Anti-Overclocker's thoughts

[LoserXLeet]

Hi guys! I recently talked to someone who was very against overclocking. I'm going to post what he had to say, because I frankly have no idea how to overclock (as much as I wish I did, I just don't understand the guides), and would like someone to interpret his thoughts, and give counter-points, if any exist. Here you go...


- the disadvantages and reduced lifespan of oc'ing anything far outweigh the advantages
- no matter how much you oc shit you still cant it thru the bus faster than 500mhz
- the on chip processing may be faster, but the throughput back to the components on the bus are still the largest bottleneck of any x86 architecture
- I find best performance on servers with the new dual core intels with 2mb or better L2 cache
- A 1mb L2 cache 2.8 is much faster than a 512k 3.2 on pure throughput
- Dont bother with oc'ing..its not worth it in the long run, and 99% of the time your increases that you'll actually feel are minimal at best
- You are also risking premature failure too
- oc'ing reduces parts life in some case to 1/2 of normal expected lifespan
- I guess what I dont understand about oc'ing is that no matter what you do, you cant increase the speed that the ram runs thru the board
- You can increase the speed of the ram, chip to chip on the ram.. but NOT on the pipeline to the board
- What you end up with a bunch of data queued (tagged) waiting to cycle back to the proc
- Sure you can overclock ram.. the ram itself.. not the pipeline thru to the core on the board
- What does it really net you? if you cant change the speed that the ram data is put to the pipeline on the proc.. why bother oc'ing?
- You can oc the proc from 3.0ghz to 4.0ghz.. you can increase the ram from 533 to 700...
- Out you cant increase the ram communication speed with the pipeline to the proc/board
- so why bother?
- You could make the ram run at a jigahertz.. but thats at the ram.. NOT the speed it communicates with the pipeline at
- I've been doing this shit for 15 yrs.. i know about oc'ing.. and I also know its not worth the 10% actual increase you get
- Doubling ram speed, doubling even processor speed does not equal double the performance
- You can overclock certain components yes.. but you cant overclock the speed at which all those devices communicate with each other on the board

 

[NEODARK]

Not to crap on your thread, but LOL...

Many of "his" points dont make any sense.

 

[haste]

ummm, that dude needs to come to this forum and see realworld results from o/cing.
maybe then he would understand.


and oh, LOL

 

[BlackTigers91]

Errr......all that "pipeline" talk:

Correct me if i'm wrong, but aren't those just where the data travels, and have no effect on how quick the processor is?

Sheesh.

 

[Firebat]

Err.........I have the feeling he doesnt really know what he is talking about.

 

[Motley]

OK, so he doesn't like to overclock, so well, leave us alone...becuase we do!!

 

[DaLurker]

I think your friend has some serious misconceptions about CPU speed, throughput, ram speed, PCI, AGP, PCI-E, Bus speeds.
I'll try to dispel these misconceptions.

- the disadvantages and reduced lifespan of oc'ing anything far outweigh the advantages
- You are also risking premature failure too
- oc'ing reduces parts life in some case to 1/2 of normal expected lifespan

Firstly, overclock does reduce lifespan and risk premature failure however note that experienced overclockers know these risks and any/all of them would recommend you DO NOT RUN and overclocked system for anything that is mission critical. The mentality as to why this is ok though is as follows: Overclockers tend to be enthusiasts. They change hardware often enough that the reduced lifespan of the CPU is not an issue. Intel and AMD generally have a MTF time of 10 years. Lets say it reduces lifespan by 50% which is generous, the CPU will still last 5 years. Most enthusiasts will have changed CPU's in 5 years. Quite simply, how many people here use Pentium III's still? Well, that was 5 years ago.

- Dont bother with oc'ing..its not worth it in the long run, and 99% of the time your increases that you'll actually feel are minimal at best
- I've been doing this shit for 15 yrs.. i know about oc'ing.. and I also know its not worth the 10% actual increase you get

The increase in overclocking can easily be measured by taking a faster CPU of the same architecture and benchmarking. Simply put, is an Athlon 3200+ faster than a Athlon 3000+? Yes it is, its 2.0ghz vs. 1.8ghz. Will you feel the difference? Well that depends on the user. The above statements basically say there is no point buying an Athlon 3200+ because it is ONLY 10% increase, and what you'll actually feel is minimal. If your friend truly believes this then he should be running the slowest processor ALWAYS since he doesn't believe the increase in clock speed will net you a difference.

Overclocking isn't voodoo, it isn't magic. it is increasing the clock speed of the CPU over what AMD determined is stable. So when I overclock a A64 1.8ghz to 2.0ghz, it is essentially the same as a 2.0ghz CPU from AMD. There is NO difference.

- What you end up with a bunch of data queued (tagged) waiting to cycle back to the proc
- no matter how much you oc shit you still cant it thru the bus faster than 500mhz
- the on chip processing may be faster, but the throughput back to the components on the bus are still the largest bottleneck of any x86 architecture
- Sure you can overclock ram.. the ram itself.. not the pipeline thru to the core on the board
- What does it really net you? if you cant change the speed that the ram data is put to the pipeline on the proc.. why bother oc'ing?
- You can oc the proc from 3.0ghz to 4.0ghz.. you can increase the ram from 533 to 700...
- Out you cant increase the ram communication speed with the pipeline to the proc/board
- You could make the ram run at a jigahertz.. but thats at the ram.. NOT the speed it communicates with the pipeline at
- Doubling ram speed, doubling even processor speed does not equal double the performance
- You can overclock certain components yes.. but you cant overclock the speed at which all those devices communicate with each other on the board
- You can increase the speed of the ram, chip to chip on the ram.. but NOT on the pipeline to the board
- I guess what I dont understand about oc'ing is that no matter what you do, you cant increase the speed that the ram runs thru the board

A little engineering background will dispel the above statements which don't make sense. The fact is the pipelines are the LEAST limiting factors, and the components are the MOST.

The CPU communicates through a Bus to all the components. The bus speeds depend on the arcthitecture. PCI bus communicates at 33mhz. AGP bus communicates at 66mhz, PCI-E bus communicates at 100mhz. The Ram bus (aka FSB, or HTT) depends on the CPU, but on all Athlon 64 platforms, the default is 200mhz. This is the speed at which the CPU communicates with the Ram and also the speed at the Ram. There is no distinction at what speed the Ram runs at, and what speed the ram communicates with the system bus at. They're the same.

Now its true there are bottlenecks everywhere along the communication buses of a computer system. The largest bottleneck is actually your hard drive. Then you have your Ram and finally your cache. The only cache that works at CPU speed is L1 cache, usually minute (8kb, 16kb). L2 cache is slightly slower but comes in larger sizes (512kb, 1mb, 2mb). Then theres Ram which can come in Gb's, and hard drives which comes in tens and hundreds of GB. Its true that if Ram were to communicate at CPU speeds, the whole system would be a lot faster but RAM can't run at CPU speeds not because of the bus, but because of the RAM. The fact however is increasing an A64's 1.8ghz to 2.8ghz (3000+ to FX56) without changing the Ram speed does increase your performance.

You can overclock the components, and the devices, however almost no device we have oversaturates the bus on which they sit. Case in point graphics cards used to be on the PCI bus. Then we found that GPU's are taking too much bandwidth so we created the AGP bus. Now we find the AGP bus may start to have limitations so we have the PCI-E bus. GPU do not make use of the bandwidth available, so overclocking the video cards will ALWAYS net you a performance gain.

Ram is the same story. The System pipeline on NF4's can usually scale up to way over 300mhz. In fact on the new DFI NF4 Lanparty UT SLI-DR Expert, they've gotten the pipe up to 512mhz. The fact is ram can't run that fast. They start getting errors.

So the conclusion is the systems LEAST limiting factor is the pipeline. The MOST limiting factors are the components.

- A 1mb L2 cache 2.8 is much faster than a 512k 3.2 on pure throughput
- I find best performance on servers with the new dual core intels with 2mb or better L2 cache

Lastly just to answer the L2 Cache myth here. More L2 cache is always beneficial, however the performance increase of more L2 ram depends solely on the CPU architecture. On Pentium 4's, the larger the L2 cache the better because of the long processing pipeline it has. Whenever there is a pipeline crunch, it has to go back and start again and having more instructions on the L2 cache is beneficial. The A64's on the other hand do not need such a large Cache because the pipeline is shorter. Its the same story as the P3's, and the Pentium M's.

The best performance on servers is the AMD Opteron and they sport 1mb cache and actually run a slower clock speed than the Xeons or P4s.


So in conclusion, overclocking does increase performance, just like buying a faster CPU increases performance. If you truly believe overclocking doesn't net any increase, then you should never spend more on a faster CPU, but just buy the slower clocked CPU. Your friend despite being in the business for 15 years has many misconceptions. RAM is the limiting factor, not the Pipeline. Throughput is deteremined by how fast your components can run, not your bus because the buses nowadays are not saturated. More cache is better, but CPU architecture determines how much cache is needed and in turn how much performance increase there is when adding more cache.

Overclocking = Faster CPU = Better performance. Always.

 

[Xilikon]

Without considering that many of the points is redundant

 

[SJetski71]

Edit: NM,

DaLurker answered it better than i ever could

 

[arentol]

- the disadvantages and reduced lifespan of oc'ing anything far outweigh the advantages
- no matter how much you oc shit you still cant it thru the bus faster than 500mhz
- the on chip processing may be faster, but the throughput back to the components on the bus are still the largest bottleneck of any x86 architecture

This talk of the bottleneck being the throughput to other components doesn't make any sense. People have already OC'ed the fastest PC processors in the world to levels as much as 50% higher than their original speed and then proven it makes their PC's perform much better So if you have a 2.8ghz P4 and people have already proven a 3.4ghz P4 OC'ed to 4.2ghz can't oversaturate the bus (99% of the time anyway) then there is NO way any OC you throw at your CPU is going to oversaturate anything.

- I find best performance on servers with the new dual core intels with 2mb or better L2 cache
- A 1mb L2 cache 2.8 is much faster than a 512k 3.2 on pure throughput

Obviously dual-core will outperform single core, and more cache is obviously better at the same clockspeed. But the 2.8 vs. 3.2mhz thing? Not true. There may be a FEW applications that rely very heavily on the L2 Cache, and those will be faster on the 2.8, but for the other 90% of programs out there the pure high clockspeed of the 3.2 will cause it to be faster.

- Dont bother with oc'ing..its not worth it in the long run, and 99% of the time your increases that you'll actually feel are minimal at best
- You are also risking premature failure too
- oc'ing reduces parts life in some case to 1/2 of normal expected lifespan

The first statement is basically right, a 10%, or even 20%, increase in CPU speed won't provide a significantly noticeable difference most of the time. In some programs it will though, especially clock-cycle based stuff. You are risking premature failure, but mostly in the initial testing phases, and so much once you have a solid, and reasonably cool, OC going. Also, that is why learning about others success with various parts before you buy them is important. As to OCing reducing lifespan by 50%, or at all, it totally depends on the amount of the OC and the parts you are using. However, it doesn't really matter anyway because the "lifespan" of most modern PC parts is so long that no OC'er has ever reached 50% of it anyway! Reducing the lifespan from 12 years to 6 years on a PC you will only have for 3 or 4 years is hardly a problem. In addition, enthusiast parts for PC's these days are DESIGNED to run for their full lifespan at OC'ed speeds, and so this is a non-issue for a serious OC'er. In addition, some CPU's are "Downbinned" and so OC'ing them actually just brings them back up to their "normal" speed, and thus they still have normal lifespans at these speeds.

- I guess what I dont understand about oc'ing is that no matter what you do, you cant increase the speed that the ram runs thru the board
- You can increase the speed of the ram, chip to chip on the ram.. but NOT on the pipeline to the board
- What you end up with a bunch of data queued (tagged) waiting to cycle back to the proc
- Sure you can overclock ram.. the ram itself.. not the pipeline thru to the core on the board
- What does it really net you? if you cant change the speed that the ram data is put to the pipeline on the proc.. why bother oc'ing?
- You can oc the proc from 3.0ghz to 4.0ghz.. you can increase the ram from 533 to 700...
- Out you cant increase the ram communication speed with the pipeline to the proc/board
- so why bother?
- You could make the ram run at a jigahertz.. but thats at the ram.. NOT the speed it communicates with the pipeline at
- I've been doing this shit for 15 yrs.. i know about oc'ing.. and I also know its not worth the 10% actual increase you get
- Doubling ram speed, doubling even processor speed does not equal double the performance
- You can overclock certain components yes.. but you cant overclock the speed at which all those devices communicate with each other on the board

Again, all this comes back to the same thing I said before. All of it only applies IF these supposed motherboard bottelnecks are already saturated at the speed you are running and with ALL your programs, and we already know that very high CPU speeds still equate to improved performance most of the time. And, of course it doesn't apply in ANY way to anyone buying a cheaper CPU and then OC'ing it to higher priced CPU speeds because we already know that at those speeds there is a proven performance increase. In addition, it still doesn't apply to people OC'ing high-end CPU's because MOST programs won't saturate the bus anyway.


One question though.. Is this all in reference to a business SERVER, or to a home PC? Because I did notice you used the term "server" and with a server you want to focus on uptime and reliability, not pure speed. So if this is about a server I would suggest not going beyond a very mild OC, 10% or less would be best. In addition, depending on what the server is doing the bottleneck could be elsewhere, like harddrive speed on a busy file server, so your friend could be correct as his overall opinion applies to YOUR circumstance, though he certainly isn't correct in general.

 

[LoserXLeet]

Actually I started the conversation with him because I wanted to know if he had previous experience overclocking. I want to overclock an opteron on a personal computer (preferably a 165, but I have a feeling a 144 will have to do, considering my budget. By the way, how well do those overclock now a days (what CPU is equivalent to an average overclocked opteron 144/165?) )

 

[malicious]

opties overclock well .. some more than others .. find a good stepping like cabne cabye caybe etc... also the week 47s seem to be doing very well.

A guy has one for sale here that does 3ghz on air..

to put that into perspective an fx-57 is stock at 2.8 and also cost like a grand wheres you can get a 3ghz cabale opty for 200-250.

as far as the OP goes .. who gives a shit .. i overclock just for the fun of overclocking.. I don't care if i get 1 more fps (which i of coarse do).

mal

 

[qbert3]

to some of us that extra framerate matters.

 

[Darkala]

this guy is an asshole =] he says the same thing several times to make it look like he knows what hes doing. ive heard those points from others, and my idea is im not going to use the same hardware more than a year, so if it burns out in 2 and wont last 5 or whatever he thinks it should, who cares?

 

[Dillusion]

To tell you the truth, you could view stock processors as 'overclocked'. Lets say AMD and intel only made two chips, a 1.8 and a 2.8 respectively. Now any and every chip that has a higher speed than these 'stock speeds' is 'overclocked' from the factory. Does that help put it into perspective? When your friend says 'only a 10% increase isnt worth it', thats very not true. Each 10% you overclock- thats a new chip from each company that they 'overlocked' themselves. So elts say you get a 40% overclock? Thats getting 4 models higher than you paid for, for doing what the factory does before they ship it to you.

 

[elm669]

Err.........I have the feeling he doesnt really know what he is talking about.

Agreed

1.6ghz vs 2.5ghz/ 225/400 vs 500/430.. yea overclocking doesn't do much..

 

[superkdogg]

This guy's friend needs to talk to my wallet about overclocking.

Winnie 3000+> $150
Motherboard of your choice >$100
GTO2 or GS, or whatever you like >$200
Some decent ram >$80
Good psu >$80

For around $500 my system is still plenty powerful because of OC'ing but if I ran it stock it would be pretty pedestrian. If I feel like it I'll start a new thread stock vs. OC'd and I won't even flash back to 12 pipes on the GPU-just "under" clock it.

 

[Justintoxicated]

This guy is so right, I notice no difference between my Opteron 165 at stock speed 1.8 ghz and at Overclocked speeds 2.7 ghz...

After all there is no noticable difference between an opteron 165 and AMD FX 60 right? So they basicaly sell FX60 just for rich people who like the name of the cpu to be FX...

I'm pretty sure he is right here, and reducing the lifespan of my CPU is also a horrible thing, thats why my 9 or 10 year old celleron 300a is still plugging away @ 464 mhz. I would be terribly sad if it died, I would have to replace it with something that is actualy USEFULL today!

BTW overclocking does not Necessarily reduce lifespan, only if you increase voltages!

he is WAY off on the memory pipeline, WTH do you think you increase FSB and HTT for, the gains from doing this are noticable, although typicaly since multipliers are now locked we do this to make the CPU faster more than anything.

All the devices communicating with the board? Huh? Most PCI devices don't even utilize the max PCI speeds, so there is no purpose to overclock them, your sound card won't produce better qualit sound if you overlock the buss it uses.

According to this guy...We should not Buy any components that are not faster than Hard Drives, Since the Hard Drive is the slowest device on the system, then why not run a 300 MHZ CPU! I mean it's all going to bottleneck on the HD anyways right?

 

[Drudenhaus]

And I'm sure he buys FX-55s/FX-57s for $125. Well thats what my Opty 144 cost and it does just under FX-57 speeds (which is $811).

 

[Justintoxicated]

yea well some people like their Cars / Trucks / ATV's / Dirt Bikes stock, and others add Superchargers , Turbos, Exhaust, and Intakes...

Same concept only for PC's it is closer to free.

 

[Retsam]

It pretty much is free. Think about it. Most opty 165's these days do 2.6ghz at the very least, which is the speed of an FX60 which costs $900-$1000.... Ive seen many opty 165's sell for $240-$280 used lately. New for like $280 as well. The speed of a $1000 cpu for $280 + price of a decent heatsink = Pretty much FREE? Cant beat that.

 

[Justintoxicated]

well I added the cost of the Liquid cooling, thats why I said pretty much free, but hey liquid and fast for less than FX

 

[Retsam]

Anyone who is agaisnt overclocking probably just lacks experience or maybe just to timid with their computer parts? Im not sure. The only argument I could see agaisnt overclocking is arguing it would decrease the life span and/or void the warranty. But whats that? Like now its only going to last for 15 years where as it would have lasted for 20? Heh.

Most people upgrade every few years anyways.

 

[kunsunoke]

- the disadvantages and reduced lifespan of oc'ing anything far outweigh the advantages

Every cpu core has a maximum speed configuration with stability, and most are overbuilt to survive even stupid builds.

Your bud can get his FX-57 or some Prescott core crapbuild from Dell for his assload of money. I can overclock a Venus or Denmark core and get better performance for far less than what he paid, and lack for nothing in terms of reliability.

- no matter how much you oc shit you still cant it thru the bus faster than 500mhz

Funny, my Venus & Venice core AMD CPUs are now doing 600 FSB easily. And they fold 24/7. I've seen some at 800 FSB on DFI boards. Your friend needs to put his rock pipe down.

- the on chip processing may be faster, but the throughput back to the components on the bus are still the largest bottleneck of any x86 architecture

LOL - as opposed to off-chip processing. And again he gives up the "fixed" front-side bus concept. Nothing could be less fixed with the advent of the PCI-E lock.

- I find best performance on servers with the new dual core intels with 2mb or better L2 cache

Uh huh. Yeah, they're faster than dual-core Opterons. No matter that they still deal with an antiquated FSB and an off-die memory controller.

- A 1mb L2 cache 2.8 is much faster than a 512k 3.2 on pure throughput

In whose crack fantasy? There aren't that many programs that take advantage of the huge L2 cache. Cache is over-rated.

- Dont bother with oc'ing..its not worth it in the long run, and 99% of the time your increases that you'll actually feel are minimal at best

As mentioned previously, there are significant gains to be had in both CPU speed and memory bandwidth, not to mention graphic processing, sound processing, etc.

- You are also risking premature failure too

So what? We risk premature failure even on a stock machine when we type during a thunderstorm without surge protection.

And if a $130 processor burns up, so what? We can buy another. We're still way ahead from a financial standpoint compared to the P4 Extreme Edition guy that dropped $1000++ on his chunk of silicon.

- oc'ing reduces parts life in some case to 1/2 of normal expected lifespan

Based on what evidence? Even if true, so what? Does he think most of us carry the processors past a year or two?

- I guess what I dont understand about oc'ing is that no matter what you do, you cant increase the speed that the ram runs thru the board

ROTF - so we can't run FSB faster than stock, huh? Show of hands, please?

- You can increase the speed of the ram, chip to chip on the ram.. but NOT on the pipeline to the board

Still living in the days of the Pentium 1. Tsk. As mentioned previously, FSB is adjustable, as is memory bandwidth. It was even in the Pentium 1 days.

- What you end up with a bunch of data queued (tagged) waiting to cycle back to the proc

Where does the "tagged" data queue up, and what precisely is the bottleneck? The FSB? We've established that FSB is adjustable.

- Sure you can overclock ram.. the ram itself.. not the pipeline thru to the core on the board

Um, if you overclock the ram you increase its operating frequency, which increases throughput to the board. If FSB is increased then the speed through the "pipeline" also increases.

- What does it really net you? if you cant change the speed that the ram data is put to the pipeline on the proc.. why bother oc'ing?

Because it makes things go faster.

The "pipeline" here may be assumed to be the FSB setting. And we can and do increase that - significantly, in fact - sometimes by as much as 50-75%.

Faster CPU speeds and faster RAM speeds = more work instructions run per unit time.

I could just as easily say "Why bother buying that P4 Extreme Edition?"

- You can oc the proc from 3.0ghz to 4.0ghz.. you can increase the ram from 533 to 700...

What in hell is his point?

- Out you cant increase the ram communication speed with the pipeline to the proc/board
- so why bother?

As established previously, the pipeline is and has always been adjustable. And we bother because we can.

- You could make the ram run at a jigahertz.. but thats at the ram.. NOT the speed it communicates with the pipeline at

No, by running the ram faster, in many cases we actually increase memory bandwidth.

- I've been doing this shit for 15 yrs.. i know about oc'ing.. and I also know its not worth the 10% actual increase you get

It appears your friend has been working from the recesses of his own alimentary canal.

Any time you hear the "I've been doing this for 15 years" line you can rest assured you're being fed excrement.

- Doubling ram speed, doubling even processor speed does not equal double the performance

Okay, that statement may be somewhat correct - a 2 GHz A64 can, on occasion, spank a 3.7 ghz P4. However, doubling performance is often not the objective. Maximizing throughput of an individual system is the objective.

- You can overclock certain components yes.. but you cant overclock the speed at which all those devices communicate with each other on the board

This is factually false. We do this every day. Even if it were true, faster parts still equals faster processing in general.

Your bud should hang out here. His knowledge is lacking. His discourse fails badly.

 

[Retsam]

lol. Good post. However that was one of the longest ive ever seen.



I like his quote of "ive been doing this for 15 years and I know its not worth the 10% improvement"


I dont know what hes been doing for 15 years, but it wasnt dealing with computers lol.

 

[artificialj]

So I think, based on his usage, that this guy is saying "pipeline" and meaning (copper) traces on the actual motherboard. Which is insane.

Saying the traces are the bottleneck is like saying that satellites are bottlenecking your solar panel because they block so much sun, as opposed to saying the flat plane design in the bottleneck.

 

[kunsunoke]

lol. Good post. However that was one of the longest ive ever seen.

You missed some of the posts we used to throw up in the Soapbox. Ah, yes, I remember when....

/me reminisces about the good old days with Asmo, Slay, Enkafan, Uncle Milty, etc.

I like his quote of "ive been doing this for 15 years and I know its not worth the 10% improvement"

I dont know what hes been doing for 15 years, but it wasnt dealing with computers lol.

Self-abuse, perhaps?

 

[Absolute0]

What can i say, the guy you talked to doesn't know shit. So ignore him.

 

[Yiffy]

Did some testing to see if overclocking is worth it, here's my cheapo $30 Celeron M, courtesy of Ebay, memory is DDR266 @ 100FSB and DDR333 @ 167FSB:

Celeron M 350J @ stock (1.3Ghz 100x13)
Doom 3, demo1:
53fps
Serious Sam SE, grand cathedral:
103fps

Celeron M 350J @ 2.17Ghz (167x13)
Doom 3, demo1
90fps
Serious Sam SE, grand cathedral:
170fps








.ouch.

 

[Retsam]

Damn a celeron lol nice!!

 

[Bbq]

- the disadvantages and reduced lifespan of oc'ing anything far outweigh the advantages

Are you really going to running your computer for 10 years?

- no matter how much you oc shit you still cant it thru the bus faster than 500mhz

Been there, done that.

- I find best performance on servers with the new dual core intels with 2mb or better L2 cache

I don't believe.

- A 1mb L2 cache 2.8 is much faster than a 512k 3.2 on pure throughput

No it's not.

- Dont bother with oc'ing..its not worth it in the long run, and 99% of the time your increases that you'll actually feel are minimal at best

Superpi: 2.0 ghz, 200mhz ram, 2-2-2-5: 32 seconds
Superpi: 2.4 ghz, 240mhz ram, 2.5-4-4-10: 28 seconds
Superpi: 2.8 ghz, 280mhz ram (1:1 lol), 3-3-3-8: 25 seconds


- You are also risking premature failure too
- oc'ing reduces parts life in some case to 1/2 of normal expected lifespan

Same point.

- I guess what I dont understand about oc'ing is that no matter what you do, you cant increase the speed that the ram runs thru the board
- You can increase the speed of the ram, chip to chip on the ram.. but NOT on the pipeline to the board
- What you end up with a bunch of data queued (tagged) waiting to cycle back to the proc
- Sure you can overclock ram.. the ram itself.. not the pipeline thru to the core on the board
- What does it really net you? if you cant change the speed that the ram data is put to the pipeline on the proc.. why bother oc'ing?
- You can oc the proc from 3.0ghz to 4.0ghz.. you can increase the ram from 533 to 700...
- Out you cant increase the ram communication speed with the pipeline to the proc/board
- so why bother?
- You could make the ram run at a jigahertz.. but thats at the ram.. NOT the speed it communicates with the pipeline at

You don't have to say the same damn point so many times. And, there is no bottle neck in between those components. Memory controller for the win!

- I've been doing this shit for 15 yrs.. i know about oc'ing.. and I also know its not worth the 10% actual increase you get

I also know that I've been fixing circuit boards for 25 years, even though I'm not 25

- Doubling ram speed, doubling even processor speed does not equal double the performance

O RLY

- You can overclock certain components yes.. but you cant overclock the speed at which all those devices communicate with each other on the board

Hyperlink oc'ing For the win

 

[Captin Insano]

As I would agree that This guy is about as full of shit as anyone here. I find it amazingly funny of all the people here that think the FX line of CPU's are nothing, but yet have to continually compare against them. Even though you may be able to OC that 144 or 146 or 165 to those STOCK levels of FX speeds the FX series of cpu still can go much further for the most part when those other chips just wont. still making the FX cpu's faster and why they are more money. Dont think im knocking anyone. Just pointing out something funny to me. Heck I have more CPU's they a guy should probably have. Just like I said found it interstingly funny of the comparison.

 

[TheAcorn]

Hi guys! I recently talked to someone who was very against overclocking. I'm going to post what he had to say, because I frankly have no idea how to overclock (as much as I wish I did, I just don't understand the guides), and would like someone to interpret his thoughts, and give counter-points, if any exist. Here you go...


- the disadvantages and reduced lifespan of oc'ing anything far outweigh the advantages
- no matter how much you oc shit you still cant it thru the bus faster than 500mhz
- the on chip processing may be faster, but the throughput back to the components on the bus are still the largest bottleneck of any x86 architecture
- I find best performance on servers with the new dual core intels with 2mb or better L2 cache
- A 1mb L2 cache 2.8 is much faster than a 512k 3.2 on pure throughput
- Dont bother with oc'ing..its not worth it in the long run, and 99% of the time your increases that you'll actually feel are minimal at best
- You are also risking premature failure too
- oc'ing reduces parts life in some case to 1/2 of normal expected lifespan
- I guess what I dont understand about oc'ing is that no matter what you do, you cant increase the speed that the ram runs thru the board
- You can increase the speed of the ram, chip to chip on the ram.. but NOT on the pipeline to the board
- What you end up with a bunch of data queued (tagged) waiting to cycle back to the proc
- Sure you can overclock ram.. the ram itself.. not the pipeline thru to the core on the board
- What does it really net you? if you cant change the speed that the ram data is put to the pipeline on the proc.. why bother oc'ing?
- You can oc the proc from 3.0ghz to 4.0ghz.. you can increase the ram from 533 to 700...
- Out you cant increase the ram communication speed with the pipeline to the proc/board
- so why bother?
- You could make the ram run at a jigahertz.. but thats at the ram.. NOT the speed it communicates with the pipeline at
- I've been doing this shit for 15 yrs.. i know about oc'ing.. and I also know its not worth the 10% actual increase you get
- Doubling ram speed, doubling even processor speed does not equal double the performance
- You can overclock certain components yes.. but you cant overclock the speed at which all those devices communicate with each other on the board

Your pal is an idiot, and I, for one, am amazed that despite 'doing this shit for 15 years' he still doesn't know what a benchmark is for.

Also, is he still using a Pentium II? What sort of lifespan does he require out of hardware? Who keeps a PC for 10 years? Last summer I filled a few K6's full of lead for target practice, they were completely useless. I could also mention that every single one of them were overclocked, and fully functional.

Bottom line is, and this is gratuitous to say at this point in the thread, but being a [H] forum member, I have to speak out and defend our art: Your buddy has no idea what he's talking about, and his opinons were 'owned' by the first three posts.

/OC'ing my gear since 1996.

 

[spine]

My £135 Opteron 146 runs faster than a £630 Athlon 64 FX-57.

Overclocking is mutha-fucking worth it, ok?

 

[superkdogg]

My £135 Opteron 146 runs faster than a £630 Athlon 64 FX-57.

Overclocking is mutha-fucking worth it, ok?

Exactly. My winchester outran and outbenched every other cpu on the retail market for quite a while despite being the cheapest available 939 proc when I got it. For $150 I got $1000 worth of performance. How is that bad?

Even better: If your "friend" is so anti-OC, have him explain this one to me. A company, let's say ATI for example has extra stock of one particular core. It used to be top of the line, but now it's a gen or two old. Still nice, but aging and there's plenty of stock of it. They decide to bios gimp the core and release it cheaper-mind you it's the same hardware, just firmware tells it not to stretch its legs. Then I get ahold of this piece, remove the nerf and get a x850xt-pe for $179 because it was labeled x800GTO2. IT IS THE SAME PIECE OF HARDWARE!

The other puzzling thing is even if it were true that the busses are the bottlenecks (which it's not fair to generalize this although it's sometimes the case), then wouldn't overclocking be doubly effective since it increases the speeds of these busses? That's why a 2.4 a, b, and c Pentium 4 show differences in performance at the same clock speed.

 

[mwarps]

By even replying I feel like I am feeding the trolls. The OP's "buddy" is a mindless idiot and needs to have his balls overclocked so maybe he can get a taste of what it's like to have a pair. Every single point in the OP is wrong, fud, or entirely non-sensical.

 

[Asian Dub Foundation]

shit when i game i can feel the difference between a 2.6ghz and 1.8 ghz

 

[Lukano]

I for one would like the OP's buddy to register an account here and help us intelligently and systematically educate him on his misconceptions and misinformation.

 

[Firebat]

I for one would like the OP's buddy to register an account here and help us intelligently and systematically educate him on his misconceptions and misinformation.

QFT

Get him over here.

 

[atom]

- What does it really net you? if you cant change the speed that the ram data is put to the pipeline on the proc.. why bother oc'ing?

This guy obviously just got his A+ certification and is throwing around lingo he really doesnt understand fully. The quoted statement is absoultely false and doesnt even make sense. Yes, overclocking can reduce your lifetime of a product. This is true. Yes, the results are often minimal. This is true. The only thing is that your cpu will last probably 10 years overclocked as apposed to 45 years running regularly. WTF? Who gives a crap! Do you still use that Apple II GS to check your email? HELL NO. In fact, because chips become so obsolete overclocking increases their useful life span. I have a 1 gighz Athlon that I have overclocked to 1.4ghz and is still good for a lot of games. I dont have money for anything else and without overclocking I couldnt game.