Why bother OC'ing?

[Jeff7181]

If you buy a CPU with the intention of overclocking it and think to yourself, "damn, now I gotta spend days finding the max overclock and ensure stability," then overclocking is not for you. When I get a new CPU I think "sweet, I get to spend a few days finding the max overclock and testing for stability." It also makes my gaming experience more enjoyable... a 10-20% increase in performance can mean the difference between being able to play a game at max detail settings with acceptable frame rates, or either putting up with poor frame rates some of the time or using medium detail settings.

Basically, if overclocking is a chore to you, don't do it.

 

[Amplifier]

I heard Overclocking doesn't work yo .

 

[Jeff7181]

Originally posted by: Amplifier
I heard Overclocking doesn't work yo .

My favorite is "sure you can overclock an XP2500 to the same speed as an XP3200 but it won't perform as well!"

 

[narcotic]

Originally posted by: Jeff7181

Originally posted by: Amplifier
I heard Overclocking doesn't work yo .

My favorite is "sure you can overclock an XP2500 to the same speed as an XP3200 but it won't perform as well!"

I know it seems funny, but technically its actually true...

*******

Originally posted by: cbehnken
Why ask stupid questions? Why does anyone buy a 3200 instead of a 3000? Why buy an x800 instead of a gf4mx?

I don't think I asked stupid questions (- but thanks anyway... ), I think you perhaps didn't understand my question, which was (putting it in your example): if you own A64 3000+, would you consider upgrading to A64 3200+....

Edit: corrected yet another error...

 

[DAPUNISHER]

Originally posted by: narcotic

Originally posted by: Jeff7181

Originally posted by: Amplifier
I heard Overclocking doesn't work yo .

My favorite is "sure you can overclock an XP2500 to the same speed as an XP3200 but it won't perform as well!"

I know it seems funny, but technically its actually true..

I await proof of this statement. If it looks, quacks, walks, and flies like a duck, it's a duck.

 

[cbehnken]

Originally posted by: narcotic

Originally posted by: Jeff7181

Originally posted by: Amplifier
I heard Overclocking doesn't work yo .

My favorite is "sure you can overclock an XP2500 to the same speed as an XP3200 but it won't perform as well!"

I know it seems funny, but technically its actually true...

*******

Originally posted by: cbehnken
Why ask stupid questions? Why does anyone buy a 3200 instead of a 3000? Why buy an x800 instead of a gf4mx?

I don't think I asked stupid questions (- but thanks anyway... ), I think you perhaps didn't understand my question, which was (putting it in your example): if you own A64 3000+, would you consider upgrading to A64 3200+....

Edit: corrected yet another error...

If the upgrade was free who wouldn't?

It's not like we sell our 3000+ or throw it away and then spend $190 on a 3200.

 

[Duvie]

Originally posted by: DAPUNISHER

Originally posted by: narcotic

Originally posted by: Jeff7181

Originally posted by: Amplifier
I heard Overclocking doesn't work yo .

My favorite is "sure you can overclock an XP2500 to the same speed as an XP3200 but it won't perform as well!"

I know it seems funny, but technically its actually true..

I await proof of this statement. If it looks, quacks, walks, and flies like a duck, it's a duck.

I agree...

Lesson #1

A Barton 2500+ is 1.83ghz with a 11x 166fsb.....

A Barton 3200+ is 2.2ghz with a 11x 200fsb...


As afr as I know they both have same SSE extensions and amount of L1 and L2 cache. the difference is the system bus only...

If you take the Barton 2500+ (like I and many did) and bumped the fsb to 200 and ran it with ddr400....hell-yes it is the exact same performance...What the heck is wrong with you??? There is no higher or lower bus, all components equal like ram and mobo I would say zippo difference...

I was lucky as I was able to do it with default vcore and settings and ran prime for 48 hours...gave it to a buddy and he runs it there to this day....as a 3200+ (bios detect it at bootup) with a heck of a savings when we bought it...

Now if you look at say my 3000+ versus a 3500+...

3000+ = 1.8ghz with a 9x 200fsb vs 3500+ = 2.2ghz with a 11x 200fsb....

Since I am top locked on multis I can only get to 2.2ghz with 9x244.....If I hold 1;1 I will be much faster. If I use a 166 divider and drop the HTT multi to 4x they are probably pretty close. If I use the 3500+ and drop its multi to 9x and set 244 fsb and run all the other same settngs it is identical.....

AXP is a bit simpler then the A64's....However if multi, system fsb, ram divder and settings are similar the systems should be identical....

 

[The Dude]

Bad thing about OC is the people that buy more than one of something and keep returning them after they beat on them hard trying to hit whatever promised land speeds. Then the vendor sells them to some other poor sucker.

 

[superkdogg]

Originally posted by: Concillian
To some extent I agree with the OP.

A serious challenge to overclockers:
- Un-overclock your processor for a week
- Play games as usual for three days or so (at the same settings you used before)
- Benchmark the games you play (at the same settings you used before)

Post up your results.

I have a 2800+ that overclocks in the 2350 range. I played for a week at 1800 MHz and this is what I found:
WoW -- No noticeable difference, FPS didn't appear to change, but I didn't use FRAPS
Battlefield: Vietnam -- No noticeable difference. FRAPS of a loop I use to benchmark showed ZERO difference in average and minimum FPS
UT2004 -- No noticeable difference. Benchmark showed ZERO difference in FPS between the two speeds.

Subjectively and objectively I couldn't tell a difference while gaming. This with a well overclocked x800pro, and I was running 1280 x 1024 with AA/AF enabled in UT and BFV, 1600 x 1200 only AF enabled in WoW.

Sure I can tell the difference when running specific benchmarks like benching games at 640x480 or SuperPi and the like, but don't fool yourself into thinking your gaming performance is actually getting better with a faster CPU. I have been an overclocker since 486 days. It's fun to find the limits, and I will continue to do so. But for everyday use, I back things off and run lower voltage to save on my utilities bill (I have 3 computers that are on 24/7 every little bit helps)

I disagree somewhat with the viewpoint of why OC, as there are cases where more CPU power is a good thing, I simply advocate thinking about how you actually benefit from OC'ing. I did that, and in my case I discovered that I wasn't getting any noticeable performance increase.

You're GPU limited-your CPU has little effect. If you want to figure out your CPU effect on gaming, set resolutions down to 640x480 and no eye candy. Granted, your contention is proven true that for somebody like you OC had no effect, but that doesn't mean that your CPU wasn't faster and better-it just means you didn't use the right tests.

I let my main rig run Einstein@home most of the time. The difference between 1800 MHz and 2700 MHz is nearly a linear increase in the amount of work that machine does. NTM that for my $150 I have a CPU that's at least equal and usually faster than the $550 4000+ and the $725 FX53. That's a pretty decent return on investment no matter how you look at it.

If I told you the speed limit on a road is 55, but you could go 80 and no police would pull you over, all you have to do is not get into an accident, how fast would you drive? Would it matter if you had plenty of time and didn't need to drive faster than 55?

For me, in the above scenario I'd go 80 everywhere, regardless of whether I was on time, early or late. It's kinda like OC'ing-why not take advantage of what's there since there's minimal risk if you do it well and correctly?

 

[Elfear]

I agree with the OP in that ocing doesn't make a whole lot of sense for people who don't notice a difference in the specific apps that they use (except maybe just to have fun). The time spent finding a stable oc can be fairly lengthy, especially if you are new at it. I think that the difference realized with ocing can be quite beneficial, even in modern games at a high resolution and max eye-candy. Going from 2.4GHz to 2.6GHz (8% increase) I saw gains of 5-8fps in HL2 at 1600x1200 4AA/8AF. Going from stock speeds of 1.8GHz to 2.5GHz (39% increase) should result in much larger gains and I run stock vcore, so theoretically, there shouldn't be a loss of cpu life either.

 

[Avalon]

Originally posted by: narcotic
well, I hear many of you are saying that a safe overclock will not shorten the lifespan of the cpu, but you are talking about overclocking in the range that I didn't feel much of a boost (aprox. 25%).
what happens when you REALLY overclock? Like Duvie... I mean, he probably feels great improvment, I read his comparison post, as most of you probably did, but the question is how healthy is it for the hardware to run at over 40% more than @stock?
Can you expect a system to run stable like this for years? (3 or more)

Well, you will only feel improvements when certain hardware designed to do something you are doing is the primary piece of hardware for accelerating that something. If you're overclocking your CPU and mem, even if you get very high of an overclock like Duvie or I, you still won't notice too much in a game unless you play at CPU bound settings. But, by then, you are already getting very high framerates, so even there the difference may be hard to feel. You may gain the ability to raise your AA or AF by a level, but that's about it. You'd want to overclock your GPU to get noticeable gains here.

Now, if you were running SuperPI or RARing a file, having a GPU overclock wouldn't mean a thing. However, that mem/CPU overclock would significantly reduce the time to complete either application.

You just have to pair your overclock with practical applications that belong to said hardware pieces that you are overclocking.

 

[Concillian]

Originally posted by: geforcetony

Originally posted by: Concillian
My main point was the observation that many people OC for the sake of OC'ing and little else when you look at the actual gain. For some this is a hobby and that's great, but I think there are people who are under the impression that they are improving gaming performance and I think these people are not under the correct impression.

Thats a good point. However, IMHO, if people are going to overclock, they are going to do it because they DO realize the benefit of overclocking. You're most likely not going to get some dude that bought an E-Machines computer overclocking just because he can. No. Most, if not all, people overclock to squeeze every last ounce of performance they can get to essentially "keep up" with the people that have $5000 machines. Not saying that there aren't those who do it just to do it, as well as those who do it just for bragging rights. I just think that people who overclock realize the amount of benefit they will get when doing it.

I dsagree. It seems like a number of people I see in normal gaming forums think they are benefitting from overclocking their processor when running at video settings that would clearly create a bottleneck with the video card. Or people who argue over framerates that are already higher than the refresh rate of the monitor.

I see a lot of hype out there, and I think a great many people get caught up in it.

Case in point is how many gamers in these very forums thought much of value memory before Zebo did his matrix? He and I were both advocating for months before that post the excellent price/performance ratio with value memory and contested that you do not lose much performance at all in applications people actually use.

I was constantly seeing posts recommending TCCD and Ballistix to some guy buying a midrange gaming system and a midrange video card. People started realizing eventually that life is better saving $100-150 on memory and dumping that into the video card.

I am not saying that the whole CPU OC world is dumb and nobody will ever see benefit from OCing CPUs, I am merely saying that in most cases gamers who OC their CPU will not see tangible benefit from it.

 

[DKlein]

I actually barely noticed any difference at all in Windows load times with overclocking. Heck, even at 1GHz Windows booted up approximately 85% as fast as with 2.46GHz. Maybe that's just me, though.

For games, I think it will depend, and I'm going to test this later with AA and R:TW. For older games, where you're not really CPU-limited, I'd doubt it will make a difference, but what about in newer games? If I have a 2600+ that runs around 2GHz stock, maybe that will be more than enough for a game like HL or AA, but what about when STALKER comes out? What about new RTS games? These games will definately be more demanding on the processor, and you'll start to see your once GPU-limited computer becoming CPU-limited. THEN you'll be able to see the difference between 2GHz and 2.46GHz, and THEN overclocking will be worth it, as you'll be saving yourself a lot of money having to buy a whole new processor (and motherboard) to get equivalent performance. I'm thinking I'm already being limited on R:TW, as it's only playing around 25-30FPS, and that's with a 6800Ultra equivalent card (well, thereabouts at least). Anyway, this is all conjecture without empirical evidence, so I'm going to go give R:TW a few runs through to see the difference.

 

[narcotic]

Answering Duvie's last post (I didn't post a quote of it, since its very long).

Basickly you're right, an overclocked axp 2500+ may perform as well as a non overclocked axp 3200+... however you're wrong about them being the same thing.
Although most people won't notice it, an overclocked cpu will generate much more calculation errors than a non-oc'ed cpu (namely floating point calculations), as well as generate more heat etc.
So, as I said, though performance might (not going to debate this issue, without having benchmarks to support it) be the same, still technically the two are different.

 

[bunnyfubbles]

I can often see some tangible differences, some can be found here on AT as well.

http://anandtech.com/cpuchipsets/showdoc.aspx?i=2330&p=2

Although I'd have to agree with you in some respects, as the difference isn't enough to justify forking over the money for the faster CPU, but with overclocking I can get that perforamnce without the loss in money. You'll definately not be dissapointed with 1.8GHz A64 gaming performance, but 2.4-2.6GHz is all the better and the difference will be all the more apparent with newer games.

 

[Concillian]

Originally posted by: superkdogg
You're GPU limited-your CPU has little effect. If you want to figure out your CPU effect on gaming, set resolutions down to 640x480 and no eye candy. Granted, your contention is proven true that for somebody like you OC had no effect, but that doesn't mean that your CPU wasn't faster and better-it just means you didn't use the right tests.

Of course you can see a difference if you remove other important factors. I'm glad you see my true point though (performance improvement in the applications I use the way I use them)

I let my main rig run Einstein@home most of the time. The difference between 1800 MHz and 2700 MHz is nearly a linear increase in the amount of work that machine does. NTM that for my $150 I have a CPU that's at least equal and usually faster than the $550 4000+ and the $725 FX53. That's a pretty decent return on investment no matter how you look at it.

You are very adept at figuring costs, have you figured the cost of the additional electricity consumed by running at 2700 MHz vs. 1800 MHz. In my case with three machines on 24/7 and folding I project I will save over $100 a year in reducing my overclocks from insane voltages to something more reasonable (undervolting some and still overclocking) based on the differences I saw from overclocking. I'm willing to donate my spare cycles, but I'm not going to create extra cycles just to donate them. That's akin to going to the store and buying clothes expressly for the purpose of giving them to Goodwill. Only rich people do that, and I'm not rich. At least in that case you would get a partial tax break on the money spent.

If I told you the speed limit on a road is 55, but you could go 80 and no police would pull you over, all you have to do is not get into an accident, how fast would you drive? Would it matter if you had plenty of time and didn't need to drive faster than 55?

This is an interesting analogy, and my brother was able to experience the answer to this question when he lived in Italy, where the police are known to ignore speeders. He drove a lot as he enjoyed going different places and seeing the sights. He would get visitors and show them Florence and Rome, etc... The Naples to Florence drive was one he did often. Eventually he found an optimum speed to drive. Such that it took very close to exactly two tanks of gas to drive between the two the result was around 80-85 MPH or so. The other option was to drive around 115 or so, which resulted in almost the same amount of time, but THREE gas stops along the way.

So the answer to your question is "I would drive as fast as I could as long as the benefits of my speed were outweighing the negatives of my gas consumption". The world is not one sided issues, everything has a trade-off, and you constantly have to wiegh advantages over disadvantages.

Some people say I tend towards over-analysis of situations. Maybe they're right?

 

[Concillian]

Originally posted by: bunnyfubbles
I can often see some tangible differences, some can be found here on AT as well.

http://anandtech.com/cpuchipsets/showdoc.aspx?i=2330&p=2

I know you agree with my main point, based on your other comments, however I wanted to point out a couple of things I think when I see a page like that:

1) In all but one of those cases the difference is not noticeable because the FPS of ALL CPUs is greater than my refresh rate of 85Hz. In the case of a 3000+, the refresh rate is over 100FPS in all but one case

2) If I'm getting 100+ FPS I'm not going to play that game at those video settings. I'm going to bump the res and AA/AF level, which will bring the GPU performance down close to or below the CPU limit.

3) In the most CPU limited case I'm at ~65 FPS average, however that doesn't tell me a whole lot. I personally know that my sensitivity in multiplayer FPS games is where the minimum FPS approaches 40 FPS. In many cases an average of 65 FPS in a single player FPS will probably be adequate, though were it multiplayer, I may want a slight boost.

I just wanted to illustrate my thought process. There are times where a 20% difference equals a 0% performance improvement, I try to find those cases and look at the data for what practical information I can get out of it.

 

[DKlein]

Okay, here are the results of my mini-experiment:
The game: Rome:Total War, Custom Battle, 2 huge armies
The situation: full rush on middle of enemy army at onset of battle, camera positioned over same area. Lasts approximately 1.5 min until army retreats.
Graphics: no aa/af, 800x600

Test 1:
Overclocked CPU 214x11.5 = 2.461GHz
Using FRAPS, average FPS = 15.029, min=6, max=30

Test 2:
CPU 214x9.5 = 2.033GHz
Using FRAPS, average FPS = 11.837, min=5, max=28

The differences are plain during the actual fighting, after the first 10 seconds of the battle. In this time, the average FPS is about 9 in the overclocked test, but 6 in the stock clocked test. So, we see a difference actually greater than the overclocking between the two tests: an overclock of approximately 21% has yielded an average FPS difference of near 26% overall, and around 33% during CPU-intensive parts. Of course, this is a very limited sampling, but it goes to show that overclocking a CPU can show a difference.

 

[Gamingphreek]

What a lot of you seem to be forgetting is we OC because we enjoy it. Most people who post here are people who enjoy working on computers. Of course the performance gains are always nice but we enjoy it, simple enough.

What does it matter if you dont get a huge performance difference, you are still getting more than you pay for.

Finally, you say it greatly shortens the life of the PC. While it most likely does shorten the life i have had 0 problems and i have had my PC OCed for 2 years.

-Kevin

 

[xtknight]

yeah...nothing we didn't already know really. i agree gpu's are better for overclocking. for people like me who make levels for games and have to compile bsp,visibility,radiance,etc... which took me 25 mins...we will notice a difference. maybe you won't feel a difference in windows but the performance will be better for applications that need it. of course you won't feel the gain in internet explorer or windows because it's already enough. apps like those are limited by other factors such as internet speed or the hard drive, not the cpu.

 

[DKlein]

Well, further testing has supported my previous results, with a closer %change between the overclock and avgFPS, though. Also, keep in mind that this is retaining the overclocking of the RAM and FSB, since I wanted to keep it purely CPU-focused.

 

[xtknight]

Originally posted by: RussianSensation
Here is another way to think about it. If you can get more performance for free why not?

Also, I bought P4 2.6 and overclocked to 3.2ghz. My last processor was 1600+. Thus, i doubled my speed. I agree with you that even 25% increase in performance is not noticeable (at least for me from 2.6 to 3.2ghz); but 100% was. So next time I upgrade I'll make sure I get somewhere along 6000+ rating (or I can get 5500+ and overclock there). Therefore, another benefit of overclocking is waiting longer before wasting more money on computer upgrades.

Finally, a lot of people feel better getting something for $150 and overclocking it to perform at FX55 $700US price.

EDIT: Also consider videocards. Say you get 6800Gt and overclock it to 420/1140. Now the argument is that you can just get 6800Ultra nad overclock that too. Problem is it'll only get to like 450/1200. The overclocking on 6800GT saves you a pile of money and will most likely get you pretty close to Ultra's maximum overclocks as well. So why pay $100 more when 2 cards will perform nearly identical when overclocked?

the thing is, overclocking is not free once you damage your hardware. safety precautions can greatly reduce this, but still...when you overclock your cpu too far and it overheats, nothing is free.

 

[Duvie]

Originally posted by: narcotic
Answering Duvie's last post (I didn't post a quote of it, since its very long).

Basickly you're right, an overclocked axp 2500+ may perform as well as a non overclocked axp 3200+... however you're wrong about them being the same thing.
Although most people won't notice it, an overclocked cpu will generate much more calculation errors than a non-oc'ed cpu (namely floating point calculations), as well as generate more heat etc.
So, as I said, though performance might (not going to debate this issue, without having benchmarks to support it) be the same, still technically the two are different.

How do you know it is erroring at all??? How do you know that the 2500+ barton would not have been qualified to be an 3200+xp by AMD's certification. According to another great debate I believe cpu are often speed binned based on market needs. At the time the 3200+ was the flagship of the AMD line but at a hefty price. The 2500+ was a screaming deal and I imagined they sold a lot more of them. However basic philosphy is and was that both intel and AMD get really good yields out of the chips and AMD ws no different on their much refined and matured 130nm process. The fact is if you look at ocing numbers and where ppl top out (like you see in winnies and even p4s) you see that from 2500+'s to 3200's and evrything in the middle there were numbers of each cpu hitting relatively same highs. That says that as a piece of silicon they are very similar. In all likelihood the 2500+ was a 3200+ calibur chip that was marked down (with the fsb) to meet the more demand for a more bargain range 100-150 dollar cpu, to work with still widely available 333fsb board and PC2700 ram. they still needed to feed that sector of the amd chipsets.


If I run a 2500+ oc'd to 3200+ at default vcore (even on a board that was undervolting by .03-.04) and I was able to run prime for 24 hours I would think your "errors" were not occuring...your calculation errors are going to come up as all sorts of errors in programs that will just halt or error out.

I am saying that a 2500+ that can do what I did above (run at 2.2ghz defualt vcore and primed) running against a certified 3200+ on same board with all setting and components will be and should be identical in performance.

The biggest issue on heat may only be a result that some of the 2500+'s came with a wimpier stock hsf,. however at default vcore each chip at identical 2.2ghz should produce the same amounf of watts of heat. basic stuff here buddy!!! If the HSF was not the same then thermal conduction could have been worse and showed higher temps at same speed. However again if those temps are below AMD max temp specifications there should be no "calculation" errors. I however like many were lucky and received a HSF that was identical to the ones being packaged with the 3200+'s so my temps were right in line with all the true 3200+...as true as diff mobos may report.


I doubt most will have 2 of the same systems with each chip but commonsense, physics, etc should be able to be clear about this...I guess not for you though.....

I see why you are asking this...question about ocing.

Another thing to consider is that a friend of mine who used to hang on these boards had a 3200+ and was only able to Oc it 150mhz before he started to need to give it more vcore. That 2500+ I had was able from my testing to do 2.3ghz at prime 24 hours. We left it at 2.2ghz only becasue we didn't want to OC the ram and FSB anymore then its rated speed. So, the certfied 3200+ was only able to clock 150mhz more, which is not much headroom and the 2500+ was able to clock 800mhzmore ??? That should be like a bat hitting you upside the head....Many chips are capable of being certified at higher speeds so there is no guarantees these 2500+'s were calculating with any errors and thus resulting is lesser performance versus a certified 3200+...there is a reason the Barton line stopped at 3200+...heat was becoming a factor already which is very likely what was starting to hold the headroom back on the chips. Later on refined barton mobiles have really pushed up there to speeds of 2.6ghz but with some vcore.

 

[Aenslead]

I don't usually game. I have a Radeon 9800SE because I heard that the 9xxx series had excelent 2D quality, contrary to the 5xxx series of nVidia. Be as it may, I do run, out of curiosity, some benchmarks;

On stock clock, 3D Mark 2k5, I had 1103 points
Now, overclocked 30%, I get 1,380 points.

Not bad, huh? I guess it DOES affect gaming.

 

[Avalon]

Originally posted by: Aenslead
I don't usually game. I have a Radeon 9800SE because I heard that the 9xxx series had excelent 2D quality, contrary to the 5xxx series of nVidia. Be as it may, I do run, out of curiosity, some benchmarks;

On stock clock, 3D Mark 2k5, I had 1103 points
Now, overclocked 30%, I get 1,380 points.

Not bad, huh? I guess it DOES affect gaming.

You need to bench some real games now