New silentt PC arrived! But it won't boot properly

[mcv]

The new PC I ordered just arrived:

Antec Solo case
GA-EP35-DS3
Intel E8400
Ninja Scythe rev B (passive)
Peak HD3850
Accelero S1
Samsung Spinpoint T166 501LJ (2x)
Seasonic S12II 380
Pioneer 215D optical drive
4x1GB Corsair Twin2x2048-6400C4DHX
Thermalright bolt-thru kit
Arctic Silver 5

Instead of assembling it myself, I had the professionals at the webstore I bought it from assemble it for me. I wanted an extremely quiet machine, so I'd selected lots of custom passive heatsinks, and having no experience installing those myself, I figured they'd do it better than I could. Also, they give 1 year warranty.

I got a strange call from them, however: they said my Ninja Scythe really did need a fan, because it' get too hot otherwise. Nonsense ofcourse; lots of people use the Ninja without a fan, and the Antec Solo should have excellent airflow to allow passive cooling. But hey, I can easily remove the fan later, so I let them assemble it with the fan. I haven't removed the fan yet.

When the PC arrived, I was a bit apalled by the mess of wires inside. Lots of PCs have a big mess of wires, so it shouldn't be fatal, but the Solo has a nice cable management system that lets you hide all the excess cables in a convenient location, instead of in the middle of the airflow path. This should be easy to fix, but I haven't done that yet, first wanting to test the machine in the way it was delivered.

The next surprise was that they hadn't used the Thermalright bolt-thru kit. Not fatal, as the Ninja works perfectly fine with its own push pins, but it's just a bit more secure with the extra backplate. I'd have to disassemble almost the entire PC to fix this, so I'm leaving it for now.

Now I'm wondering, if they didn't use the bolt-thru kit, did they even use the Arctic Silver? They said they would, but I have no idea how I can check.

Anyway, according to CMOS setup, system and CPU temps are a cozy 32 degrees C, fans are spinning at 1000 rpm (probably more than is really necessary), and the system isn't anywhere near as quiet as I thought it would be. I don't have any data on GPU temp or speed.

DRAM timings set to auto should detect SPD values, right? My RAM should have 4-4-4-12 timings, but was somehow set to 5-5-5-18. I originally changed this to 4-4-4-12, but I've changed it back to auto again because it wouldn't boot properly. In fact, I've set CMOS to failsafe settings (no idea what it is, but it sounds safe), although I did enable USB mouse and keyboard, because it's a bit hard to anything without those.

Problem is, my system stil doesn't boot properly!. It did boot from the install CD (64 bit Ubuntu 7.10), but once I finish installation and boot from harddisk, it just keeps booting. No error message that it can't boot, just restarts the boot cycle everytime it should start loading my OS.

So what could be the problem? Did the Ubuntu installation mess up? Sounds unlikely. It isn't known for its easy installation for nothing. Did the assembly crew mess up? They didn't sound quite as experty as I'd hoped, and didn't use the bolt-thru kit that I ordered. (No idea if they applied the arctic silver). Should I invoke the warranty? It'd probably take a week before I have my machine back, and I want to start using it now. Monday at the latest. Should I try to fix it myself? I don't even know what the problem is. Even without the arctic solver, CPU temp seems fine, and while the airflow is partially obstructed by a big knot of cables, system temp is fine. There are things that require attention to get the best performance (and lowest sound) out of the PC, but nothing that should prevent it from booting.

Any ideas what could be the problem here?

 

[mcv]

After some more checking, I noticed that Grub (linux boot loader) does start, and it even mentions the kernel, but after that, it reboots.

Why would a fresh linux install not work? No idea, but I booted from the CD again. This works fine (perhaps there's a problem with the harddisk?), and I noticed it contains a memtest utility. So now I'm running Memtest86 v1.70 and it's found 187 errors. It also found 187 errors on the second pass (not sure why multiple passes are necessary). I have absolutely no idea what these errors mean, but I fear faulty memory modules. Or perhaps I hope faulty memory modules, because they're easy to remove and replace. I guess I'll try removing some to see if that fixes the problem.

 

[QueBert]

errors mean bad memory, or possibly a bad MB, the best way to troubleshoot this, pull all out but 1 stick of memory and re-run mem test. If it passes with 1 stick, put a 2nd in and test with 2, and so on until you find the bad stick or bad memory slot.

I recommend you run it at least 3 hours per stick, overnight is best but with 4 sticks that could take some time. Start out 3 hours until you find the problem. you should get zero errors in Memtest, even 1 is a red flag.

 

[QueBert]

errors mean bad memory, or possibly a bad MB, the best way to troubleshoot this, pull all out but 1 stick of memory and re-run mem test. If it passes with 1 stick, put a 2nd in and test with 2, and so on until you find the bad stick or bad memory slot.

I recommend you run it at least 3 hours per stick, overnight is best but with 4 sticks that could take some time. Start out 3 hours until you find the problem. you should get zero errors in Memtest, even 1 is a red flag.

 

[mcv]

Why should I run it at least 3 hours? The first errors show up after a few seconds of testing. A single pass is less than an hour and it's immediately obvious that it's going to find lots of errors.

I spent all day yesterday testing the memory sticks, testing them all, testing only two, testing each individually, testing them in different positions, and always I got exactly the same errors. Except that I got more of them when testing less memory. I got about 186 errors each pass when testing all 4 memory sticks, but when testing only 1 or 2, I often got hundreds of errors.

Is the memory simply completely rotten, or could it be a different error? www.memtest86.com says that memory errors can also be caused by other things, although it doesn't say what or how I can find out.

 

[QueBert]

sorry I should have been more clear, if you get errors stop right away that's a no go. if you get no errors keep running for 3 hours.

if you get errors with all sticks in all slots, it's either A. bad MB or B. mem isn't compatible with MB. if you can get another brand of memory I'd try that. But it's one of the 2 things. i

 

[DSF]

I just posted this in your other thread, but I'll put it here so you'll see it either way. Find someone with a working stick of DDR2 and test it (and just it) in your motherboard with Memtest. If it passes, then the problem is most likely the RAM. If it doesn't, then the motherboard is probably to blame.

 

[mcv]

I'm still hoping for the incompatibility. A friend said that Corsair DHX memory is not something normal people should use, and using it without being an expert overclocker is asking for trouble. Whether that means it really is an incompatibility, I don't know, but that's what I hope. I'm certainly no expert overclocker.