The Inquirer-Home

Push a 2.4GHz Athlon64 X2 4800+ to 3.6GHz

Oh yeah, only the BIOS works fine... not the Windoze
Wed Jul 06 2005, 10:05
WILL A SINGLE core (over) clock better than the dual core in the same process and core design? Of course! That should be the case. Look at the dual core Pentium at 3.2GHz versus a single core at 3.8 GHz, or the 2.4GHz dual core Athlon64 vs its new 2.8 GHz single-core cousin.

This week I landed a simple quickie test setup from AMD with Athlon64 X2 - the 4800+ top model) and Asus A8NSLI-Deluxe mainboard, plus added 1 GB of RAM in two Corsair DDR400 XMS3200XL Pro DIMMs, the golden reference for CL 2-2-2-5 ultimate-speed Athlon64 DIMMs (oh yes, OCZ now has a nice new Gold Edition GX series supporting CL 2-2-2-5 at DDR433, more about that later), plus GeForce 7800GTX and HIS IceQ II Radeon 850XT Platinum. The GeForce refused to work above the 2.8 GHz CPU clock for some reason, so the IceQ II took over. All that fits in the masculine looking CoolerMaster Praetorian 730 casing with matching RealPower 550W EPS power supply.

The Praetorian case, besides sturdiness and the mean looks, also has a great mainboard drawer for frequent board swaps that us poor hacks have to perform often.

More about that gorgeous hardware later in the week - in the meantime, I tried to have some fun and replaced the generic fan combo delivery by AMD with my house favourite - the old Corsair HydroCool 200EX with an Athlon64 adaptor that I never tried before.

The goal: see how far up it goes within an hour - the Athlon64 X2 clock speed, of course. The bottom line: as long as it boots fine to BIOS and I can play around within it, verify it runs at that speed, and safely reboot back for another round of changes without any hang or crash, I'm fine - this does not include the Windows boot *yet* since I wanted to check the very limits of the hardware itself.

So, the air conditioner goes on as well - it was 33 C outside my test room, so no way to avoid it, and we start the game. The initial temperature on the HydroCool was 30.5 C on the boot, going up to 31.5 C after booting into Windoze and running Sandra. The temperature stayed the same, and we also ran Sandra fine, when upping the mem clk to 433 MHz and therefore the CPU clk to 2592 MHz, while still running at CL 2-2-2-5. Not bad - above 6.3 GB/s in the Sandra memory benchmark.

Move then a notch up: 466MHz memory, 2790MHz CPU, latency still 2-2-2-5. The BIOS is up fine, the temperature is 32 C, but Windows doesn't proceed. I change the HyperTransport multiplier from 5X (1166 MHz HT) to 4X, and Windows boots fine, but Sandra hangs. OK, almost there, will need to do some tweaking to get this level stable - an attractive target, since it is equal to having dual FX-57 cores with a sped-up memory to feed them both faster!

Twenty minutes left - forget about Windows, let's just crank the CPU up: 500 MHz RAM setting, and the BIOS automatically re-sets the latency to 3-4-4-8, and no I can't change it even in the manual mode (I suspect the DIMM could take 2.5-3-3-7 easily at least). Never mind anyway, we booted the X2 4800+ at 3 GHz sharp. and the temperature shows 33 C only! HT, even though at 4X setting instead of 5X, runs now at full 1 GHz anyway. We change it back to 5X to see if it will boot back fine, and it does - in any case, 1.25 GHz HT is within the HT 2.0 spec.

The next run was 533 MHz memory setting, i.e 3.2 GHz CPU. It also booted fine in BIOS, we checked all the screens, and memory latency settings stayed the same 3-4-4-8. Then, move to 550 MHz memory, or 3.3 GHz CPU clock. Again it goes into the BIOS finely, the temperature is 33.5 C and stays so for those few minutes we run it, but the BIOS has slowed down the memory latency to 3-5-5-10 now. The HT 5X setting didn't work here, had to go down to HT 4X from then on.

Three minutes left of that hour, change to 600 MHz memory and reboot. It boots fine into the BIOS again - a 3600 MHz, 3.6 GHz, Athlon64 X2 CPU! If I could make it stable, it would mean having a 7200+ speed grade according to the AMD marketspeak. The temperature jumped to 34.5 C... still reasonable. I guess something like a liquid nitrogen cooler would be needed now to ensure really stable long-term operation at such speed, but at least we're on the way - after all, booting cleanly into the BIOS and being able to manipulate it safely into these speeds is way better than, everytime you 'overclock' too much, the system just hangs, and you have to clear CMOS or whatever to get it up again. No such problems here.

Yesterday, I got hold of the FX57, and was kinda excited expecting, what, a 4GHz boot up? So a quick fix followed - same cooler, components, memory and full-blast aircon going right at the open system case. Funnily, it was identified as FX25 by the BIOS.

The system booted fine at the default 2.8 GHz, no issues at 3.02 GHz either (when memory set to 433 MHz) or 3.0 GHz (when multiplier set to 15) - Windows booted in both cases fine. Then, we went to 3.2 GHz (multiplier 16), boot was fine but Windows didn't go ahead. Then, multiplier 16 plus 433 MHz 2-2-2-5 memory, at 3456 MHz, BIOS boot still fine - and that was it. Any further attempts to break this level just didn't go through, resulting in black screen. Whether upping the multiplier without overclocking the rest, or leaving the multiplier at 16 but upping memory to 466 MHz - no go.

But something really funny happened after I set the multiplier to 18 and had to clear CMOS to reset the system - it booted up showing, believe it or not, 6104 MHz. I let it proceed to boot with that setting, of course it hung dry right after, without any video signal. I have no explanation for this 'phenomenon'.

So, in summary, it seems that the 2.4 GHz X2 could reach higher theoretical speed than the 2.8 GHz FX57? Either I got an odd pair of chips, or something else's going on - anyway, after all the physical labour, the 6.1GHz thing made my day. µ


Share this:

blog comments powered by Disqus
Subscribe to INQ newsletters

Sign up for INQbot – a weekly roundup of the best from the INQ

INQ Poll

Heartbleed bug discovered in OpenSSL

Have you reacted to Heartbleed?