If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below. |
|
|
Thread Tools | Display Modes |
#1
|
|||
|
|||
Unlock AMD X3 455?
I've got a GA-880GA-UD3H, and an Athlon II X3 455.
I'm using a stock heatsink, and a 500W power supply. Unlocking the CPU reveals a fourth core. Given the PS and the heatsink, should I have any issues if the CPU is unlocked/ |
#2
|
|||
|
|||
Unlock AMD X3 455?
red floyd wrote:
I've got a GA-880GA-UD3H, and an Athlon II X3 455. I'm using a stock heatsink, and a 500W power supply. Unlocking the CPU reveals a fourth core. Given the PS and the heatsink, should I have any issues if the CPU is unlocked/ The CPU support chart for the motherboard, handles up to 140W processors, so that's not going to be a problem. http://www.gigabyte.com/support-down....aspx?pid=3789 The disabled core, was probably wasting power just sitting there. When they're disabled, it doesn't mean they're not still connected to VCore. There was a proposal at one time, sort of a "what-if", to turn off power on individual cores. But to do that, might require separate race tracks (power rail rings) on the silicon die. And I don't really think they like that idea too much. There is a limit to how many race tracks you can have around the edge of the die. So I would assume the locked core was still receiving power. They could remove the clock entirely from that core, or use clock gating, as other means of reducing power. But on modern silicon, that doesn't necessarily reduce the power on that core to zero either. My expectation would be, your idle power won't change, while your 100% load power will go up slightly. I wouldn't expect the change to be major. ******* You need to test the "before" and "after" cases. Ideally, you'd want test software, which tests execution of all possible instructions. Intel and AMD will have these in-house (after the FDIV bug, such applications are crucial, and should be quite detailed in what they check). I'm not aware of anything like that, available for end users. Only the factory can thoroughly certify the core (with test vectors for 100% of chips, or software tests as a functional test of correctness of design). The next best thing, is to use an app that looks for execution errors. As an old timer, I'd use Prime95 (stress test option), as it does some math, and knows what the answer should be. Prime95 makes the processor hot, so it also tests the adequacy of the cooling system. http://www.mersenne.org/freesoft (download page) First, you'd run Prime95 on the three cores, without unlocking the fourth. This establishes the baseline conditions for the processor. You check CPU temp with Speedfan (almico.com), while Prime95 is running. Prime95 starts a test thread per core, so there will be three test threads for the baseline run. You shouldn't get any errors, in a four hour run of Prime95, with your three cores. You don't want any of the test threads to stop on an error. If the test is clean after four hours, you can stop it (stop, and exit). If your processor isn't stable and error free with the three cores, then stop right there. You're not ready to unlock. Things would only get worse if you unlocked. You need to fix the reason for the errors first, before you go any further. During the test run, you note the temps, and whether the cooling in your system is good enough. Say we aim for 60C or 65C as an upper bound for the Tcase perhaps. (Tcase and Tdie can be different, and you have to be careful not to freak out, by measuring the wrong one and jumping to conclusions. The die gets hotter than the case, and some measurement methods are measuring the silicon die, which can be 25C hotter than the casing.) On modern systems, the temp measurement is more likely to be the silicon die kind. So if the three-core run is error free, and there seems to be some cooling margin, then you're ready to unlock. Before unlocking, you back up your Windows C: drive. That protects you in case the fourth core causes damage to the Registry or the like, on the Windows partition. (Doing the Prime95 tests with a Linux LiveCD, is a way to avoid this, but then the temperature measurement step is more complicated.) So with Windows backed up, you enable the fourth core, and do another Prime95 run. Did one of the test threads stop ? Did it stop quickly ? Perhaps the fourth core really stinks. Or, if the four hour run passes with just as much ease and the baseline run, you're home free. (Prime95 only covers a fraction of all the possible instructions on the processor, so someone at the factory would laugh at this test as an "acceptance test". But what can we do ?) I wouldn't expect the temp rise, on a Prime95 run with the fourth core enabled, to go that much higher. As long as you have a bit of headroom when three cores were running, it'll probably be OK. As for a 500W supply, that would only be an issue, if you had a high end video card drawing over 200W. Then you might be concerned. If you have a low end video card, again, I don't expect a problem. If you really wanted to know, you'd calculate an estimate of the system power, to be sure. If your video card has no PCI Express connectors on the end of it, for example, you've got nothing to worry about. Such cards are 50W or less. Paul |
#3
|
|||
|
|||
Unlock AMD X3 455?
On 3/7/2012 8:43 PM, Paul wrote:
[too long to quote] Thank you, Paul, for an informative, thorough, and detailed reply. It was very useful. -- red floyd |
#4
|
|||
|
|||
Unlock AMD X3 455?
On 3/7/2012 8:43 PM, Paul wrote:
red floyd wrote: I've got a GA-880GA-UD3H, and an Athlon II X3 455. I'm using a stock heatsink, and a 500W power supply. Unlocking the CPU reveals a fourth core. Given the PS and the heatsink, should I have any issues if the CPU is unlocked/ The CPU support chart for the motherboard, handles up to 140W processors, so that's not going to be a problem. http://www.gigabyte.com/support-down....aspx?pid=3789 [redacted for length] ******* You need to test the "before" and "after" cases. [redacted for length] I suspect it will be OK. I accidentally ran for a couple of weeks on an unlocked CPU without realizing it... I only noticed after a reboot that Linux was reporting 4 cores. The only issue I had was that the USB keyboard sometimes wasn't being properly reset by the BIOS, and I had to reboot multiple times to get into setup. Thanks again, red floyd |
#5
|
|||
|
|||
Unlock AMD X3 455?
red floyd wrote:
I suspect it will be OK. I accidentally ran for a couple of weeks on an unlocked CPU without realizing it... I only noticed after a reboot that Linux was reporting 4 cores. The only issue I had was that the USB keyboard sometimes wasn't being properly reset by the BIOS, and I had to reboot multiple times to get into setup. Thanks again, red floyd As long as you run a decent test to prove the fourth core works, it should work great. AMD sometimes takes perfectly good CPUs, and uses them to fill that SKU (make three core processors). And in other cases, the CPU really does have a defective core. So there's no way to know what you're really dealing with. Looking at the reviews on Newegg, sometimes gives a hint as to which type they're shipping at the moment. Three core CPUs were invented, as a way to increase yield from the four core production line. But if the four core yield is good, they can't provide enough defective ones, to make the three core product. And that's when they ship some of the good ones to fill demand. Since the defective core, is connected to the same power source as the good cores, at least you know it isn't so defective, as to render the entire chip dead. But it could still have an "at-speed fault", and then you need software which tests all ~1000 x86 instruction combinations, to know for sure you have a good one. And I'm not aware of any applications (for free) that do that. While running Prime95 is fun, it would only cover a small portion of all the instruction set. You could execute a SQRT() some day, and the answer could come back wrong, from the bad core :-) Even on "good" CPUs, they occasionally find an instruction sequence which causes "noise problems" on the silicon die. And the computation ends up wrong. But when these are discovered, they're found by doing a test at the factory, which involves instruction sequences which don't happen in real programs. So even what we consider to be "good working CPUs", actually have hidden weaknesses. It's the fact that compilers only use 20% of the instruction set, and seldom achieve max code density for FPU sequences, that prevents those bugs from causing end users a problem. Paul |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
The return of the unlock | Jim[_31_] | Ati Videocards | 0 | December 27th 10 06:43 PM |
455 TomTom GPS 45583 | Morales | Homebuilt PC's | 0 | July 16th 08 10:27 PM |
3800+ x2.....unlock...... | Davide | Overclocking AMD Processors | 3 | April 4th 06 06:17 PM |
HP DJ2000c err 455 | val | Printers | 0 | June 21st 04 04:30 PM |
Unlock a week 39 or later AMD CPU | LBJGH | Overclocking AMD Processors | 1 | January 1st 04 07:57 PM |