes of near 300W possibly could reach 300W themselves.
But if it happens the power supply probably won't turn off necessarily, it's an argument underway on Jonnyguru right now, so this type of discussion is still in progress. I learned some months ago that the protection IC would be able to detect instantaneous power compared to the secant/average over a time interval. This is why OCP is set so much higher, it should be able to take in these spikes well without turning off the PSU. It does all depend on the exact PSU though and the exact hardware and often we don't have enough information on either to make these judgements.
Quest for Silence says he ran GTX 460 SLI on an Enermax ErPRO 80+ 350W power supply
http://www.jonnyguru.com/forums/showthread.php?t=13766&page=3 for a good period of time and the PSU is still running. The GTX 460 is a 160W GPU for the base model. So as these 5 paragraphs show it's not such a simple matter it involves understanding the following things:
1) Average gaming power consumption over a brief time interval (i.e. 10s) in an average game
2) Average gaming power consumption over a brief time interval (i.e. 10s) in an intense game *cough Assassin's Creed
3) Ultra mega load power consumption over a brief time interval (i.e. 10s) such as in Furmark
4) Transient spikes of the GPU and CPU when gaming (instantaneous spikes, milliseconds) in an average game
5) Transient spikes of the GPU and CPU (instantaneous spikes, milliseconds) in ultra mega load (highest single power draw at an instantaneous moment)
6) Desirable trigger point of OCP/OPP of the power supply.
7) Error tolerance of this trigger point.
The shunts that a protection IC uses to measure the power on the rails are not always perfect. They have error tolerance. You can have a 450W power supply where the 12V rail overcurrent protection kicks in at 600W for one person's unit and then 560W for another person's. Unfortunately, unless digital protection circuitry is used like in the AX1500i, analog circuitry will always have these inconsistencies.
But we are missing so much information from my above list. 1 and 2 are pretty well known along with 3. But 4 and 5 are often not known too well. This is why you have to be very careful when reading any power conumpstion data from any website. Power from the wall, highest transient spike, average power over a brief time interval, and highest average power interval from a series of briefer intervals (i.e. 1s) is all very tricky.
All this can be portrayed in this oscilliscope shot of the GTX 1080. Metro Last Light 4K. Note this is an overclocked one since it is aftermarket. You can see that small amount of purple where a brief spike almost hits 400W. This is a GTX 1080 here, Nvidia's efficient GPU. But people have run a GTX 1080 on 450W power supplies and been perfectly fine. Those spikes seem to be in millisecond territory, about 0.1 seconds for each. So they are extremely brief. Capacitors and any energy-storing things in a PSU should be able to adapt to these, which is where transient response testing comes from in HardOCP and Toms/TPU power supply reviews.