Overclocking effects on CPU longevity

bigben2k · 02-10-2003, 10:57 AM

(Struggling to find the right Forum)

Here's an interesting tid-bit of (unverified) information.

In short, CPUs are spec'd to run for ten years, at stock voltages. (True? Anyone with contacts at Intel or AMD?).

What we don't have is a measure of the effect of OC'ing...

Personally, I don't believe that I'll be using the same PC for ten years, but I would take comfort in knowing that it'll run for "about" 5.

airspirit · 02-10-2003, 11:44 AM

IMHO, I don't mind if my CPU only lasts 3 years. Honestly, in the crowd that we're in, how many of us would be bothered if our CPU went out faster than that? By the time that $200 processor is showing age in a couple of years, you can slap a brand spankin' new one in for $30. For the amount of speed you gain over the life of both chips, you're coming out ahead.

For certain people I build machines for, I give them this logic laid out. I ask them to order in importance the following three items: speed, price, and longevity. Once that is done, I explain that I can get them a 1700+ TBredB for $50, a 2100+ TBredB for around $100, or a 2700+ for around $300. With the first, I can get it to run slightly slower than the two more expensive chips, but it will most likely die in 2 years. With the other two, I don't have to push them as hard, so they'll probably last much longer, though in three years they'll be thinking about an upgrade anyway.

Note: I ALWAYS explain to my customers the values and risks of overclocking their machines either by just a little or a whole lot. I have never had anyone refuse to have it buffed. Every once in a while I'll get someone who is leery, so as I'm building the machine and stress testing it at higher speeds I have them sit next to me and watch. Once they realize the relatively safety of a moderate overclock, even the most conservative people insist I leave it faster. Even at my business, every computer I've built for them is overclocked at least 20%. The price/performance advantage makes this the logical move.

Needless to say, I've never had a single person go for a 2700+. Most go for the 1700+ knowing that even if it toasts in a year, they can buy better than the 2400+ to replace it and STILL come out ahead financially, and some go for the 2400+ because they don't want to be bothered with upgrading for a while. The amount of importance they put on longevity also affects how far I push the chip in regards to how far it'll go (I typically keep it at 50-80 Mhz below peak stable speed, though for longevity people I lower it another 100 Mhz to lower the voltage further).

In all honesty, CPUs can almost be considered a disposable commodity anymore: you can OC a cheap one to the speed of an expensive one, and even if you fry three or four of them, you'll still be ahead pricewise. If you can buy a computer for $500 that performs like one for $1000, why would you care if in a year in a half you have to replace a $50 part or two? In the end you come out ahead.

airspirit · 02-10-2003, 11:45 AM

On another note, a constant upgrade cycle gives us geeks something to do on the weekends besides drown in cheap beer.

bigben2k · 02-10-2003, 11:58 AM

Well, that's kinda my point.

If you run a 1700+ to 2100+ specs, but you knew that you could expect a decreased longevity from 10 years to *say* 5 years, then it's worth it. but if the longevity is down to *say* 1 year, then maybe it's not so worth it...

With your $ figures, assuming that the user will use his PC for 3 years, and it's overclocked to last 1 year, you'd be better off spending $100, rather than three times $50.

But if the resulting longevity is more than 1.5 years, then it's even: that's the break-even point.

Of course this doesn't take into account the cost of the cooling solution.

What I'm trying to get to here, is to build a business case (cost justification) of OC and watercooling a PC. I think we all know that watercooling is *mostly* not worth the investment (IMHO), but it is a perfectly good hobby! Spending $300 on a kit is *definitely* out there.

Does anyone have any data on the decreased longevity resulting from overvolting/overclocking/temperatures?

(actually, I posted temp effects on longevity recently... now where did I put that up?)

airspirit · 02-10-2003, 12:10 PM

From what I've seen on forums (totally unscientific), on AMD processors, a overvolting of .3V reduces life expectancy to about 1.5 Years when temperatures average around 50C (2.05 on a palomino, 1.9 or thereabouts for TBreds). Lower temps from watercooling and phase change cooling tend to reduce that effect a bit.

Two things you failed to account for in your example was the downward march of CPU prices per speed index, and the intangible cost benefit of having a faster machine immediately. The example should be:

One 2400+ now for $200 OR
One 1700+ now for $50 (at 2400+ speeds) and a 2700+ for $50 in 1.5 years.

In the end, you're ahead in every category, unless the hassle of swapping chips is not worth the extra money, in which case you buy the highest speed you can afford, and leave it stock, though this will end up being a much more expensive solution.

bigben2k · 02-10-2003, 12:31 PM

Point taken! I'll add it to my business case

airspirit · 02-10-2003, 01:22 PM

Business case?

pHaestus · 02-10-2003, 01:36 PM

I abuse CPUs so I don't bother buying expensive ones

I am running my 1700+ TBredB at 2.05V and 2400MHz now. I will continue to run it at such speed until it "doesn't go". I'll keep you posted.

I have been overvolting the crap out of processors for a LONG time now, and never had a CPU die. I typically run them flat our til I want to upgrade, and then put them in systems for family members clocked back to whatever they will do near default voltage. I have yet to have the first CPU turn up dead: the Tbird 1.0 that ran at 1.625 and 2.1V, my old Celeron2 566@952 2V are both still purring along years later.

bigben2k · 02-10-2003, 01:39 PM

Quote:

Originally posted by airspirit
Business case?

Cost benefit analysis.

Are there CPU price statistics anywhere? Google time...

airspirit · 02-10-2003, 02:46 PM

Check two places: pricewatch and Newegg.

Pricewatch gives you a general "historical" overview of both the newest and oldest chips, while Newegg gives you a look at verey reasonable prices on the correct steppings for the newest lowbies (like 1700+ TBredB chips). Those two sites will give you a peek at the exponential rise of cost per speed. At the low end, there isn't much price difference per raise in speed, though at the top it jumps by huge percentages per minor increase.

You could do it really easily: plot the 1600+ through the 3000+ for current processor pricing with the X Axis showing speed and the Y Axis showing cost.

Here is some data:
1600+/1400/$56
1700/1466/53
1800/1533/65
1900/1600/89
2000/1666/79
2100/1733/94
2200/1800/118
2400/2000/147
2600/2080/247
2700/2166/330
2800/2250/395
3000/2166/590

These are OEM costs from Newegg (the 3000+ price is speculated from rumor). You make the call. If I can make that $53 1700+ run as fast as that $247 2600+ ... you get the picture, right? It's an easy sell to a customer if you sell OVERCLOCKING services along with computer building. They just need to be told the bad with the good to CYA.

02-10-2003, 10:57 AM	#1
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Overclocking effects on CPU longevity (Struggling to find the right Forum) Here's an interesting tid-bit of (unverified) information. In short, CPUs are spec'd to run for ten years, at stock voltages. (True? Anyone with contacts at Intel or AMD?). What we don't have is a measure of the effect of OC'ing... Personally, I don't believe that I'll be using the same PC for ten years, but I would take comfort in knowing that it'll run for "about" 5. __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

02-10-2003, 11:44 AM	#2
airspirit Been /.'d... have you? Join Date: Jul 2002 Location: Moscow, ID Posts: 1,986	IMHO, I don't mind if my CPU only lasts 3 years. Honestly, in the crowd that we're in, how many of us would be bothered if our CPU went out faster than that? By the time that $200 processor is showing age in a couple of years, you can slap a brand spankin' new one in for $30. For the amount of speed you gain over the life of both chips, you're coming out ahead. For certain people I build machines for, I give them this logic laid out. I ask them to order in importance the following three items: speed, price, and longevity. Once that is done, I explain that I can get them a 1700+ TBredB for $50, a 2100+ TBredB for around $100, or a 2700+ for around $300. With the first, I can get it to run slightly slower than the two more expensive chips, but it will most likely die in 2 years. With the other two, I don't have to push them as hard, so they'll probably last much longer, though in three years they'll be thinking about an upgrade anyway. Note: I ALWAYS explain to my customers the values and risks of overclocking their machines either by just a little or a whole lot. I have never had anyone refuse to have it buffed. Every once in a while I'll get someone who is leery, so as I'm building the machine and stress testing it at higher speeds I have them sit next to me and watch. Once they realize the relatively safety of a moderate overclock, even the most conservative people insist I leave it faster. Even at my business, every computer I've built for them is overclocked at least 20%. The price/performance advantage makes this the logical move. Needless to say, I've never had a single person go for a 2700+. Most go for the 1700+ knowing that even if it toasts in a year, they can buy better than the 2400+ to replace it and STILL come out ahead financially, and some go for the 2400+ because they don't want to be bothered with upgrading for a while. The amount of importance they put on longevity also affects how far I push the chip in regards to how far it'll go (I typically keep it at 50-80 Mhz below peak stable speed, though for longevity people I lower it another 100 Mhz to lower the voltage further). In all honesty, CPUs can almost be considered a disposable commodity anymore: you can OC a cheap one to the speed of an expensive one, and even if you fry three or four of them, you'll still be ahead pricewise. If you can buy a computer for $500 that performs like one for $1000, why would you care if in a year in a half you have to replace a $50 part or two? In the end you come out ahead. __________________ #!/bin/sh {who;} {last;} {pause;} {grep;} {touch;} {unzip;} mount /dev/girl -t {wet;} {fsck;} {fsck;} {fsck;} {fsck;} echo yes yes yes {yes;} umount {/dev/girl;zip;} rm -rf {wet.spot;} {sleep;} finger: permission denied

02-10-2003, 11:45 AM	#3
airspirit Been /.'d... have you? Join Date: Jul 2002 Location: Moscow, ID Posts: 1,986	On another note, a constant upgrade cycle gives us geeks something to do on the weekends besides drown in cheap beer. __________________ #!/bin/sh {who;} {last;} {pause;} {grep;} {touch;} {unzip;} mount /dev/girl -t {wet;} {fsck;} {fsck;} {fsck;} {fsck;} echo yes yes yes {yes;} umount {/dev/girl;zip;} rm -rf {wet.spot;} {sleep;} finger: permission denied

02-10-2003, 11:58 AM	#4
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Well, that's kinda my point. If you run a 1700+ to 2100+ specs, but you knew that you could expect a decreased longevity from 10 years to say 5 years, then it's worth it. but if the longevity is down to say 1 year, then maybe it's not so worth it... With your $ figures, assuming that the user will use his PC for 3 years, and it's overclocked to last 1 year, you'd be better off spending $100, rather than three times $50. But if the resulting longevity is more than 1.5 years, then it's even: that's the break-even point. Of course this doesn't take into account the cost of the cooling solution. What I'm trying to get to here, is to build a business case (cost justification) of OC and watercooling a PC. I think we all know that watercooling is mostly not worth the investment (IMHO), but it is a perfectly good hobby! Spending $300 on a kit is definitely out there. Does anyone have any data on the decreased longevity resulting from overvolting/overclocking/temperatures? (actually, I posted temp effects on longevity recently... now where did I put that up?) __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

02-10-2003, 12:10 PM	#5
airspirit Been /.'d... have you? Join Date: Jul 2002 Location: Moscow, ID Posts: 1,986	From what I've seen on forums (totally unscientific), on AMD processors, a overvolting of .3V reduces life expectancy to about 1.5 Years when temperatures average around 50C (2.05 on a palomino, 1.9 or thereabouts for TBreds). Lower temps from watercooling and phase change cooling tend to reduce that effect a bit. Two things you failed to account for in your example was the downward march of CPU prices per speed index, and the intangible cost benefit of having a faster machine immediately. The example should be: One 2400+ now for $200 OR One 1700+ now for $50 (at 2400+ speeds) and a 2700+ for $50 in 1.5 years. In the end, you're ahead in every category, unless the hassle of swapping chips is not worth the extra money, in which case you buy the highest speed you can afford, and leave it stock, though this will end up being a much more expensive solution. __________________ #!/bin/sh {who;} {last;} {pause;} {grep;} {touch;} {unzip;} mount /dev/girl -t {wet;} {fsck;} {fsck;} {fsck;} {fsck;} echo yes yes yes {yes;} umount {/dev/girl;zip;} rm -rf {wet.spot;} {sleep;} finger: permission denied

02-10-2003, 12:31 PM	#6
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Point taken! I'll add it to my business case __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

02-10-2003, 01:22 PM	#7
airspirit Been /.'d... have you? Join Date: Jul 2002 Location: Moscow, ID Posts: 1,986	Business case? __________________ #!/bin/sh {who;} {last;} {pause;} {grep;} {touch;} {unzip;} mount /dev/girl -t {wet;} {fsck;} {fsck;} {fsck;} {fsck;} echo yes yes yes {yes;} umount {/dev/girl;zip;} rm -rf {wet.spot;} {sleep;} finger: permission denied

02-10-2003, 01:36 PM	#8
pHaestus Big Player Making Big Money Join Date: Aug 2001 Location: irc.lostgeek.com #procooling.com Posts: 4,782	I abuse CPUs so I don't bother buying expensive ones I am running my 1700+ TBredB at 2.05V and 2400MHz now. I will continue to run it at such speed until it "doesn't go". I'll keep you posted. I have been overvolting the crap out of processors for a LONG time now, and never had a CPU die. I typically run them flat our til I want to upgrade, and then put them in systems for family members clocked back to whatever they will do near default voltage. I have yet to have the first CPU turn up dead: the Tbird 1.0 that ran at 1.625 and 2.1V, my old Celeron2 566@952 2V are both still purring along years later.

02-10-2003, 02:46 PM	#10
airspirit Been /.'d... have you? Join Date: Jul 2002 Location: Moscow, ID Posts: 1,986	Check two places: pricewatch and Newegg. Pricewatch gives you a general "historical" overview of both the newest and oldest chips, while Newegg gives you a look at verey reasonable prices on the correct steppings for the newest lowbies (like 1700+ TBredB chips). Those two sites will give you a peek at the exponential rise of cost per speed. At the low end, there isn't much price difference per raise in speed, though at the top it jumps by huge percentages per minor increase. You could do it really easily: plot the 1600+ through the 3000+ for current processor pricing with the X Axis showing speed and the Y Axis showing cost. Here is some data: 1600+/1400/$56 1700/1466/53 1800/1533/65 1900/1600/89 2000/1666/79 2100/1733/94 2200/1800/118 2400/2000/147 2600/2080/247 2700/2166/330 2800/2250/395 3000/2166/590 These are OEM costs from Newegg (the 3000+ price is speculated from rumor). You make the call. If I can make that $53 1700+ run as fast as that $247 2600+ ... you get the picture, right? It's an easy sell to a customer if you sell OVERCLOCKING services along with computer building. They just need to be told the bad with the good to CYA. __________________ #!/bin/sh {who;} {last;} {pause;} {grep;} {touch;} {unzip;} mount /dev/girl -t {wet;} {fsck;} {fsck;} {fsck;} {fsck;} echo yes yes yes {yes;} umount {/dev/girl;zip;} rm -rf {wet.spot;} {sleep;} finger: permission denied

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)