New PC

[Steve Kalle]

Your experience seems very odd. Have you checked your BIOS and Windows settings to prevent them from spinning down? Have you replaced your sata cables, tried different power cables, tried different sata ports on the mobo?

I have almost 20 1TB Seagate 7200.12 drives and 4 2TB 5900rpm LP drives between 2 PCs and none of them spin down except for one 2TB in an external case. I also have 4 2TB Hitachi drives that never spin down. And I have a 2-day old 1TB 7200.12 that does not spin down (I am testing it before adding it to the array).

But then again, you are using the Intel onboard Raid, and that could be the culprit. Most of my drives are attached to hardware Raid controllers and only a few are attached to the Intel ICH10R for cache and backups.

[Randall Leong]

Originally Posted by Steve Kalle

Your experience seems very odd. Have you checked your BIOS and Windows settings to prevent them from spinning down? Have you replaced your sata cables, tried different power cables, tried different sata ports on the mobo?

I have almost 20 1TB Seagate 7200.12 drives and 4 2TB 5900rpm LP drives between 2 PCs and none of them spin down except for one 2TB in an external case. I also have 4 2TB Hitachi drives that never spin down. And I have a 2-day old 1TB 7200.12 that does not spin down (I am testing it before adding it to the array).

But then again, you are using the Intel onboard Raid, and that could be the culprit. Most of my drives are attached to hardware Raid controllers and only a few are attached to the Intel ICH10R for cache and backups.

The problem is that I have no hardware RAID controller card at all (and good ones cost at least $500 for an 8-port model). And I have already checked the settings. The problem here is quality control on the part of the manufacturers. It is very unusual that I get five faulty drives - but it does happen.

[Craig Coston]

Originally Posted by Randall Leong

I had difficulty in running six 2GB sticks of memory at the memory's full advertised speed. I had to drop the speed down to the next-lower speed class in order to run stably. As such, the memory is running at only DDR3-1240 speed instead of the advertised 1600 speed.

Running six 2GB sticks of memory at Intel's official maximum memory speed of DDR3-1066 is no problem. But running those same six sticks of memory at anything above DDR3-1333 speed is a crapshoot (meaning that it may or may not work). Any memory speed above 1066 is considered overclocked when it comes to the memory controller on lower-end (non-Extreme) LGA 1366 platforms.

As for using the other SATA III port, you could (but I personally wouldn't recommend it, especially since the Intel SATA II controller is of higher quality than any current third-party controller chip).

I have no problems OC'ing 6 sticks of Corsair Dominator DDR3-1600 on my Asus board. Are you sure that your experience wasn't limited to your particular hardware?

[Randall Leong]

Originally Posted by Craig Coston

I have no problems OC'ing 6 sticks of Corsair Dominator DDR3-1600 on my Asus board. Are you sure that your experience wasn't limited to your particular hardware?

Well, I have only the XMS, not the Dominator.

Second, I have a $200-ish Gigabyte motherboard. I might replace it with an Asus motherboard. I have trouble running 6 sticks at higher than DDR3-1400 speed on my current system.

[Craig Coston]

You're still using the Gigabyte? You may want to preface your comments that seem to generalize the performance of DDR3 (or any other facet of your particular motherboard) with the fact that is your experience with that particular board, not in general. It may mislead some. I personally have built 3 i7 machines with all RAM slots loaded, all on the Asus P6T or x58 boards, and all with the RAM overclocked. All worked fine, the two that weren't my machines ran Corsair XMS.

[Randall Leong]

Originally Posted by Craig Coston

You're still using the Gigabyte? You may want to preface your comments that seem to generalize the performance of DDR3 (or any other facet of your particular motherboard) with the fact that is your experience with that particular board, not in general. It may mislead some. I personally have built 3 i7 machines with all RAM slots loaded, all on the Asus P6T or x58 boards, and all with the RAM overclocked. All worked fine, the two that weren't my machines ran Corsair XMS.

I meant to say that there are too many variables involved when it comes to running with all six DIMM slots filled at any memory (RAM) speed faster than Intel's official maximum speed (DDR3-1066 in the case of the LGA 1366 i7 CPUs). The motherboard, the CPU, the RAM itself. In my particular case, with my particular setup (and yes, I'm still using the Gigabyte motherboard for now), I could not get my CPU overclock above 3.7GHz without the CPU overheating, even with a good CPU HSF (cooler). I tried setting the CPU core voltage to a lower setting, only to have the system crash at even stock speed. (The default CPU core voltage settings as detected by the Gigabyte motherboard are 1.28125V for the i7-920 and 1.25625V for the i7-950.) My previous Intel-branded X58 (DX58SO) motherboard performed better at lower CPU core voltages, but it has only four DIMM slots.

[Kyle Trambaugh]

Originally Posted by Noa Put

Just tested with crystaldiskmark and checked the sequential read and write speeds, the western digital 1tb black drives reach a read and write speed of about 135mb, when I activated the marvel raid controller and link 2 equal WD 1tb black drives to a raid 0 I get 260mb read and 200mb write speed, not bad for a motherboard controller I think :)

Noa,

Can you give me the step by step on this as I have never set up a raid 0... I have the same mobo and HD's. Specifically what cable do i need etc.

Right now I have the system dive in the SATA III #1 slot, the video storage in the SATA III #2 slot, and my export drive on a SATA II port....going to order another WD Black 1TB soon, for the RAID 0 video "drive"

Thanks in advance...

[Randall Leong]

Originally Posted by Kyle Trambaugh

Noa,

Can you give me the step by step on this as I have never set up a raid 0... I have the same mobo and HD's. Specifically what cable do i need etc.

Right now I have the system dive in the SATA III #1 slot, the video storage in the SATA III #2 slot, and my export drive on a SATA II port....going to order another WD Black 1TB soon, for the RAID 0 video "drive"

Thanks in advance...

Forget about the WD Blacks in a RAID array (as Steve Kalle stated). These drives (all versions manufactured since late 2009) have the spin-down issue which degrades the performance and reliability of even a RAID 0 configuration. I stated otherwise in one of my previous posts, which is misleading.

If you must get a consumer desktop drive, go with a non-WD hard drive. If you do choose a WD, you will have to get one of the much-more-expensive RE drives in order to even use RAID 0, let alone RAID 5/6/3. Be advised, however, that although the WD RE drives are optimized for RAID, they are relatively ill-suited for single-drive or JBOD setups because they actually perform slower than comparable non-RAID-optimized drives due to the extra overhead.

[Randall Leong]

Originally Posted by Randall Leong

Well, I have only the XMS, not the Dominator.

Second, I have a $200-ish Gigabyte motherboard. I might replace it with an Asus motherboard. I have trouble running 6 sticks at higher than DDR3-1400 speed on my current system.

An update:

I finally got my system (still with the Gigabyte motherboard) running stably (at least in Prime95) at 3.84GHz with the memory speed at DDR3-1600 (with all six DIMM slots filled). I had failed to re-adjust certain bus speeds and voltages in the BIOS, which is the reason why the system went into a reboot loop or locked up during the loading of Windows.

One potential problem: The CPU temps at this setting just exceeded 90°C (with the thermal cutoff point being 100°C) at 100% load.

[Craig Coston]

Randall, you're going to fry that system at those temps. What CPU cooler do you have? I can stress test my i7 920 @ 4.2GHZ and not go over 72 degrees c. I didn't want to push it much further than that. The system runs stable at those settings too. Using a Noctua D14 cooler, and the Noctua paste that came with it. Something is wrong if you are pushing over 90c at that speed. I would back it off immediately.

[Randall Leong]

Originally Posted by Craig Coston

Randall, you're going to fry that system at those temps. What CPU cooler do you have? I can stress test my i7 920 @ 4.2GHZ and not go over 72 degrees c. I didn't want to push it much further than that. The system runs stable at those settings too. Using a Noctua D14 cooler, and the Noctua paste that came with it. Something is wrong if you are pushing over 90c at that speed. I would back it off immediately.

I have a different Noctua cooler - an NH-U12P SE2.

One more problem: I spoke too soon. I have backed off the overclock, and also backed off the memory speed. At that aforementioned overclock, Prime95 ran stably, but Adobe Media Encoder crashes.

I also did a new test, and stress tests at the CPU's stock speed (yes, I now have an i7-950 which was priced low enough not to pass up) still allowed the CPU temperature to reach 70°C at even stock speed.

Now, I'm undecided as to what to do next. I never got the opportunity to replace anything yet. I originally wanted to replace either the motherboard and/or the CPU cooler. But I could also replace the memory.

And for the record, the Gigabyte motherboard actually "overvolts" the CPU at default settings. With my current i7-950, the CPU core voltage is set to a marked 1.25625V by default (1.232V after Vdroop). I tried lowering the CPU voltage, only to not run stably at higher than stock speed.

So much for this lower-end Gigabyte X58 motherboard. My previous Intel DX58SO was actually more stable than the GA-X58A-UD3R even though it has only four DIMM slots (primarily because it lacks both SATA 6 Gbps and USB 3.0 onboard).

UPDATE: I lowered the voltage setting to a labeled 1.23125V (actual: 1.2V to 1.22V) at 155x24. The CPU temps still reached 80°C at this setting after running Prime95 for six hours.

[Randall Leong]

And as it turned out, I could not get either of my i7 CPUs stable above 3.5 to 3.6 GHz without raising the voltage above a marked 1.2V.

Intel has recently replaced its DX58SO motherboard with a newer DX58SO2, which now has the six DIMM slots that most other X58 boards have. It now also has an onboard SATA 6 Gbps controller with two SATA 6 Gbps ports. Still no onboard USB 3.0 capability yet. But since that motherboard costs as much money as an Asus P6X58D-E, I'd pick the Asus in that price range.

On the Sandy Bridge front, the initial Sandy Bridge platform - the mainstream LGA 1155 - promises to be even faster than current i7 platforms clock for clock. Unfortunately, it will continue to be saddled by an insufficient number of full-bandwidth PCIe 2.0 lanes, especially since some of the theoretical maximum of 24 PCIe 2.0 full-bandwidth lanes (16 on the CPU, plus up to eight on the P67 PCH) will be eaten up on most such motherboards by a PCIe-to-PCI bridge controller, an onboard USB 3.0 controller and an onboard Realtek or Marvell NIC that also steals one PCIe lane for itself. Therefore, the current LGA 1366 platform will continue to be the choice for a serious video editing system until the release of LGA 2011 and/or LGA 1356 later in 2011.

[Randall Leong]

Originally Posted by Craig Coston

Something is wrong if you are pushing over 90c at that speed.

I think I found the problem: It's not my CPU cooler or motherboard per se, but I might have had the CPU cooler orientation wrong. In general, tower CPU coolers work best when oriented perpendicularly to the direction of the CPU retention mechanism. I had the cooler mounted parallel to the retention mechanism. And the Gigabyte motherboard, like many other "good" motherboards, has the CPU retention mechanism oriented vertically (when the motherboard is installed in a tower case). My old Intel motherboard, on the other hand, has the CPU retention mechanism oriented horizontally when that board is installed in a tower case. This might explain why my Intel board was more stable than my current Gigabyte board ever has been.

As a result, I will try reorienting my system's CPU cooler.

I will report back with the results.

UPDATE: Just this simple reorientation of the CPU cooler resulted in a roughly 9-degree drop in the CPU temperature. Whereas the old orientation that I used delivered an 81°C maximum result at 3.72 GHz, the new orientation (with the fan blowing upwards towards the case's top fan) delivered (so far) a 72°C maximum at these same speeds and voltages.

[Craig Coston]

Are you using thermal paste? If so, what brand? I tested between Arctic Silver 5 and the Noctua branded one that was included with my cooler and the Noctua branded one seemed to be significantly better. Make sure you are following the instructions on the paste and orienting the paste in a way that it lines up with the chips inside the CPU properly. Your temps are still way too high, even at 9 degrees cooler.

[Randall Leong]

Originally Posted by Craig Coston

Are you using thermal paste? If so, what brand? I tested between Arctic Silver 5 and the Noctua branded one that was included with my cooler and the Noctua branded one seemed to be significantly better. Make sure you are following the instructions on the paste and orienting the paste in a way that it lines up with the chips inside the CPU properly. Your temps are still way too high, even at 9 degrees cooler.

I took the heatsink off, and discovered that the thermal paste did not spread evenly. The center of the CPU got much less paste than the edges despite me using a single line down the center (vertically). This led me to believe that the heatsink was not sitting perfectly flat on the CPU no matter what I tried: The parts that got more thermal paste actually have poor contact with the heatsink.

And yes, I was forced to use Arctic Silver V because I ran out of the Noctua paste, and I could not find it in the stores that I frequently shop at. AS5 requires as much as a 200-hour curing time (this means that the system must be powered on and off during this period, in increments of several hours).

By the way, Craig, what program did you use for temperature readings? I use RealTemp, which supposedly reads the temperatures on each individual core. And RealTemp usually reads higher than many other temperature monitoring programs.