Athlon XP Overclocking: Results

[Guest Max Cowgill]

No solid benchmarks (yet) but as soon as I get the time to d/l 3dmark and if I ever get Sandra working I'll post the results. Results:Athlon XP 1700+ (AGOGA Palomino core)stock rating: -11x multiplier -133MHz FSB -1.75V Vcore-retail HSF-1.466GHz o/c'd: -stock multiplier -150MHz FSB -stock voltage -retail HSF-1.650GHz (16.666MHz shy of true Athlon XP 2000+ rating)o/c'd temps w/open case: idle: 42C (CPU) 34C (case)full load under Prime95 stress testing: 48 C peak (CPU) 38C (case)100% stable (so far, <>)under full load Prime95 stress testing @1.65GHz/150FSB. I checked out the Athlon XP overclocking database at vrzone (http://www.vr-zone.com/guides/AMD/AthlonXP/) and, considering the circumstances (stock Vcore, retail HSF, no case fans, CL2.5 DDR RAM) I'd say this is a pretty darn good overclock, especially compared to a lot of the other oc's out there using exotic cooling setups, higher Vcores, and modded mobos! It appears to me as though no one on that list has a stock Vcore with a retail HSF and has gotten anywhere near the overclock I have... Going from 1.466GHz to 1.650GHz (basically 3 full speed grades higher) is more of an overclock than I expected, and I think with a *REAL* cooling setup I could get a monster o/c out of this thing...Ok, that's enough out of me for tonight. Just thought I'd share my results with you all :)Max Cowgill

[John_Tavendale 453]

Nice overclock :)I can't get my damn FSB past 140, it won't post. I have good solid hardware, cool temps.. something is holding me back.. I think it's my PCI modem maybe.My system:-XP 1700+ (AGKGA core. Maybe that's holding me back) running @ 11 x 140 = 1540Mhz cooled by a Swiftech MCX460 with an 80mm DeltaAsus A7V333 Motherboard256MB Corsair XMS3000 DDR. Garanteed to run @ 185Mhz Cas2.Leadtek GeForce 4 Ti4400 @ 300/650Creative Audigy DEPCI modem

[ShezA 176]

What is your view on the recent releases by Intel of the 2.4 P4? Apparently the new Northwood chips are rock solid and very fast.Do you think Intel is now going to storm ahead, and AMD will fall behind?Naughty topic.Shez

[John_Tavendale 453]

Say after me AMD AMD AMD AMD AMD :-lolNot too sure. AMD are bringing out some new CPU's soon I believe.But I know of a few AMD users switching back from the Dark side..

[Guest Max Cowgill]

Thanks John. Funny you should mention the PCI modem thing, since I have one too... Haven't noticed any instabilities while online either... Maybe I just got lucky? Then again, my PCI modem is the only physical PCI device I have (although the onboard sound and LAN appear as PCI devices). Have you tried another modem? hehe, it's kind of funny that I'm getting a full 110MHz more than you with a retail HSF vs. your aftermarket HSF w/Delta fan ;) Sorry, don't mean to rub it in your face... I'm sure you're at least a bit upset with your overclock. Perhaps your RAM timings are too agressive? I have mine pretty conservatively set (although still pretty darn fast according to Sisoft Sandra 2002 (will post results momentarily)). It could also be your Audigy card (I assume that's a PCI device, since I haven't heard of an integrated Audigy chipset as of yet).Max Cowgill

[Guest Max Cowgill]

>What is your view on the recent releases by Intel of the 2.4 >P4? Apparently the new Northwood chips are rock solid and >very fast. >>Do you think Intel is now going to storm ahead, and AMD will >fall behind? >>Naughty topic. I'm not touching this one with a 10 foot pole, especially not now that I actually own an AMD CPU, and don't just have to rely on 2nd/3rd-hand info anymore ;) I used to be a pretty avid AMD enthusiast (despite having only owned one AMD CPU and it not even being an Athlon), you can imagine how much of one I am now, considering my success so far. As far as the Northwood P4's go, I think they're a nice upgrade over the previous Willamette P4's, but I don't think the architecture Intel has chosen for the P4 is the right one for today's applications. Perhaps in the future the P4(and beyond) architecture will prove to be faster than the Athlon(and beyond) but I still think that the Athlon architecture is just as fast (if not faster) as the P4 architecture in most of today's applications. Synthetic benchmarks aside (don't get me started on synthetic benchmarks, they're usefull, but not the end-all-be-all measure of performance) the Athlon is generally faster than the P4 in most real-world apps. just my two cents,Max Cowgill

[John_Tavendale 453]

Max,Yep my audigy is PCI. I tried the RAM settings at 2.5, 3, 3, 3, 6. Turned off Turbo, increased voltage and no go. I might rip out my modem and soundcard and see if that makes a difference.

[Guest Max Cowgill]

I finally got Sandra 2002 to work (silly missing .dll file) and here are the results, along with the results of an Athlon XP 1800+ on KT266A chipset w/PC2100 DDR RAM @ 133MHz FSB (DDR), and P4 2.0GHz on I850 chipset w/PC800 RDRAM @ 100MHz FSB (quad-pumped). Tests were run 5 times each with no background tasks running and the results were the medium score from each group of tests.My system (Athlon XP @ 1.65GHz w/150MHz FSB (DDR)):CPU Arithmetic Benchmark:Dhrystone ALU MIPS: 4552Whetstone FPU FLOPS: 2280CPU Multimedia Benchmark:Int EMMX/SSE it/s: 9026Float SSE it/s: 10366Memory Bandwidth Benchmark:Int EMMX/SSE MB/s: 2151Float EMMX/SSE MB/s: 2018P4 2.0GHz w/100MHz FSB (quad-pumped):CPU Arithmetic Benchmark:Dhrystone ALU MIPS: 3688Whetstone FPU/SSE2 FLOPS: 1040/2440CPU Multimedia Benchmark:Int EMMX/SSE it/s: 7875Float SSE it/s: 9625Memory Bandwidth Benchmark (note: results from 1.6GHz P4, not 2.0GHz):Int EMMX/SSE MB/s: 2450Float EMMX/SSE MB/s: 2450Athlon XP 1800+ system (1.53GHz 133MHz FSB (DDR)):CPU Arithmetic Benchmark:Dhrystone ALU MIPS: 4240Whetstone FPU FLOPS: 2124CPU Multimedia Benchmark:Int EMMX/SSE it/s: 8370Float SSE it/s: 9760Memory Bandwidth Benchmark:Int EMMX/SSE MB/s: 1925Float EMMX/SSE MB/s: 1767Extrapolations:CPU Arithmetic Benchmark:Dhrystone MIPS: my system vs. Athlon XP 1800+ = ~7% fasterDhrystone MIPS: my system vs. P4 2.0GHz = ~23% fasterWhetstone FPU FLOPS: my system vs. Athlon XP 1800+ = ~7% fasterWhetstone FPU FLOPS: my system vs. P4 2.0GHz = ~119% faster/~6% slowerCPU Multimedia Benchmark:Int EMMX/SSE it/s: my system vs. Athlon XP 1800+ = ~7% fasterInt EMMX/SSE it/s: my system vs. P4 2.0GHz (SSE2) = ~15% fasterFloat SSE it/s: my system vs. Athlon XP 1800+ = ~6% fasterFloat SSE it/s: my system vs. P4 2.0GHz (SSE2) = ~8% fasterMemory Bandwidth Benchmark:Int EMMX/SSE: my system vs. Athlon XP 1800+ = ~12% fasterInt EMMX/SSE: my system vs. P4 1.6GHz (SSE2) = ~14% slowerFloat EMMX/SSE: my system vs. Athlon XP 1800+ = ~14% fasterFloat EMMX/SSE: my system vs P4 1.6GHz (SSE2) = ~21% slower :-eek What does all this mean? Well, it means I've got a pretty quick system ;) Oh, you wanted analysis? Fine... In x86/x87 ops (int/float) my CPU is ~7% faster than a 1.53GHz Athlon XP which is almost parallel to the ~8% clock speed difference between the two CPUs (the different FSB speeds don't affect this test, as all the ops are carried out within the L1/L2 cache of each CPU, and should never have to access system memory due to the small code size of this specific benchmark). Even more interesting is the result compared to the P4. My CPU is ~23% faster than the P4 in pure x86 (int) ops, and ~119% faster in pure x87 (float) ops :-eek while being ~6% slower than the P4 when the P4 uses SSE2 ops.Nothing too interesting in the Multimedia benchmark... My CPU is ~7% faster than the Athlon @ 1.53GHz in Int MMX/SSE iterations, and ~6% faster in float SSE iterations, while being ~15% faster than the P4 2.0GHz in Int MMX/SSE (vs. SSE2) iterations, and ~8% faster in Float SSE iterations (vs. SSE2) iterations.Memory performance is quite interesting, however. A 17Mhz (~13%) increase in FSB speed vs. the Athlon XP 1800+ yields ~12% increase in Int memory bandwidth and ~14% increase in Float memory bandwidth, once again, demonstrating a near-linear performance increase in comparison to a "regular" Athlon XP on a 133MHz FSB. Unfortunately, the results of my system compared to the P4 system in memory bandwidth don't really stand up (despite the results only being from a 1.6GHz P4, not even a 2.0GHz P4) being ~14% slower in Int memory bandwidth and ~21% slower in Float memory bandwidth. Ok, so what DOES all this really mean? For the purposes of o/c'ing Athlon XP's via the FSB, one can expect a near-linear increase in performance in both CPU performance, and memory bandwidth performance. I'd rather not delve into the Athlon XP vs. P4 results at this time, except to say that they are entirely different architectures, designed with different goals in mind. I'm thinking of updating my (now aging) P4 article with these results, and perhaps analyzing them a bit further...Let me know what you think,Max Cowgill

[Guest Max Cowgill]

>Max, >>Yep my audigy is PCI. I tried the RAM settings at 2.5, 3, 3, >3, 6. Turned off Turbo, increased voltage and no go. I might >rip out my modem and soundcard and see if that makes a >difference. That's interesting John... I've been running in "Normal" mode with Turbo disabled this whole time and I just went into my BIOS and enabled Turbo: I could boot into windows but when I ran the Sandra memory benchmark again it rebooted, so I know that at least on my system, the normal RAM timings are crucial to achieving my 150MHz FSB (at least with this CL2.5 PC2100 RAM). I'm surprised even with those RAM settings you can't get above 140MHz FSB... Definitely try removing those PCI cards (and make sure you remove them from Windows as well) and see if that helps. I think 150MHz truly is the limit on my current setup, I can sit here and benchmark/play FS2k2 for hours with no lockups but when I bump the FSB up to 152MHz I can't even get into Windows (haven't tried 151MHz yet but I'm happy with 150MHz and 100% stability).Max Cowgill

[ShezA 176]

MY God! :-eek That is one interesting conclusion...I guess I got my answer.CheersShez

[Guest]

A word of advice please Max?I have an Athlon 1gig @ 1.33 (10x133)on a KT7A running stable around 40C, 28C idle, but this is it's limit before failure to POST.I havn't tried overclocking the FSB.Will overclocking the FSB give me better results?ThanksPH

[Guest PaulL01]

Hi Max,I would highly recommend getting 3Dmark2001se set up and to run it as getting all the way through is a real good indication as to the stability of your setup if you cant make it through, lower memory timing, if that doesn't to it, put the memory timing back and up the voltage, just 0.10v at a time, watch your temps after you overcome that wall (you may end up needing faster memory) you can do the same thing with your CPU, adding voltage one step at a time though it will bring a little heat (its up to you to keep it cool >45-47C is good while under load) it will also give you stability, keeping it cool does too as well of course will enable it to have a nice long and happy life. :)Also, If you really want to get the most out of yer setup, unlock your CPU, bring the multiplyer down and then bring the FSB up to match yout old speed as you will be speeding up the whole system even more, as long as your hardware can take it. :)Lets us know what settings end up working for you after you run 3Dmark2K1.Great fun heh!Paul

[Guest PaulL01]

>Nice overclock :) >>I can't get my damn FSB past 140, it won't post. I have good >solid hardware, cool temps.. something is holding me back.. >I think it's my PCI modem maybe. >>My system:- >>XP 1700+ (AGKGA core. Maybe that's holding me back) running >@ 11 x 140 = 1540Mhz cooled by a Swiftech MCX460 with an >80mm Delta >Asus A7V333 Motherboard >256MB Corsair XMS3000 DDR. Garanteed to run @ 185Mhz Cas2. >Leadtek GeForce 4 Ti4400 @ 300/650 >Creative Audigy DE >PCI modem High John,True, your AGKGA is not the best for an O/C but it will do mutch better than 1540, You will need to increase the voltage to get it to run better at O/C speeds, though you should be able to move it in 0.025V increments happiness for stuborn AGKGA types usually occures at +0.100 (1.85V) then it will be finding a balance between stability vs speed vs heat and noise that you like. :)If you hit the "wall" early then the best thing you can do is to unlock your CPU and bring the multiplier down so that you can run higher bus speeds for the rest of your sytem, there alone is a 5-10% performance boost just waiting to happen.Let us know how goes it. :)Paul

[Guest PaulL01]

>A word of advice please Max? >>I have an Athlon 1gig @ 1.33 (10x133)on a KT7A running >stable around 40C, 28C idle, but this is it's limit before >failure to POST. >>I havn't tried overclocking the FSB. >>Will overclocking the FSB give me better results? >>Thanks >>PH Yes, Overclocking the FSB is the best result as the whole system benefits, Memeory/AGP/PCI all get pumped. you will find though that the wall you hit may be your memory, you may need to lower its timing/ up its voltage slightly etc.Since you can hit 1.33 you could bring your multiplier down to 9.5 and FSB up to 140 and you should be OK, if you cant do 9.5 and have to setle for 9.0x then be prepaired to be limmited to around a FSB of 143 and a sum of 1.287mhz.Paul

[Guest Max Cowgill]

Paul, I do plan on running 3dmark2k1se, but not for stability testing purposes. I've already run the Prime95 "torture testing" and Sandra 2002 "burn-in wizard" several times with nary a hickup. 150MHz FSB seems to be the sweetspot for my system; 100% stability for max. overclock. I may try upping the Vcore and Vmem settings a bit though, to see if I can push things a bit further. One problem with my current setup and running 3dmark2k1se: this POS TNT2 M64 doesn't have native DX8 support so it wouldn't be able to do some of the tests anyway, so that kind of negates the purpose of using 3dmark2k1se as a stress testing tool. Max Cowgill