Is overclocking a method of making equivalent processors?

[legocitytruck]

If two processors have all the same specs other than clock speed i.e., same amount of cache, same manufacturing tech... If you overclock the lower speed processor to the high sped processor's level, have you in essence created the same two processors and saved yourself a little money along the way? Even though both processors have the same specifications and now clock speeds, is thee anything else that will keep the overclocked processor from being exactly as fast as the more expensive standard clocked processor?

 

[Spoelie]

No, in fact it's usually the other way around:
In order to run the overclocked processor at the same speed, there are other parts that need to run faster as well, granting the overclocked processor a small (sometimes immeasurable) performance advantage. The exception are the extreme/black editions which have unlocked multipliers.

Energy consumption/heat will usually be higher on overclocked parts tho.

 

[alyarb]

energy consumption will be higher, but not prohibitively higher. 45nm dual cores can easily go from 2.5 to 3.5 ghz at the cost of a dozen or so watts under load. finding the "g spot" among memory, buses, and voltages involves running some of these components out of spec, which is why the overclocked system would not be identical to a stock-clocked higher-freq system. however, when it comes to the user's experience, a Q9450 at 3.2 GHz is indistinguishable from a stock-clocked Q9770.

 

[apoppin]

Originally posted by: legocitytruck
If two processors have all the same specs other than clock speed i.e., same amount of cache, same manufacturing tech... If you overclock the lower speed processor to the high sped processor's level, have you in essence created the same two processors and saved yourself a little money along the way? Even though both processors have the same specifications and now clock speeds, is thee anything else that will keep the overclocked processor from being exactly as fast as the more expensive standard clocked processor?

The idea of overclocking is exactly that; what is better is running at a higher frequency than what is even offered - e.g. there are no C2Qs that are clocked by Intel at 4.0 GHz, yet mine is there - allowing me to compete more evenly with the new core i7 in gaming and thus hold off my upgrade for a bit longer while not suffering.

the only thing that is a disadvantage is that Overclocking voids the manufacturer's warranty in many cases and often you have to run with higher voltage to keep your CPU at the higher clocks - so you get more heat, less lifespan and usually you have to get an aftermarket cooler to handle the higher temps generated by overvolting and overclocking

over all, as a balance, it is for most of us experienced O/C'ers, a win to overclock

 

[Eureka]

Originally posted by: legocitytruck
If two processors have all the same specs other than clock speed i.e., same amount of cache, same manufacturing tech... If you overclock the lower speed processor to the high sped processor's level, have you in essence created the same two processors and saved yourself a little money along the way? Even though both processors have the same specifications and now clock speeds, is thee anything else that will keep the overclocked processor from being exactly as fast as the more expensive standard clocked processor?

Technically its what it is... if the core and specs are the same, and the speed is higher, its just manufacturer overclocked ( or if you think about it the other way, the lower cores are actually underclocked). This could be from any reason from slightly worse chips to just creating variety in the market.

 

[ShawnD1]

Originally posted by: legocitytruck
If two processors have all the same specs other than clock speed i.e., same amount of cache, same manufacturing tech... If you overclock the lower speed processor to the high sped processor's level, have you in essence created the same two processors and saved yourself a little money along the way?

The way you've written it, yes you can do that. Overclocking 2.5ghz to 3ghz makes it as fast as a processor that is factory set to 3ghz.

The other side of the coin is that the "better" processor really is better. If you had 2.5ghz and 3ghz models of the same basic processor, it's reasonable to assume the 3ghz model will overclock higher or use less power to overclock to the same level. Intel might be able to get away with randomly binning things just to offer different price points (hence people getting 80% overclocked celerons), but it's reasonable to assume that AMD processors are sold based on how good the processor is. People can unlock the cores and overclock them, but it often requires significant voltage increases, which means significant heat increases. A quad core that is actually sold as a quad-core might take 60W of power, but a tri-core that you've unlocked and overclocked to the same level of performance might take 100W just to get that same level of performance.

 

[soccerballtux]

Originally posted by: ShawnD1

Originally posted by: legocitytruck
If two processors have all the same specs other than clock speed i.e., same amount of cache, same manufacturing tech... If you overclock the lower speed processor to the high sped processor's level, have you in essence created the same two processors and saved yourself a little money along the way?

The way you've written it, yes you can do that. Overclocking 2.5ghz to 3ghz makes it as fast as a processor that is factory set to 3ghz.

The other side of the coin is that the "better" processor really is better. If you had 2.5ghz and 3ghz models of the same basic processor, it's reasonable to assume the 3ghz model will overclock higher or use less power to overclock to the same level. Intel might be able to get away with randomly binning things just to offer different price points (hence people getting 80% overclocked celerons), but it's reasonable to assume that AMD processors are sold based on how good the processor is. People can unlock the cores and overclock them, but it often requires significant voltage increases, which means significant heat increases. A quad core that is actually sold as a quad-core might take 60W of power, but a tri-core that you've unlocked and overclocked to the same level of performance might take 100W just to get that same level of performance.

Eh, not significantly higher-- less than +10% vcore for me. At 3.5Ghz I only had to add +.12vcore. That's with the 4th core unlocked too. Free quad for me!

 

[Fox5]

An overclocked processor might perform slightly better than the stock, since you're overclocking the bus speed and thus reducing latency to memory. It would be nearly indistinguishable.

If you have a processor with unlocked multipliers (like AMD's black editions) then you can just set the multiplier to the same as another processor and they should be exactly the same for all intents and purposes. (as long as there were no other differences between the two)

 

[aigomorla]

okey what your asking is.... in a very rough sense..

Can a i7 920 be as fast as a i7 975? The answer to that is yes, but it will cost ya.

To get a 920 moderately as fast as a stock 975, it would require you to push your entire system, ram board and all. While the 975 would be in its ideal possition.

What you need to realize tho, is the 975 has a higher climb wall then the 920, so if both were tuned the 975 would gobble a 920.

If you get lucky, u might get a golden 920, but this has luck to do with it.

Im still betting the fastest quadcore processor is inside a dell not overclocked. >.<

 

[ShawnD1]

Originally posted by: soccerballtux
Eh, not significantly higher-- less than +10% vcore for me. At 3.5Ghz I only had to add +.12vcore. That's with the 4th core unlocked too. Free quad for me!

The reason this is possibly incorrect is because people forget to take a baseline voltage to see what is required for the stock speed.

Let's say you buy a processor and the "auto" voltage is 1.25V. After some testing, you find that the maximum overclock requires a voltage of 1.30V in order to pass linpack. A lot of people will incorrectly claim that this was a voltage increase of 4%, calculated as 1.30V divided by 1.25V. The reason this is wrong is because 1.30V is a minimum stable voltage whereas 1.25 is a wildly over-estimated voltage to ensure the processor will not have math errors when set to auto. If the processor is run at stock speed but using minimum voltage, just like you would do for an overclock, you'll find that many processors can run at much lower voltages than auto.

As a real world example of this, my E6600 has an auto voltage of 1.25V but the actual minimum voltage for stock speed is 1.10V. The maximum overclock I can get with this motherboard requires a CPU voltage of 1.30V. While this does fall within the "auto" range and people will say it's done at stock voltage (implying 0% voltage increase), the reality is that overclocking my E6600 requires me to boost the minimum voltage by 18%, calculated as 1.30V divided by 1.10V.

If you test your unlocked processor, I'm very confident that you will find a similar story. If AMD's auto setting adds an extra 15% voltage at stock just to be safe and then you throw another 10% on top of that, this stuff starts to add up. 115% of minimum voltage is auto, then you're 110% of that for a total of 26.5% voltage increase.

I don't know if semiconductors follow the same electricity rules as normal circuits, but power consumption usually increases exponentially with voltage. A voltage increase of 26.5% would cause a power increase of 60%. Then again that might sound reasonable depending on how good the overclock is.

 

[Idontcare]

Originally posted by: ShawnD1
I don't know if semiconductors follow the same electricity rules as normal circuits, but power consumption usually increases exponentially with voltage. A voltage increase of 26.5% would cause a power increase of 60%. Then again that might sound reasonable depending on how good the overclock is.

We had a pretty cool thread on this a while back, at least I thought it was neat. Checkout the last 5 or 6 posts in this thread: http://forums.anandtech.com/me...id=28&threadid=2278938

 

[Rubycon]

Think of what your overclock is in percentage and add that to 100 when your overclocked CPU is running at 100% load!

Xeon W3540 running at 155%. It would be neat if the task manager in Windows worked that way. :laugh:

 

[dflynchimp]

Originally posted by: Rubycon
Think of what your overclock is in percentage and add that to 100 when your overclocked CPU is running at 100% load!

Xeon W3540 running at 155%. It would be neat if the task manager in Windows worked that way. :laugh:

eh, but performance doesn't always scale linearly with clockspeed, because the average PC is chalk full of potential bottlenecks (HDD, Memory, etc). To get a 1:1 ratio of clock speed vs performance you'd have to increase performance of every other component by an equivalent percentage.

Even then there's the OS and software side of things, which I'm no expert on but can at least say for sure that it's never an ideal scaling.

 

[taltamir]

Originally posted by: ShawnD1

Originally posted by: soccerballtux
Eh, not significantly higher-- less than +10% vcore for me. At 3.5Ghz I only had to add +.12vcore. That's with the 4th core unlocked too. Free quad for me!

The reason this is possibly incorrect is because people forget to take a baseline voltage to see what is required for the stock speed.

Let's say you buy a processor and the "auto" voltage is 1.25V. After some testing, you find that the maximum overclock requires a voltage of 1.30V in order to pass linpack. A lot of people will incorrectly claim that this was a voltage increase of 4%, calculated as 1.30V divided by 1.25V. The reason this is wrong is because 1.30V is a minimum stable voltage whereas 1.25 is a wildly over-estimated voltage to ensure the processor will not have math errors when set to auto. If the processor is run at stock speed but using minimum voltage, just like you would do for an overclock, you'll find that many processors can run at much lower voltages than auto.

As a real world example of this, my E6600 has an auto voltage of 1.25V but the actual minimum voltage for stock speed is 1.10V. The maximum overclock I can get with this motherboard requires a CPU voltage of 1.30V. While this does fall within the "auto" range and people will say it's done at stock voltage (implying 0% voltage increase), the reality is that overclocking my E6600 requires me to boost the minimum voltage by 18%, calculated as 1.30V divided by 1.10V.

If you test your unlocked processor, I'm very confident that you will find a similar story. If AMD's auto setting adds an extra 15% voltage at stock just to be safe and then you throw another 10% on top of that, this stuff starts to add up. 115% of minimum voltage is auto, then you're 110% of that for a total of 26.5% voltage increase.

I don't know if semiconductors follow the same electricity rules as normal circuits, but power consumption usually increases exponentially with voltage. A voltage increase of 26.5% would cause a power increase of 60%. Then again that might sound reasonable depending on how good the overclock is.

true... also there is the whole clockspeed issue.
The same processor at the same voltage, runnning at 3ghz will consume more power than running at 2ghz. This is the whole reason the processor switches its multiplier down when idle.