"Upgrading" Corsair power supplies

Morbert

I recently bought a HX750 PSU for my PC after the old PSU died. Unfortunately I don't think the PSU is powerful enough (my last one was 1000W). Does anyone know if I can send it back and pay the difference for a higher wattage PSU? I've contacted their customer support but they haven't gotten back to me yet.

Solitaire

What's your system spec?

TouchingVirus

The system is an Alienware machine and from what I remember the specs are this following:
Intel QX9450
DDR2 Ram (8gb?)
ATI Crossfire X2
2 SATA hard drives in a RAID array.
Fancy case with plenty of fans & lights. Oh and it's liquid cooled.

The old PSU was a piece of sh*t generic 1000W from Dell, provided as a replacement for the original PSU which also went bust. According to any PSU calc the draw, including all the fancy lights, is only 720W at 100% load, so I bought Morbert a 750W when I was fixing the machine.

Morbert - take this one over to the Komplett forum - http://boards.ie/vbulletin/forumdisplay.php?f=242

I remember asking Ryan in there if I could return the PSU if it didn't fit your case and he said I could, but nothing was mentioned about returning it if it wasn't powerful enough

marco_polo

Would need to know the exact ATI card models?

Unregistered.

TouchingVirus wrote: »

The system is an Alienware machine and from what I remember the specs are this following:
Intel QX9450
DDR2 Ram (8gb?)
ATI Crossfire X2
2 SATA hard drives in a RAID array.
Fancy case with plenty of fans & lights. Oh and it's liquid cooled.

The old PSU was a piece of sh*t generic 1000W from Dell, provided as a replacement for the original PSU which also went bust. According to any PSU calc the draw, including all the fancy lights, is only 720W at 100% load, so I bought Morbert a 750W when I was fixing the machine.

Morbert - take this one over to the Komplett forum - http://boards.ie/vbulletin/forumdisplay.php?f=242

I remember asking Ryan in there if I could return the PSU if it didn't fit your case and he said I could, but nothing was mentioned about returning it if it wasn't powerful enough

If it has enough PCIe power sockets (not using adapters) for the graphics cards then chances are that it is powerful enough. However - is anything overclocked?

Solitaire

TouchingVirus wrote: »

Intel QX9450

Do you mean Q9450? Or QX9650??

ATI Crossfire X2

What cards? The only setup that would make the HX750 even moderately unhappy would be 2x HD4870X2s...

Fancy case with plenty of fans & lights. Oh and it's liquid cooled.

What WC setup?? What lights - LEDs, CCFLs??

The old PSU was a piece of sh*t generic 1000W from Dell, provided as a replacement for the original PSU which also went bust. According to any PSU calc the draw, including all the fancy lights, is only 720W at 100% load, so I bought Morbert a 750W when I was fixing the machine.

In my experience Dell's "piece of sh*t generic" PSUs are mid-grade Delta or HiPro units that are pretty goddamn formidable - I'd be surprised if one went, if two went I'd start to wonder if it was getting nuked by rippling reflected loads off bad caps on the cheap-ass proprietary Dell motherboards!

Morbert - take this one over to the Komplett forum - http://boards.ie/vbulletin/forumdisplay.php?f=242

We haven't even worked out if the HX750 isn't sufficient yet!

I remember asking Ryan in there if I could return the PSU if it didn't fit your case and he said I could, but nothing was mentioned about returning it if it wasn't powerful enough

Depends on how long you've had it. If it works but you just don't like it there may be a 15- or 30-day limit on returns, and you only get 80-85% of the value back due to restocking fees (those are standard btw). You might not even need to return it though

Morbert

If it has enough PCIe power sockets (not using adapters) for the graphics cards then chances are that it is powerful enough. However - is anything overclocked?

Nothing is over clocked. And from what I've also read and calculated, 750W should be more than enough for the components.

As TV said, I had to return the system before because the PSU died. When I got the PC back, it worked for about a a week before dying again, and I was out of warranty. So maybe it's the PC itself that's the issue.

Either way, if I can't return it, I am building my own PC around it. TV: I was sickened when I heard Eug's beast was only around a grand.

Morbert

Solitaire wrote: »

Do you mean Q9450? Or QX9650??

It's the Q9450.

What cards? The only setup that would make the HX750 even moderately unhappy would be 2x HD4870X2s...

I did some calculations with 2x X2s and yeah it was still well below 750.

What WC setup?? What lights - LEDs, CCFLs??

This I am unfortunately not able to tell you at the moment. All I can say is I chose the "minimum" number of lights when ordering the PC. But will have a look.

In my experience Dell's "piece of sh*t generic" PSUs are mid-grade Delta or HiPro units that are pretty goddamn formidable - I'd be surprised if one went, if two went I'd start to wonder if it was getting nuked by rippling reflected loads off bad caps on the cheap-ass proprietary Dell motherboards!

I don't know if two different PSUs went or if one went and they just fixed the same one and put it back in. It was a strange problem with my PSU insofar as, if I could manage to turn on the PC, the PSU would work and the PC would run without complaint (To make sure, I even left it on for a month at one stafe). What would happen is the computer would only work once every 10 "tries". Then once every 50, then once every 500, and eventually never at all.

With this PSU, the computer turns on fine all the time, but cuts out if I do anything more intensive then a flash game.

We haven't even worked out if the HX750 isn't sufficient yet!

Oh yeah, certainly not rushing out yet. Just want to see if the option is there. All I'm using the PC for now is reading documents, and the PC is fine for that.

[edit]-Another useful bit of info is that the "CPU hot" LED will flicker on even when doing things no more demanding than browsing the internet. And I've just noticed that it will stay on constantly if I play anything remotely more demanding that youtube (i.e. flash games will cause it to stay on).

TouchingVirus

Again lads, this is all from my memory. I'm exactly 200 miles away from the computer I'll get you the specifics later tonight if I can get Morbert to run Aida/SiSoft Sandra/Everest

The case is an Alienware Area 51. I have no idea about the watercooling setup, I just know it takes a 12v 4-pin to power it. The lights are LEDs, there are no CCFLs. Total of about 15 LEDs, they're powered via PCBs that take 4-pin molex connectors. Grand total of 4 fans, all are 200mm or so.

The old PSU was a Newton N1000P-00 1000W power supply. From my own research it's not quite formidable as problematic. I have no idea what the first PSU was, but I'm pretty sure it was water ingress that caused it's failure and not a general PSU fault.

The GFX cards may be 4870x2. I'll verify later if Morbert is online and I talk him through getting the true specs.

By their very nature I think Alienware machines are overclocked. It might be something I have overlooked when I bought the new PSU.

TouchingVirus

I wonder if it's the CPU thermal cutout that's triggering the power cut then, and not the PSU. Your CPU shouldn't be running hot

Morbert

TouchingVirus wrote: »

I'll get you the specifics later tonight if I can get Morbert to run Aida/SiSoft Sandra/Everest

Will do.

Morbert

TouchingVirus wrote: »

I wonder if it's the CPU thermal cutout that's triggering the power cut then, and not the PSU. Your CPU shouldn't be running hot

One of the coolant pipes is very hot to touch. GPU temp havers around 60 degrees C when browsing the internet but don't know about the CPU temp.

Solitaire

I still don't understand why Alienware charge more for a lower spec than Dell :eek: I keep forgetting we're not talking about a real Dell here

In any case that is a Delta PSU - recognised the model number format and JonnyGuru backs it up. JG suggests they're good high-end units from internals and performance. However, when they get the "Newton/Alienware" branding those PSUs have a bunch of connectors sliced off and replaced with proprietary Dell connectors :eek: Bad Dell!! Didn't you learn your lesson from the last time you did that, let alone the BTX fiasco?!?!? :mad::mad::mad: In any case something about the proprietary bits - connectors or mobo - may be causing an issue with the Deltas because when they're sold off cheap second-hand and modded to work in a conventional PC they work fine

And if the CPU is getting hot something is very wrong. Check that the WC loop isn't leaking and that the CPU waterblock is still properly attached - it may have started to work loose and/or need a fresh application of thermal paste. Is that WC a loop of discrete components or an Asetek-type "all-in-one" pre-built module like the Siberian Tiger/Corsair H50 etc.?

TouchingVirus

Something went wrong in the PSU innards, it was buzzing and clicking by the time I got my hands on it. I know from the stickers and codes on it that it was definitely a Delta and they're a decent brand but I remain skeptical about the Newton branding after the one experience there and that of a few others with Alienware PCs around the internets.

It looked like an all-in one watercooling solution. There was the fan that was mounted to the case, the radiator to the fan with two pipes coming off and going onto the waterblock on the CPU.

Morbert, call me tonight and we'll get the full specs of your machine. If you want I'm down your neck of the woods this weekend so if you're around I can take the machine back to mine and troubleshoot it with the folks here on boards. I'm also closer to Komplett

TouchingVirus

Right then, here are the official specs:

CPU: Intel(R) Core(TM)2 Quad CPU Q9650 @ 3.00GHz
RAM: 4GB DDR3-10666MHz (2x2GB) - Elpida EBJ21UE8BAFA-AE-E (7-7-7-20 4-27-8-4)
Hard Drives: 2xSATA Hard drives, 7200RPM (ST3500620AS)
Optical Drive: HL-DT-STDVD-RAM GH22NP20
GFX: ATI Radeon HD 4870 X2 (2x 800 SM4.1 750MHz, 2x 1GB DDR5 4x900MHz, PCIe 2.00 x16)



Two interesting bits from the Report:

< Temperature Sensor(s) >
Board Temperature: 113.00°C
CPU 1 Temperature: 86.00°C td
CPU 2/Aux Temperature: 28.50°C

< Voltage Sensor(s) >
CPU 1 DC Line: 1.07V
CPU 2/Aux DC Line: 3.00V
+3.3V DC Line: 3.29V
+5V DC Line: 7.31V
+12V DC Line: 9.48V
Standby DC Line: 3.29V
Battery DC Line: 3.29V

Warning 2523: Memory bus speed exceeds memory rated speed.
Reduce memory bus speed.
Warning 2518: Mainboard temperature is too high.

That CPU temp is way too high IMO, especially considering at most all that was happening was a flash video playback. I think the board temperature can be ignored, there's no way the board is 113 degrees as Morbert says he can touch it fine.

More concerning is the reading on the +5v rail, a staggering +7.31v - what's using up that much power. There are no peripherals plugged in, no USB, no firewire, nothing. The fans/LEDs are connected to the +12v rail

I'm going to guess from the Memory bus speed warning there that the machine is overclocked.

Solitaire

What the hell is giving those readings?!?!? :eek: Is that HWMonitor drowning in failsauce again? Because a PC can't power up with those 5v and 12v rail voltages, not in a million years And that's ignoring the fact that according to the report the mobo is a small puddle of melted plastic on the floor :P

TouchingVirus

SiSoft Sandra provided the report, I think it's drowning all right

Morbert is going to test with Prime95 now, apparently the idle temp is 50-60C and doing anything spikes it to 80C+, flash videos on youtube put it up to 90C+. There's no way a cooling system that works puts out those temps on a quadcore

Morbert

Solitaire wrote: »

What the hell is giving those readings?!?!? :eek: Is that HWMonitor drowning in failsauce again? Because a PC can't power up with those 5v and 12v rail voltages, not in a million years

Hmm... Would that explain my PC not powering up 99.99% of the time with the older PSU, but being fine if it actually managed to?

And that's ignoring the fact that according to the report the mobo is a small puddle of melted plastic on the floor :P

It was all grabbed with SiSoftware. The CPU temp readings are similar to those reported by "Core Temp". They are 78/79 C when idle, but slowly increase in temp to past 100 when even a small task is undertaken. I have a feeling this is what's causing the PC to shut down rather than the PSU. I have not mustered up the courage to deliberately induce a shutdown yet to see if altering the cutoff in the bios confirms the theory. (Plus I haven't figured out how to access the bios yet. The bootup flashes by too fast, so I'm systematically trying all the usual suspects)

Morbert

TouchingVirus wrote: »

Morbert is going to test with Prime95 now, apparently the idle temp is 50-60C and doing anything spikes it to 80C+, flash videos on youtube put it up to 90C+.

I should have clarified. The GPU temp 50-60 when idle. The CPU temp still goes mad.

TouchingVirus

F2+DELETE at the same time. Hammer it from bootup

Unregistered.

TouchingVirus wrote: »

F2+DELETE at the same time. Hammer it from bootup

Is there any water in that water-cooling rig?

Morbert

Unregistered. wrote: »

Is there any water in that water-cooling rig?

The pipes are hot to touch, and the radiator fan in the back is blowing out plenty of hot air.

Solitaire

Before I was so rudely interrupted by both Boards and my PC crashing... :rolleyes:

Morbert wrote: »

Hmm... Would that explain my PC not powering up 99.99% of the time with the older PSU, but being fine if it actually managed to?

Nope - at those voltages the PC simply wouldn't function. At all. Ever.

Can you check what VID CoreTemp gives you while the CPU is loaded? That would be the stock voltage for that particular chip. Now check what real-time Vcpu CPU-Z gives you. I have a funny feeling that, even at stock clocks, your CPU is being massively overvolted :eek:

Morbert

Solitaire wrote: »

Before I was so rudely interrupted by both Boards and my PC crashing... :rolleyes:



Nope - at those voltages the PC simply wouldn't function. At all. Ever.

Can you check what VID CoreTemp gives you while the CPU is loaded? That would be the stock voltage for that particular chip. Now check what real-time Vcpu CPU-Z gives you. I have a funny feeling that, even at stock clocks, your CPU is being massively overvolted :eek:

Ok, I will have a look tonight at doing this. As well as confirming it is the thermal cutoff that's causing the shutdown.

If my CPU is being overvolted, is that a difficult problem to fix? That would sound like a PSU fault, but considering I've gone through three of them...

Solitaire

Just dig into the BIOS settings and set the CPU voltage back to Auto or even try the exact VID from CoreTemp if you don't want to OC it (yet). Then see how you get on with the temps

Morbert

Solitaire wrote: »

Just dig into the BIOS settings and set the CPU voltage back to Auto or even try the exact VID from CoreTemp if you don't want to OC it (yet). Then see how you get on with the temps

Hmm... All the voltages are auto at the moment already. CoreTemp tells me the VID for all cores is 1.2125 V at 100% load. Temperature jumps to 110 when fully loaded.

Morbert

Here's a pic of what CPU-z is telling me

Morbert

Eh... woopsy... I change the Core voltage to 1.212 in the BIOS and it ramped up the temperature, so much so that the PC would cut out 5 mins after loading windows (The one time I managed to get CoreTemp up and running, it reported temps of +110!). I've set it back to auto and it's ok (still running hot but stable).

I don't really understand the colour coding. High voltages were red, mid-range voltages were green. Low voltages were gray. 1.212 was a grey voltage so I assumed the temp would lower.

[edit]-Oh the only other thing I did was change the multiplier from 9x to 6x. I don't know why, but the BIOS told me it was 9x and CPU-Z told me it was 6x

Anti

The 5v and 12V rails will not run at what is being read. They only have a 10% tolerance. so 5.5v for example should be the most it would run before cutting out.

Here is a screengrab of my my temps and voltages across the board. Notice how precise they are. And this is my cpu being maxed out on 4 cores at the moment with the F@H mp client(spec in sig).



At the moment im thinking there is certainly not enough water in the loop, a kink, or the block has lost full contact. Or that thermal paste has all gone to ****. As for the weird voltages and lack of powering on, im guessing something is up with the pwm's on the mobo. Can you have a look around the cpu socket and check for large clumps of dust as this could be a bad contact/short.

Anti

Morbert wrote: »

[edit]-Oh the only other thing I did was change the multiplier from 9x to 6x. I don't know why, but the BIOS told me it was 9x and CPU-Z told me it was 6x

This is called intel speedstep. When the cpu is not under load the voltage and multiplier will drop to keep the heat low... in theory. Best just to turn it off tbh.

Unregistered.

Anti wrote: »

At the moment im thinking there is certainly not enough water in the loop, a kink, or the block has lost full contact. Or that thermal paste has all gone to ****...

This but if it were the paste or the cooler not making contact then the pipes wouldn't be hot. If it's a kink or the pump has failed then the water could be gently simmering in there. I'd suggest getting your hands on the stock cooler for LGA775 quads and try that.