Sign in with
Sign up | Sign in
Your question
Solved

Minimum PSU for my new Build

Last response: in Systems
Share
October 10, 2012 9:52:16 AM

I was hoping someone could help me with my HTPC/Gaming rig build. I have on board GPU cross fired with my discrete card but after a few hours of gaming the system restarts. I am sure the two possible reasons for this :

- Overheating
- PSU Overload or Issue

Can someone help me to pinpoint the issue and how to conclude what actually is the reason of frequent crashes. My build is:

CPU: AMD A8 3870K 3GHz (OCed to 3.6GHz)
GPU: Sapphire Radeon HD 6670 DDR5 1GB (OCed to 1100MHz)
PSU: 430Watts Seasonic 80+ Bronze
RAM: GSkill 2x4GB 1600MHz (running at 1866MHz dual mode)
MB: MSI A75MA-G55
CPU Cooler: CM 212 EVO

I prefer to OC my CPU/GPU through AMD Overdrive as the BIOS tool OC Genie can't even OC my system to 3.1GHz and it fails to boot.

I appreciate any help I can get.

More about : minimum psu build

October 10, 2012 12:57:22 PM

Its not a PSU being overloaded, those components only sip power.
Though that doesnt rule out it may just be broke.

Monitor your system temps while running Prime95 and Furmark, that will tell you your max temperature. If either get above 80-90c, there is an issue.

Back off your overclocks, maybe they are unstable and fail after extended usage.
m
0
l
October 10, 2012 1:32:04 PM

^ What manofchalk said. Heat could be a part of the problem, I've never really heard someone pushing a HTPC hard like this. Check temps and voltages here, you shouldn't have a power supply problem. The phenomenon that normally happens is that voltage/power requirements for components scale upward with heat. So what may be happening is that the settings you have are good for a cool and underutilized computer, but as heat builds inside the case and your components become hotter, they may be reaching a point where supplied voltage becomes insufficient - hence the delayed restart. I too would back off of your overclocks a bit, but definitely re-check you temperatures here. If you don't wish to scale back your overclocks a bit, make sure your voltages are sufficient and that the timings/settings for your 1600 RAM running at 1866 are solid.

*I've run higher power demands than this on a 400 watt PSU and had no issues. I would be surprise if you're even using close to 350 watts.
m
0
l
Related resources
October 10, 2012 2:22:22 PM

matt_b said:
^ What manofchalk said. Heat could be a part of the problem, I've never really heard someone pushing a HTPC hard like this. Check temps and voltages here, you shouldn't have a power supply problem. The phenomenon that normally happens is that voltage/power requirements for components scale upward with heat. So what may be happening is that the settings you have are good for a cool and underutilized computer, but as heat builds inside the case and your components become hotter, they may be reaching a point where supplied voltage becomes insufficient - hence the delayed restart. I too would back off of your overclocks a bit, but definitely re-check you temperatures here. If you don't wish to scale back your overclocks a bit, make sure your voltages are sufficient and that the timings/settings for your 1600 RAM running at 1866 are solid.

*I've run higher power demands than this on a 400 watt PSU and had no issues. I would be surprise if you're even using close to 350 watts.



Thanks Matt. So can I rule the PSU issues out of question here and concentrate and monitor heat issues only? Can you advice on voltages I should be hitting on each components and the kind of OC I should look to be considered as a SAFE LEVEL? AMD APUs, specially the 'K' series ones, I thought could be OC considerably higher. But yes considering the cramped place my rig sits and only 3 fans, including the CPU cooler I am sure heat is the culprit here. Any specific tools I should consider to note temps during idle, mid and heavy load and how should I plan to get a cooler system overall. FYI its a brand new rig, only weeks old.
m
0
l
October 10, 2012 2:29:09 PM

manofchalk said:
Its not a PSU being overloaded, those components only sip power.
Though that doesnt rule out it may just be broke.

Monitor your system temps while running Prime95 and Furmark, that will tell you your max temperature. If either get above 80-90c, there is an issue.

Back off your overclocks, maybe they are unstable and fail after extended usage.


Thanks manofchalk. Can you please tell me how do I monitor temps? Do I run Prime95 & Furmark or may be run a game in the background? Any tools you recommend to monitor the whole OC process and reach a stable level for the whole system. I am considering to use the system for gaming for sometime till get my full on gaming rig. So I'd like to be on the higher performance side. Thanks again.
m
0
l
October 10, 2012 3:45:00 PM

Using the smallest possible PSU is like running your car in first gear all the time. Think about it.
m
0
l
October 10, 2012 4:08:11 PM

ram1009 said:
Using the smallest possible PSU is like running your car in first gear all the time. Think about it.


I expected this reply sooner or later :) 
Actually I had this PSU with me and wanted to bring this to use. But seriously, isn' t 430 watts enough for this build of mine. What do you feel is a non-overkill PSU for me? Any demerits of using a less powered PSU? How is it going to harm the system anyway.

Thanks.
m
0
l
October 10, 2012 4:24:40 PM

AbhieSpeaks said:
I expected this reply sooner or later :) 
Actually I had this PSU with me and wanted to bring this to use. But seriously, isn' t 430 watts enough for this build of mine. What do you feel is a non-overkill PSU for me? Any demerits of using a less powered PSU? How is it going to harm the system anyway.

Thanks.


Seasonic is perhaps the TOP quality psu maker.
430w is plenty, and A Seasonic psu will run at peak wattage even higher than advertised.

Start by backing off all of your overclocks.
Not all cpu or graphics cards can accept high overclocks.
Run a stress test using something like prime95 and furmark.
Monitor your cpu temperatures with realtemp, and the graphics card with GPU-Z

Verify that you have installed the latest chipset and graphics drivers along with any patches to your game.

m
0
l

Best solution

October 10, 2012 6:46:15 PM

ram1009 said:
Using the smallest possible PSU is like running your car in first gear all the time. Think about it.

I see too many people with a setup like this that may be using 350 watts and they get a 650+ watt unit - this is stupid. The rule of thumb is pick a unit that gives around 25%/30% margin of safety. A PSU should run at it's peak efficiency in this range. Running a PSU that is absolute overkill can be just as hard on internal components as over-stressing one - there's a member on here that could word this far better than I could but this is the same summary.

So here's an example, there's no way to tell what the system uses without measuring, so using TDP:
HD6670 = 66 watts
AMD 3870k = 100 watts

So 166 watts is going to be the most wattage used from the two most power-hog components by a long shot. For absolute overkill, let's say the rest of your stuff uses 100 watts (very unlikely) - We'll say your system will probably be about 266 watts at the absolute most (stock settings for example). So a 430 watt PSU at a typical 80% efficiency rating would be pulling about 537.5 watts from the wall to convert at 100% load. Give a safe 30% margin of safety, and the PSU is supplying you 301 watts or 376 watts from the wall. Given the already bloated power consumption number of 266 watts in the system example, you really aren't even close to maxing this thing out as this comes in at a whopping 62% load.

In the end, is it better to go bigger than smaller on a PSU - yes. Choose one, like you have, that matches what it's intended purpose is though.

With the total of 3 fans in your system though, how many more and what size(s) could be added?
Share
October 10, 2012 7:15:57 PM

manofchalk said:
Back off your overclocks, maybe they are unstable and fail after extended usage


This. On the other hand, it's up to you if it matters that much. I have my components pushed as far as they'll go without dangerous overvolting... they're cooled as well as can be, and run stable just fine... but have to reboot every 6 hours or so. Personally, I don't mind this - I just set up a little timer on my computer that tells me at 5 hours and 30 minutes, because a 19 second reboot ever 6 hours doesn't bother me, and it gets me a significant power boost without harming anything.
m
0
l
October 11, 2012 5:11:53 AM

Prime 95, Furmark, Core Temp, MSI Afterburner and games are what I use to test my various overclocks.

Prime 95 will stress the CPU to 100%, serving as both your extreme temp scenario and stability test. Use a program like Core Temp to monitor the temperature while this is happening.

Furmark will stress the GPU to 100%, so you have the extreme temps. But in my experience, isn't good for telling a stable overclock. I can run an unstable overclock with no visual artifacts or oddities with Furmark. As soon as I enter a game the display driver crashes, and im shunted back to the desktop. So games or something like 3D Mark 11 would be your stability test.
MSI Afterburner is to overclock my GPU and monitor temperature.

I noticed you have clocked your memory as well. I havent ventured into that territory yet, so dont know you could use there.
I imagine a Memtest run, a Prime95 blend test (which stresses RAM as well) and a RAM intensive activities like video editing and rendering would be a decent enough test to see if its stable.
Unless you have really cranked up the voltages in the memory, I wouldn't worry about temperature.
m
0
l
October 11, 2012 1:31:16 PM

To add to what Manofchalk said, I would also add GPU-Z as another graphics card utility for monitoring. For realistic video card testing though, the Heaven Benchmark is an excellent simulated load as well. Furmark is regarded in the same light as IBT for your processor in the sense that they are both good synthetic benchmarks and you're going to get maximum heat from them, but they can be unrealistic as they are just that - synthetic.

Quote:
Unless you have really cranked up the voltages in the memory, I wouldn't worry about temperature.

Rarely is temperature for RAM a problem, it has to do with stability. Problems/crashes develop If timings are too tight, speed is too high, voltage is not enough, memory controller voltage isn't compensated for higher bandwidth, etc. Overclocking RAM can be a very finicky process.

Other processor stress tests that I personally believe are stressful from most to least would be:
IBT, Linpack, OCCT, Prime95. Whatever you test with, the general rule of thumb is try and stay at or below 75* C (some chips accommodate higher and some lower) for the more stressful tests - definitely better to be on the below side of this number though. Another rule of thumb is to try not to make CPU voltage more than 0.2 volts from stock. So if you're running something like 1.2 volts at stock (non overclocked), try to stay at or below 1.4 volts - this is the extreme side. RAM if rated at 1.5 volts, really shouldn't deviate much from this. With memory controllers on the processor on most architectures now, anything past the 1.5 volt range anyway starts to stress it a little too much, even the sticks of RAM rated at 1.65 volts could bring issues. These again are general figures I'm telling you and values more on the extreme side. The actual numbers vary from part to part, so some could be more forgiving and others not so much. Most manufactures won't post actual numbers for this reason, so keep this in mind. Your best bet is stay well below.

The last thing to look at is voltage specs. Make sure your PSU isn't dropping too low on your voltages. All (3.3, 5, and 12 volts) should be close to or higher than rated. It's not a good sign if you're stress testing and your 12 volt rail is telling you something like 11.5 volts.

You are right about overclocking well for these chips. What that means though is you HAVE to have a good cooler and good airflow. In the situation with the HTPC case here, you're producing a lot of heat, you need to get it out of the case as well though. If you cannot exhaust any more heat than currently (observed while stress testing), you're faced with no other option than to back your clocks down a bit.
m
0
l
October 18, 2012 6:04:52 AM

Ok. Here is what seems to be the issue. I backed off the RAM OC and things are working fine. The GSkill Sniper 1600 2x4GB RAM doesn't seem to like the OC. I was able to play FIFA13 for 3-4 hours w/o any issues. No question is how do I OC my RAM? How much difference will 1600 -> 1866 is gonna make? Any guidance here will be appreciated guys. Thanks for the help already.
m
0
l
October 18, 2012 8:57:37 PM

AbhieSpeaks said:
Ok. Here is what seems to be the issue. I backed off the RAM OC and things are working fine. The GSkill Sniper 1600 2x4GB RAM doesn't seem to like the OC. I was able to play FIFA13 for 3-4 hours w/o any issues. No question is how do I OC my RAM? How much difference will 1600 -> 1866 is gonna make? Any guidance here will be appreciated guys. Thanks for the help already.

There is a difference but noticeable is debatable. Like I said though, overclocking RAM is very complicated if you're looking to push it as far as you can. It's best to read up on how RAM operates. The two biggest bottlenecks in computers these days is the storage drive and memory, so fooling with RAM helps in the regard.

As for differences (ns = nano seconds and lower is better):
9 CAS 1600 MHz RAM = 11.25 ns
9 CAS 1866 MHz RAM = 9.65 ns

More than likely though, you will have to increase the CAS timing on your RAM when going for higher frequency:
10 CAS 1866 RAM = 10.65 ns
11 CAS 1866 RAM = 11.79 ns
You can see though, that DDR3 does love higher frequency more than tighter timings, but If you must reach 2 CAS levels higher than your default RAM timing for stability if the other timings and voltage do not help, then you're becoming counter-productive.
m
0
l
October 22, 2012 9:33:14 AM

Best answer selected by AbhieSpeaks.
m
0
l
!