Issues with Hardware Swapping

[2.71828183]

So, I have a few prospective computer upgrades to do, but first I need to figure out if they're going to hurt anything.

The first is a processor upgrade. Both mine and my dad's computers use AM3-compatible motherboards, and he's having processor-related (as far as I can tell) lag issues with his Athlon II X2 240. So I thought I'd give him my Phenom II x4 945, and buy myself a Phenom II x4 965 or 970 BE, which are both sub-$200 right now.

Both systems are running Windows 7, and my dad's has the "OEM" type with DRM tied to the specific computer it's on.

Can I just swap the processors around (with the computers off, obviously!) and turn everything on and have it work? Or do I need to do something to tell the computer I'm going to swap its processor? Is Windows' overly-moronic DRM (is there any other kind?) going to cause me problems?

Tangential question: Is the 970 BE worth the $25 premium (minus $10 rebate) over the 965 BE, for 100MHz faster clock speed? Right now, I'm leaning towards no, because it seems like a steep price hike for incremental gains, but I'd like a second opinion.

The second upgrade I'm planning is a memory upgrade. I currently have 4GB of memory in two slots (specifically, this), and I'd like to get another 4GB of memory. Can I just pick out something cheap and go with it (like this, the cheapest DDR3 on Newegg at the moment), or do I have to get something that matches the originals? For what it's worth, I'm a little unhappy with the original RAM because I never was able to run it stably above 1066 MHz, even though it's advertised to run at 1600 MHz.

[GeorgeH]

Popping in the new CPU should work just fine. OEM Windows activation has three parts; a SLIC entry in the BIOS and a license file and code inside Windows. None of them should be in the least bit bothered by a CPU change. If you decide to upgrade the motherboard, there is an individual license key printed on a sticker attached to the PC that will let you install Windows using generic install media.

If you want a 970, buy a 965 and bump the multiplier. It might take 30s and you are pretty much guaranteed that it'll work without having to touch voltages or anything else (imagine the nerd rage if a 965 would only overclock by 100MHz without lots of work.) Also, even theoretically a 970 is only ~3% faster than a 965, so you'd be very right that its performance benefits are going to be very small, to the point that I'd say they're completely indistinguishable without running multiple benchmarks.

That said, I'd very strongly suggest upgrading to the X6 1075T instead of a slightly faster X4. You'll get 50% more cores at the same 3.0GHz 945 clockspeed and the ability to Turbo to 970 3.5GHz speeds for loads that don't stress all six cores.

As for RAM, mixing and matching is just fine (although the collective will run at the speed of the slowest pair), although I'm pretty sure I'm the moron that told you "RAM is RAM" - so grab some grains of salt and whatnot.

[2.71828183]

Good to know that installation is going to be simple, thanks.

I could actually afford a x6 1090T without going over-budget, but I'm not sure if I really need six cores. Benchmarks seem to put the 1090T on-par with the 965 for lightly-threaded loads, and between 10-30% better for heavily threaded loads. The price difference is currently $60 including rebates; will I notice a substantial performance improvement in real-world usage?

[Endless Mike]

Depends what you're doing. Gaming, probably not. Film editing, very possibly.

[GeorgeH]

As Mike sans termination said, it depends on your compute tasks. The main reason I think you should go with a hex core is that without overclocking a 965 isn't really going to blow away a 945 at anything ever, while a hex core has a chance to do so either now or in the future if software or your compute needs change.

The one thing I should have noticed earlier is that the Athlon X2 240 is a 65W or 45W CPU, and the Phenom II X4 945 is a 95W or 125W CPU; the OEM motherboard may be completely incapable of powering a 945.

[2.71828183]

Valid point. I think I'll go with it and find out. $60 isn't really all that much extra, in the grand scheme of things.

I didn't even think of checking power requirements, but I checked and the processor is in the compatibility list, so I'm good. Both mine and my dad's motherboards are Gigabyte 785G's, so even though his is a different form factor, I figured the compatibility lists would be the same.

[2.71828183]

After some fiddling around and a BIOS update (which I missed when I looked up my motherboard on the manufacturer's website, otherwise I would've done it first and not had to swap the processor again to do it), I got the x6 processor working. (I haven't swapped my dad's yet.) The hardest part of the whole ordeal was fiddling with the mounting clips on my aftermarket heatsink, and I wouldn't even be griping about that if I didn't miss the BIOS thing and have to do it twice.

I ran a benchmark in POV-Ray v3.7 beta: the values are pretty much exactly what they "should" be given the clock speed differences. Single-core benchmark dropped from 5m 45s to 4m 49s, and the all-core benchmark dropped from 1m 28s to 55s. Interestingly enough, my monitoring program never shows a single-core frequency above 3200MHz (instead, with a single core load, it shows all six cores at 3200MHz and about 10-15% utilization), but given the single-core speed, the Turbo feature seems to be working as expected.

I'm toying with the idea of doing a mild overclock, but I don't want to sacrifice stability. I have plenty of cooling (Scythe Kabuto aftermarket heatsink), but I don't want to sacrifice its current near-silence by making the fan run at high speed all the time either.

I've never done overclocking before; this is about what I know about it. How much might I be expected to gain from overclocking within those constraints? Is it even worth messing with?

[GeorgeH]

How much you're going to gain and it being worth it are up to you and your CPU. I haven't overclocked an X6 before, but I'd play around with AMD's overdrive utility and see what happens. Google will find you tons of OC guides, but here are my three golden rules for casual overclocking:

Don't mess with voltages. It's just not worth it unless you know what you're doing, and even then the amount of testing you have to do for the minimal gains is generally pretty silly (a rough expectation is exponential increases in voltage/power for linear increases in clockspeed.)

If you've got an unlocked multiplier, play with that first (and pretty much exclusively); that way you don't have to worry about testing anything but the CPU for stability.

Once you find your highest stable overclock, take it down one notch. That means either dropping the multiplier by one notch or reducing the total overclock by ~10% (eg. if you go from 3->4GHz, drop back to 3.9GHz.)

[2.71828183]

Thanks for the tips. That OverDrive program is much more comprehensive and intuitive than what came with my motherboard.

So far, I've gone from stock multiplier settings of 16x (18x turbo) to 18x (I haven't changed turbo settings yet), going in half-steps and testing for stability as I go. I've reduced to 17x for the time being.

So far, I haven't run into any stability problems. What I have run into is a few quirks that I want to make sure are okay before proceeding. One, while running the stability test, I noticed that occasionally one core would drop a half step or even a full step in multiplier for one refresh cycle, then go back up. This effect was present when I tested 17.5x and 18x, and I didn't notice it on any of the earlier tests. Is this some kind of throttling behavior, or just a normal hiccup?

Temperatures, as far as I can determine (the hottest temperature is listed as a system temperature, not even a CPU temperature; I assume that it actually is a CPU temperature and the labels are wrong) have stayed below 60°C on all tests. Interestingly, the processor seemed to produce the highest temperatures and highest maximum fan speeds on the 17x test, and slightly less on the two higher-clocked tests. Subjectively, though, the fan noise was more noticeable on the higher-clocked test, so I conclude that the average fan speed was higher for the higher clock rate.

The CPU voltage seems to automagically rise proportionally when the multiplier is increased. Is it supposed to do that? So far it hasn't exceeded the max voltage specified for the turbo setting.

Also, my screensaver came on during one of the stress tests and promptly crashed (with a mildly humorous error message something like "Blank Screen is not responding"), though I'm pretty sure that's just because it was starved of CPU cycles and something timed out.

Completely (I assume, since it happened before I started) unrelated to overclocking but related to the recent upgrades, there's something making an extremely intermittent two-toned squeaky chirpy noise (of about a second's duration, maybe once every ten to twenty minutes or so). I suppose it has to be the processor fan, the only moving part in what I messed with, but it's a distinctly un-fan-like sound. I'm kind of stumped at the moment, so I guess I'll poke my hand in there and jiggle the fan mounts around in case one is squeaking against the heatsink or something.

[GeorgeH]

Chirping - Guess 1, power regulation circuitry on the motherboard; could have been caused by BIOS update or by rapidly changing power demands of X6 CPU (throttling up/down.) Guess 2, power regulation circuitry in the PSU; could be caused by shifting power demands on rail x due to new CPU. Guess 3, system speaker going off; could be caused by system software or by BIOS freaking out due to some error state (possibly a temperature alarm that you can set in the BIOS.)

Multiplier shifting - Dunno. If it only happens at 17.5/18x, my guess would be the Turbo code freaking out due to overlap in base clock and Turbo frequencies; personally I'd disable Turbo until I found a stable overclock I was happy with, then add it back in only if it made sense to do so when compared with that overclock.

Temperatures - Core and CPU temperatures are what you want to look at. My personal "that's too high" threshold when fully loading an AMD CPU is <=60C for the core temp, and I'd be very happy with mid 50's. I tend to ignore CPU temperature; IMHO it's not really an accurate gauge so much as a dummy light, and as long as it's not triggering thermal panics I"m happy.

Automagic voltage increase - Just fine. If you want to be slick you can probably decrease the automagic value a bit and still be stable while consuming less power; I'd just avoid increasing it on your own if the automagic setting isn't stable.

Screen Saver - Sounds like a software bug, not a hardware issue. It's generally a good idea to disable power saving and other "idle" routines in the OS while doing stability tests.

[2.71828183]

Sorry for the huge reply lag; real life kind of caught up with me.

Re chirping: I'm almost positive now that your guess 1 is correct.

What I've done with overclock so far is set my multiplier to 17x (3400MHz) for normal use, and 19x (3800MHz) for Turbo. I've also undervolted the processor by 0.025V (I think; don't want to reboot and get into the BIOS right now to check that). The reason being, voltage increases proportional to clock speed. At 3800MHz, it thinks a "normal" voltage is 1.475V, instead of the 1.3V or whatever it normally expects for Turbo. Voltage goes down again as the processor throttles back; the modest undervolt is the most I could do and have the processor stable at its 800MHz idle frequency. It runs at 1.225V at that frequency. I don't think my motherboard's BIOS has an option to disable this feature. It may be linked to AMD's Cool'n'Quiet, but I don't want to get rid of that.

Anyway, the reason I'm using high Turbo clocks and not increasing the main clock multiplier much is for power and heat reasons. Not so much the processor, though: temperature readings say it idles at a minimum of 18.88°C (almost exactly room temperature; just barely believable considering the CPU HSF draws most of its intake air directly from outside the case) and I'm getting a max of 38.12°C over both three-core Turbo and six-core regular loads, which is nowhere near concern levels. Unless the temperature sensors on all six cores are reading consistently low, that is. I'm not too skeptical of this, because I can touch anywhere on the CPU heatsink during a load test and it's no more than mildly warm.

The problem, however, lies elsewhere in my case. TMPIN0 reads a min (idle) of 35°C and max of 41°C; TMPIN1 reads a min of 31°C and a max of 48°C; and TMPIN2 reads a min of 39°C and a max of 55°C. I've seen TMPIN2 up to 61°C during higher-clocked (18x) six-core loads.

It would be nice to know exactly where on the motherboard these sensors are, but I have my guesses. I'm guessing northbridge, southbridge, and VRMs, in that order, based on the highly scientific finger test. Since I have a low-noise (and thus low-airflow) computer, none of these components gets much airflow. The VRMs in particular are in kind of a dead spot for airflow, and they're un-heatsinked to boot. Even though I have a down-flowing CPU fan, the airflow through the heatsink is pretty slow even at high CPU load, and it's immediately exhausted out the rear of the case by the rear case fan.

It would be nice to get these components some air, which might allow me to raise the main multiplier that much more. I don't want to sacrifice silence to do it, though. I have an intake fan in the front of the case, which blows air over the HDDs (which remain cool to the touch). I'm thinking of moving that directly below the CPU fan, which should point some air at the NB and SB. I don't know how that'll affect HDD temperatures, but I plan to migrate mostly to an SSD fairly soon, so that might be a mitigating factor (the HDDs will still see some use, but only when I'm running backups or accessing files too large to keep on the SSD). The other problem with this is there's not actually a place designed to mount a fan where I want to put it, and it's in the middle of a flimsy panel so I might have to cook up some reinforcements to keep it from vibrating and making noise.

The VRMs, I'm not too sure what to do with. They're in an awkward position to route airflow to with any kind of fan, since they're mostly occluded by the side of the CPU heatsink. I suppose I could buy some of those stick-on heatsinks (like so). That won't get any more airflow up in there, but maybe convection and increased dissipative surface area will do its part.

Memory is also pretty warm, but not uncomfortable to the touch like the components I'm concerned about.

Thoughts? Ideas? Am I worrying about nothing?

[GeorgeH]

Overheating VRMs obviously aren’t a good thing. Sometimes motherboard makers will assume that you’ll be using a stock heatsink that blows perpendicular to the board, providing airflow to the nearby VRMs. A tower cooler that blows parallel to the board combined and the increased power demands of a fancy X6 might be enough to cause issues.

Unfortunately I can’t say much more than that, though. Those temps really don’t look bad to me (and your PC probably isn’t going to blow up anytime soon) but if you’re planning on keeping that hardware for a long time those heatsinks are pretty cheap insurance. I’d personally try to rig up some kind of fan before going that route, as IMHO “passive == quiet” and “fan == noisy” are false equivalences, but YMMV and whatnot.

[2.71828183]

I do have a down-blowing (not a tower) heatsink, but it uses a low-speed fan with large fins, so very little air gets past it (and my rear case fan exhausts most of that immediately). My problem is I can't figure out a way to get a fan in there. I have four fans (two case fans, a PSU fan, and a CPU heatsink fan), all of which are much quieter than my HDD, but looking from the side of the case, the VRMs are shadowed by the edge of the CPU heatsink. Likewise, the exhaust fan doesn't pull much air across them, because it's spaced away from the motherboard by all the ports.