One of my computers was struggling a bit running Windows 7. It is a 3.4GHZ/1GB RAM that I'd built back in 2005. Ran XP like a treat but struggled under a newer operating system.

Thought I'd try upgrading it as it's got a fairly decent graphics card and a large hard disk.

Ended up buying some extra RAM off eBAY 512MB (only cost about £13), then overlocked the fooker in the BIOS. By upping the FSB (front side bus speed) I turned a 3.4GHZ PC into a 3.6GHZ (the fastest speed it seems to be stable at). It performs like a completely different machine now.

The motherboard (Aopen) is designed to be easily overclocked and even comes with a smart util that enables you to do it online (see screenshot).

All you need to do is go into your BIOS when the comp starts (press Del on most machines). Find the "Frequency/voltage control" page and increase the FSB speed. Try it at different settings to see what's stable. It's very easy to do in most Award BIOS.

If you want to run Windows 7 on your PC, here's what it takes:
1 gigahertz (GHz) or faster 32-bit (x86) or 64-bit (x64) processor.

1 gigabyte (GB) RAM (32-bit) or 2 GB RAM (64-bit).

16 GB available hard disk space (32-bit) or 20 GB (64-bit).

DirectX 9 graphics device with WDDM 1.0 or higher driver.

This could be taken as the bare minimum, But we all know that if you give the bare minimum thats all you will get in return.

Very interesting. My comp is dead at the moment and I am getting it fixed soon (no graphics) I have always toyed with the idea of overclocking it but never really knew what to do and to be honest was a bit worried about it blowing up.

Overclocking isn't usually worth it, increasing the processor speed by 5% or so but this increases the power by much more which means the guts of the processor get HOT. It is rare that the processor is the bottle neck, so the 5% increase would hardly ever be realised anyway.

The biggest speed increase is by getting a faster hard-drive, that's the lowest speed component (apart from the idiot operating it) in the system and much used by windows.

What you say is incorrect. Intel processors are very good thermal management and will slow down their operations if the processor gets dangerously hot.

http://www.intel.com/support/processors/pentium4/sb/cs-007999.htm

Check this vid from Tom's hardware that shows an Intel processor dealing with overheating and an AMD processor trying to do the same thing!

[youtube]NxNUK3U73SI[/youtube] NxNUK3U73SI

What you say about the HDD isn't correct either. Hard drive performance is usually noticable in a home environment when the computer performs paging operations. This is where the computer swaps parts of memory out to disk when RAM is getting tight. As disk access is slower than RAM, this will cause a noticable performance drop as the data has to be read from the slower disk back into RAM.

Windows Task manager will tell you all about this.

The million dollar question is it depends on what you want to do. I manage a 12TB, 10,500 email environment so fast disk access really benefits an exchange environment as lots of data is being transfered to and from the disks as users operate their Outlook clients.

On the other hand I have a farm of servers that perform data processing processing operations to OCR scanned invoice numbers to match against purchase orders. It uses loads of processing power to do this and fast disk access doesn't make any difference whatsoever to this application.

The same would be true in a home environment if you used CAD type applications or rendered images. A faster processor and more RAM will defo benefit a graphic intensive OS such as Windows.

As for bottlenecks, I could talk all day about data buses but it would probably send everyone to sleep.

The point I'm trying to make is, if you were thinking of chucking out a slow PC, try overclocking it first and adding some cheap RAM to get a bit more life out of it.

Neil, I think you have confirmed what I said about overclocking, the speed will be reduced downward by more than the 5% increase when the processor gets hot. Additionally the thermal management system is only designed to work correctly when the processor is operated within its prescribed specification, if it is overclocked, parts of the processor that are not close to the thermal sensor could get hotter than they should.

Windows does a lot more than paging with the HDD, switch your paging file off and many applications will still be thrashing the HDD. It is not uncommon when a program is working hard that the processor is only at 10% because it is waiting for the HDD and yet there is loads of physical RAM still free.

Overclocking can be quite unpredictable, I have had systems that wouldn't work when ambient temperature went down, and others that wouldn't work when ambient temperatures went up etc etc.

The HDD escpecially if IDE is by far the item that overall slows the system down, I was suggesting that uprating the HDD would benefit much more than overclocking by 5% and be less risky.

Obviously I agree that a shortage of RAM makes a very significant difference, but on the other hand a surplus of RAM can actually slow things down a bit.

As a slight aside, I have a system where the processor is overclocked by 100% and this was done by choosing the lowest speed processor, the higher speed processor which would only need to be overclocked by 33% to achieve the same speed is renowned for failing because the test parameters were tighter on the higher speed processors and if they were capable of being overclocked at 33% were re-badged with an enormous mark-up in price.

The overall increase in speed by overclocking is not worth-it unless you are really desperate or just want the prestige.

Most modern fans are adequate to cool a small increase in temperature, if not additional cooling can easily be added.

Processor manufacturers like Intel don't make different speeds of processor. There is only one model, say a Pentium 4. It's the testing of the processor during the post manufacturing stage that determines what the multiplier is locked at to determine the frequency (speed). You seem to be aware of this from info in your previous post.

A 3.4GHZ CPU is unlikely to have any probs operating at a stable speed of 3.6GHZ. This is why many motherboards similar to mine offer a very easy way of overclocking and reporting on the system health as per above screenshoot.

Common things that will cause slow performance are:

Badly written applications
Anti-Virus
Disk fragmentation
Processor
Bus speed
RAM
Caching effectiveness
Disk spindle speed
Disk Connectivity
Motherboard architecture

In an estate of 10,500 machines, I'd have to say the biggest cause of performance problems in order are:

1) Loads of services running in Windows
2) Old operating systems that are badly fragmented and have had lots of programs installed (needing re-imaging)
3) Badly configured anti-virus running on PC's under the minimum spec
4) shit applications.

I wouldn't get a new PC out of the box and overclock it, but it's ideal for a computer that's reaching the end of its useful life.

Ok putting some figures on the table to give some sort of perspective.

On both my laptop and desktop, I uprated the HDDs to the next higher spin speed and with more and better built in cache. The boot time improved by about 40% in each case, the application start times achieved similar figures.

Upping my memory (on XP) from 512MB to 1G gave a disappointing increase of about 15% in boot time and little difference in applications (but I don't tend to have many applications running at the same time).

If I upped the processor speed by 6% it would have an insignificant difference on boot time or application start time as both are mainly limited by reading and writing data to the HDD.

Browsing the internet, most of the bottle-neck is my broadband speed, but there is significant HDD activity because of the disk based caching.

Video editing, even though the processor speed should help me out here, there is still an extreme amount of HDD activity.

I was struggling to think of an example where a 6% increase in processor speed would significantly make any difference to me, however here are a few.

Repetitive mathematical things I do occasionally where there is low amounts of data to transfer to or from disk - then I may get a whopping 6% improvement, but that is a bit specialised.

The formatting stage of medium sized pictures, PDFs or DOC files, files big enough to require a lot of processing, but not big enough to throw Windows into a significant amount of disk swapping.

Software code efficiency has an awful lot to do with how applications perform. There used to be a saying once; don't bother optimising your code as Intel will bring out a faster processor.

I program in Visual Studio a lot and increase my processor speed and RAM did make a noticable difference to me.

Much Microsoft stuff centres around the re-use of code, DLL's and the .NET framework are a prime example of this. This makes it easier for the programmer to quickly produce software, known as RAD - rapid application development.

The most efficient code to write in short of binary, is one that runs as close to the processor as possible, this is called assembler. The screenshot gives an example of assembler language. Debug comes with every version of Windows and Dos.

Steve Gibson (http://www.grc.com/stevegibson.htm) writes all his stuff in assembler. This causes the program files to be very small and efficient, although writing complete programs in assembler is way beyond the scope of most developers.

Regarding internet, broadband speed isn't everything as I'm sure you'll agree; DNS response times, latency, contention ratios, USB transfer, home routers, cachcing and web page accelerating all have an impact.

We sell broadband and internet filtering to schools in work and it always makes us laugh when ISPs try to compare our 10meg sync fibre service to 8 meg ADSL contended at 50:1.

When I said broadband speed I meant the speed the internet data arrives at my computer through whatever means, it is the most significant speed limit on my surfing with my system, if the data arrived twice as quick I would probably get a near twice speed improvement, if I doubled the speed of my computer, the difference would hardly be noticeable in comparison.

I am an assembly programmer and what you showed there is mostly gobbledygook because the majority of it is data code not program code and so the disassembly is not of a program.

Something like visual studio or generic programming will be more processor speed dependant, but is not typical PC usage. Even still is overclocking by 6% to obtain a (say) 3% overall increase in speed, is it worth it, especially if the processor does get a bit warmer and may reduce its speed by 30% now and then? If time is that precious you would gain more by uprating your kettle

Steve Gibson is a genius, despite upsetting Microsoft on a number of occasions and them taking no notice of his warnings for years, they have eventually incorporated some of his suggestions.

Doubling the speed of your internet link wouldn't make much of a differnece. I struggle to tell the difference between my DSL at home and work's 1 gig fibre for typical browsing. HTTP is a stateless protocol and most web pages don't really put that much of a strain on the browsing computer even with flash or java based web apps.

The example was to show x86 assembler mnemonics and not that of a working program (I like our constructive arguments, but not that much that I'd write a working program as an example :-)).

I can only comment on how overclocking benefits my PC and not anyone else's. It does improve the speed at which my computer runs and compiles programs in the software that I use most. If I get the time or inclination I might benchmark it before and after and post the results.

I've seen a few recent example in work of application processor troubles. As standard web apps don't do much for mass data entry, most developers either use Windows clients or Java for delivery of their applications.

As most people like thin clients these days, Citrix provides a great way of delivering apps to desktops without the full client install. The common alternative to this is Java.

I've seen so many Java apps max out servers and it's not that much to do with the processor speed, but because of the fact many Java apps aren't written to take advantage of multiple processor or cores.

Stuff like Microsoft SQL server and Exchange seem to do a wonderful job on multi-processor servers, but Java in many apps just whack the hell out of one core and don't spread the load amongst all cores effectively.

It's not just the single CPU that benefits the speed of a computer, but the computer's ability to offload other instructions elsewhere. The Amiga was an early example of this with its custom chipset. I remember buying my first 287 to run Autocad and the moderator of this forum helping me install it about 20 years ago!

Microsoft aren't particulary enthusiastic about Gibson, but they do listen to him.

Oh Neil, You take me on a fond old trip down memory lane,They were the days! It is you that's helping me these days for which I am off course very grateful. Thank you ;-}

I never used a seperate co-processor, jumped from a 286 to a 386DX, but on the 286 I used that little co-processor emulator (it was only a few K long) that worked nearly as fast as a 287 but cheated because it didn't do things to the same precision.

I might try some benchmarks later in the week, except I'll be cautious and underclock instead of overclock to find the differentials. My system doesn't have a high speed HDD system just PATA.

Java has got multi-threading (so I guess by now it would take advantage of multi-processors), but as you say, people rarely use it effectively, was just reading up on it the other day in a 2000 Sun book (Core Java Vol II) probably somewhat outdated but still beyond my ken.

I was careful to say that if the internet data arrived twice as fast as oppose to the link bandwidth being twice as much.

I used Citrix for a while to remotely run later versions of office than I had, I was quite impressed how smooth/fast it was unlike previous forays with Xwindow terminals.

Is this the computer I use to "Swipe in" at work today by any chance (PB 286)?

Is that still going? That thing must be 15 years old. I remember having to write special code to grab threads on Windows NT to bring the app to the front so it read the code into the clock app when you swiped your card. Ahh the joys of VB6!

As for your server, that was my dad's old PC from 1997. I'd noticed that the SCSI card packed in a few years ago but it was still working on the Win 2000 software RAID on IDE.

Your boss had his money's worth out of that kit. It's done a cracking job hosting your files and SQL databases for 10 years. Guess it shows that something does not have to be mega fast or expensive to fulfill a particular need.

I remember the co-processor emulator. I had it for my 486SX. The interesting thing was my 286 with 287 could render images faster than the 486sx.

I quite liked Interlnk that came with DOS 6, I was able to network the two computers together using a serial link. I learn't a lot when I fuct those DOS computers. When Defrag messed up the FAT, that was a whole load of fun!

Was it not the 486DX rather than the 486SX that had the built in maths co proccessor? It is going back a while that one Neil?

By the way there is a new server there ready and waiting - You know where!

I think it was built into both and disabled on the SX

that's correct, it was the same chip, but I think the SX had a missing pin. The DX upgrade was actually a full blown chip, bit of a waste of money. SX's are actually DX's which failed the FPU tests.

My 486 was an Olivetti, the monitor was a bit naff (until it broke and they issued a replacement). The onboard Oak graphics card was a pile of pants too.

Nigel it makes me laugh when Boss Hogg calls me into fix that server, I just reseat the processor and card, hit it a couple of times and it's good to go for another 2 years!

Would you believe that it's actually 486 chips that run parts of the systems on the Hubble telescope!

http://www.nasa.gov/mission_pages/hubble/servicing/SM4/news/status_update_20081015.html

Been water-cooling an AMD Phenom II X4 965 Black for six months now, up at 3780Mhz (210x18) (stock is 3400Mhz 200x17), well over the performance of a Phenom II X6 1055 six-core with temperatures only around 5c over ambient, so my experience of overclocking has been superb, been at it for years, always with AMD's, never a failure yet. The latest gen (Phenoms) have thermal management to reduce clock speeds if needed, but if you have to rely on that, you're doing something wrong in the first place! That video is a VERY old Socket 478(?) Athlon, not a good time for AMD.

I'm running a twin-loop system on my Phenom rig at the moment with Water on the graphics too (EVGA Nvidia 280 GTX HC). As well as a watercooled server (just CPU, off the shelf cooler), and a watercooled CPU/Graphics (Nvidia 8800 Ultra SC) Ubuntu Linux box, though that is an Intel Core 2 Duo E4700 at stock speeds.

Couldn't recommend liquid cooling highly enough as even if you only use your machine at stock speeds, you hugely reduce the amount of noise (and dust-bunnies) generated, feel free to PM me if you want to discuss the liquid way, I build mine into tower cases (cut my own slots, vents etc) , the early Antec cases are best as they are huge cases, very strong, with space for good quality radiators and fans.

H2O inside

Edit -> Quite ironically, the only processor I've ever toasted was an Intel E2180, got a bit keen with the ol' VCore ... Intel protective systems - fail. My bad.