Experiments done, lessons learned.

just over 1.5 years ago i built my last PC. it also happen to be my first time doing so. then, as now, i was short on cash, a bit shorter then now obviously :biggrin:

the point is, i only bought what i did not have laying around in spare parts. so i got a tower, a motherboard and a processor, the rest i had.

my goal, was to buy the biggest i could afford, which i thought was the 2.6 celeron i ended up with.

the purpose to this whole thread? well it is contained within the next 2 screens.




screen #1 is my machine as it was when i did the critique. it was a 2.6 celeron with 512 megs of PC133 ram and a 64 meg radeon 9000 video card.

screen #2 is its current configuration of 1.8 P4 with 512 megs of PC2100 and the radeon 9000 card.

the moral? i could have originally gotten the P4 chip, for the price of the 2.6 (since at that time the 2.6 was the biggest celeron speed and more expensive than now)

the parts were given to me from a machine also recently upgraded. the map now runs at twice the frames per second, due to only two parts that i swapped.

bottom line, bigger, or faster doesn't mean better. :smile:

I'm almost positive that difference is RAM related rather than CPU dependant.

Those stats pertain to the video card mostly ... unless I'm reading them incorrectly.

Tracer is correct, by the way. Video card = 1st prominent determination of latency in games, Ram = 2nd, then the Bus, then the CPU, cache, and HDD are all about the same in "effect".

But, the other thing you have to remember is that a Celeron isn't "bigger" then a P4 even if it says it runs "faster" because that term is really relative when talking about CPUs. But, they have different registers. If I remember correctly, the Celeron is one less a multiplication register then the P4 (Or something like that). Multiplication is a very useful operation because it's faster most of the time (or something, it's been awhile since I did that low level of work).

I remember a friend of mine talking about the new L68 (Mac) CPUs and how they don't "have" addition registers. Just lots of multiplication and division with increment and decrement. I'm not sure if that's true though.

Just so you know.

stands looking at monitor with glassy eyes

uhhh, ok then.. BTW, the FPS went from 8 to 17 in that exact position. :smile:

mumbles

you guys are to damned smart. makes me wonder why you hang out here, and with people as obviously far behind as i. :/

sometimes you guys put me in mind of those people who give to the needy. they dole out small amounts of cash to the homeless. in this case, you pass along tid-bits of knowledge instead.

Sounds pretty true, but always look at the amd chips, ?100 for a 3ghz
intel equal that is 64bit now :razz: (by intel equal, i mean the 3000 at
the end is what speed of an intel chip it is equal to, it actually runs
at 2ghz)

Clock speed is becoming less important anyway, The size of the tracks
on the chip are, along with the size of the board. 4ghz is currently
the max processor clock speed because how big the tracks are and the
speed of light. over 4ghz, you can have a 1 at one side of the board,
and a 0 at the other, so weirdness will occurr, and the weirdness is
unknown to me.

anyway, im buying thatplus a board that supports the 800mhz FSB, and
512 of ram when i can, and i'm pretty pleased with my ATI raedon
9200SE's preformance at HL2, vis lag in places, as my HD needs
seriously defragging, and its got to load stuff.

Probably one of the best ways to make your PC run faster is to defragment and increase the vitrual memory space.

Happy computing!

<DIV class=quote>
<DIV class=quotetitle>? quoting Dred_furst</DIV>
<DIV class=quotetext>4ghz is currently the max processor clock speed because how big the tracks are and the speed of light.</DIV></DIV>
well that settles it, i refuse to play HL2 on a 4ghz machine. it would be akin to my sex life, over before the lights even go out.... but thats a story best left to another thread.. :/

You'll probably be getting AMD 64 4500+ very soon that runs at 3ghz imo :biggrin:

AMD has by far the more powerful chips.

If anyone has noticed, the AMD chips run hotter because of the smaller parts, so the bigger heatsinks.

Last time I checked the P4 chips were smaller ... I don't know how large the 64 bit AMDs are.

Orph, FPS is directly related to BUS speed and probably the only really noticable effect a CPU has on video performance (everything is "directly" related to the BUS, what I mean here is video performance).

By the way, don't even bother with getting chips as high as 4Ghz ... really, there will be new architectures in the next few years. I'm thinking everyone's next upgrade (which should normally last you 2 - 4 years if you put together a good set of parts) will be their last X86 machine ... at least, that's what I'm hoping.

by smaller, i also mean the track size, the amount can be packed in. I
know AMD pack in a lot more computing into their chips than intel.

by smaller, i also mean the track size, the amount can be packed in. I
know AMD pack in a lot more computing into their chips than intel.

I would like to buy some of this "computing", please. Where can I find some?

by computing i mean power to compute, i.e. power.

You'll probably be getting AMD 64 4500+ very soon that runs at 3ghz imo :biggrin:

AMD has by far the more powerful chips.

If anyone has noticed, the AMD chips run hotter because of the smaller parts, so the bigger heatsinks.

AMD chips don't run hotter than intels AFAIK.

<DIV class=quote>
<DIV class=quotetitle>? quoting Dred_furst</DIV>
<DIV class=quotetext>by computing i mean power to compute, i.e. power.
</DIV></DIV>

You mean more transistors. As far as I know Intel and AMD do not use significantly different process sizes. What are they at now? 0.13-0.09 Micron?
I would also challenge the idea that the speed of light is a major limiting factor in CPU clock speed. Surely heat dissipation issues play a bigger role at this point.

the wavelength of light is "reduced" when the hz of the changes size.
elecricity travels at the speed of light, and so has the same maximum
speed. the length/width of the current ATX boards are equal to this
"reduced" wavelength. therefore any higher frequency means the same
track can have both a 1 and a 0 on the same track.

Happy understanding!

So basically there going so fast that you cannot tell if it's a one or a zero?

no, it knows wether its a 1 or a zero, its that one end of the board sees a 1 and the other end sees a 0.

nods and agrees

<DIV class=quote>
<DIV class=quotetitle>? quoting Dred_furst</DIV>
<DIV class=quotetext>the wavelength of light is "reduced" when the hz of the changes size. elecricity travels at the speed of light, and so has the same maximum speed. the length/width of the current ATX boards are equal to this "reduced" wavelength. therefore any higher frequency means the same track can have both a 1 and a 0 on the same track.

Happy understanding!
</DIV></DIV>

Boy that made sense. You seem to have left out some critical noun. I think I'd understand a bit better if you were to use complete sentences.

The wavelength of light has no bearing on it's velocity... I've no idea what you are trying to say. I don't know much about how computer chips function, but I am clear on the basic physics.

The "wavelength" of an electric signal is determined by the clock frequency. 4Ghz means 4 billion (4x10⁹) cycles per second. An electric signal (or light) could travel 7.5 cm in one cycle at that frequency (0.25 ns). I presume they use a square wave signal (clock pulse) so you would have a high signal for .125 ns (3.75 cm).
The way I understand your statement is that you have both a high and a low signal in the track at one time. I'll ignore the fact that it is impossible to avoid this at some time interval, and just think about whether it is possible to have all high or all low. I am envisioning a track as a single circuit within the chip. Given that, the only way for the track to avoid having states of all high and all low sequentially is if it is longer than 3.75 cm. I find it hard to believe that these circuits are that long.

It's funny, that no one mentioned the whole problem with these "tests" is one chips is a celeron while other is a P4.

Celeron is an OBSOLETE chip compared to P4 or any of the AMD
XP's. Celerons have horrid architecture and ONLY 128 or 256K of
L2 Cache, back in the day of the sub ghz processors, this didn't mean
much, seeming how games would rarely use the 128K of cache, while apps,
would, now however, Many games use a whole 512, and even more.

Most of the results you have their come from the P4 having the better
architeture, and moving away from ghetto PC133 SDRam and going to
PC2100 DDR.

yeah Celerons are basically P4's that were rejects and are missing some math instruction sets. From what I have heard the AMD Semprons are a better choice for a budget chip; they look comparable to a pre-Barton Athlon XP from what I can see, and are not 'reject' chips. Still, a P4 or Athlon64 is optimal and will maul either budget chip in the same range.

To give you an idea.. my friend has a Celeron 800mhz... but my old Pentium II 400mhz is a noticably better machine, even though it runs at half the speed.

by track, i mean a wire on a circuit board (eletronics)

and i suppose all tracks are going to be less than 3.5cm on the processor as they are about 3.5cm big.

Thanks for doing the maths! thats added to my understanding :biggrin: explains other factors too :biggrin:

The wavelength of light has no bearing on it's velocity.

This is only true in free space. Once you have waves propagating in a
medium, like light in glass, or RF signals through copper, you get an
effect called dispersion; different wavelengths (frequencies) of the
signal travel at different speeds.

For example, blue light is slower than red light when it's travelling
through most transparent media. That's why a prism splits up light into
different colours.

Now if you have a very brief pulse, it requires lots of bandwidth, i.e.
a wide range of signal frequencies which add together to make up the
pulse shape. If the bandwidth is broad enough, the different frequency
components travel at different speeds, and the shape of the pulse gets
distorted. Travel too far along and the pulse gets smeared-out.

Imagine a 0 and a 1 travelling down a wire. If the dispersion is too
high, the 1 gets broader and broader until it overlaps the 0. At the
other end of the wire, there's no way to tell which was the 1 and which
was the 0.

Plus electricity doesn't travel at the speed of light. I believe the electromotive thingummyjigger does, but the electrons themselves can't.

<DIV class=quote>
<DIV class=quotetitle>? quoting willow</DIV>
<DIV class=quotetext>It's funny, that no one mentioned the whole problem with these "tests" is one chips is a celeron while other is a P4.
</DIV></DIV>
Yeah. I've got 2 PCs...one a Celeron 2.4, the other a P4 2.4. Otherwise, the bits inside are identical. I run the P4 in dx9 mode and get roughly the same FPS as the Celeron which is in dx8 mode. I appreciate that software, XP setup, etc may have a bearing but that's quite a marked difference.

Plus electricity doesn't travel at the speed of light. I believe the electromotive thingummyjigger does, but the electrons themselves can't.

Force?

And besides, processor speeds are not even close to being
limited by the speed of light. Methinks dred_furst isn't very clear
about the topics he's talking about.

We are nowhere near close the theoretical limitations of the current
architectures. TB is dead on when he says heat dissipation is a bigger
issue. That, and also the fact that semiconductors do not behave too
predictably at the frequencies we have reached.

Edit: Semiconductor junctions, to be more accurate.

More edits:

yeah Celerons are basically P4's that were rejects and are
missing some math instruction sets.

Jinx, you have any links explaining this some more? AFAIK the
instruction sets supported by P4 williamette, P4 Northwood and
the celeron are identical (MMX, SSE, SSE2). And they can most
definitely not be missing any of the core instructions or the built in
arithmetic co-proc instructions. That's just silly.

Some more edits:

Willamette,
Northwood
(improved 0.13micron version.),
Celeron
(Willamette core, only difference is L2 cache). Most detailed reference
I could find.

Plus electricity doesn't travel at the speed of light. I believe the electromotive thingummyjigger does, but the electrons themselves can't.

Force?

And besides, processor speeds are not even close to being
limited by the speed of light. Methinks dred_furst isn't very clear
about the topics he's talking about.

We are nowhere near close the theoretical limitations of the current
architectures. TB is dead on when he says heat dissipation is a bigger
issue. That, and also the fact that semiconductors do not behave too
predictably at the frequencies we have reached.

Edit: Semiconductor junctions, to be more accurate.

I must admit i tried to explain something that my dad explained to me,
that could be false anyway, so i could well be talking rubbish :biggrin:
anything I say (oh wait, im going round in circles, time for me to shut
up)

back on with something that isnt high-level physics:

Yes, i guess P4's are faster that Celerons.

the wavelength of light is "reduced" when the hz of the changes size.
elecricity travels at the speed of light, and so has the same maximum
speed. the length/width of the current ATX boards are equal to this
"reduced" wavelength. therefore any higher frequency means the same
track can have both a 1 and a 0 on the same track.

Happy understanding!

electricity doesn't travel at the speed of light, only the effect of
electricity is instantaneous. electrons in a wire only travel a few
centimetres in an hour through a wire.

<DIV class=quote>
<DIV class=quotetitle>? quoting Leperous</DIV>
<DIV class=quotetext>Plus electricity doesn't travel at the speed of light. I believe the electromotive thingummyjigger does, but the electrons themselves can't.</DIV></DIV>

Electric SIGNALS travel at the speed of light. Not electrons. It's like turning on a faucet. The water comes out immediately, but that does not mean the water traveled all the way from the well in that fraction of a second.

Rof said:
Now if you have a very brief pulse, it requires lots of bandwidth, i.e. a wide range of signal frequencies which add together to make up the pulse shape. If the bandwidth is broad enough, the different frequency components travel at different speeds, and the shape of the pulse gets distorted. Travel too far along and the pulse gets smeared-out.

Imagine a 0 and a 1 travelling down a wire. If the dispersion is too high, the 1 gets broader and broader until it overlaps the 0. At the other end of the wire, there's no way to tell which was the 1 and which was the 0.

That is an interesting implication of the uncertainty principle I had never considered, but does it apply to electric signals? It seems more applicable to optical computing or fiber-optics. Surely even modern electronics can be treated classicaly. I kind of doubt they could have continued to build faster computers as quickly as they have if quantum effects such as this were becoming prominant.

At high enough frequencies (>100 MHz or so), electrical signals can
be usefully thought of as electromagnetic waves travelling down a
waveguide (the wire). In fact, you can even treat simple DC circuits
like that if you want to, though there's no point in doing so.

Dispersion is a classical effect, and it and the idea of short-duration
signals requiring a wide bandwidth works for any wave (water waves,
sound waves, light waves, radio waves, etc.). However, it is indeed
related to the uncertainty principle, when you get down to it.

The wavelength of light has no bearing on it's velocity.

This is only true in free space. Once you have waves propagating in a
medium, like light in glass, or RF signals through copper, you get an
effect called dispersion; different wavelengths (frequencies) of the
signal travel at different speeds.

For example, blue light is slower than red light when it's travelling
through most transparent media. That's why a prism splits up light into
different colours.

Now if you have a very brief pulse, it requires lots of bandwidth, i.e.
a wide range of signal frequencies which add together to make up the
pulse shape. If the bandwidth is broad enough, the different frequency
components travel at different speeds, and the shape of the pulse gets
distorted. Travel too far along and the pulse gets smeared-out.

Imagine a 0 and a 1 travelling down a wire. If the dispersion is too
high, the 1 gets broader and broader until it overlaps the 0. At the
other end of the wire, there's no way to tell which was the 1 and which
was the 0.

The underlined word: Do you mean distortion?

Dispersion seems to indicate an even spreading out of different
frequencies. This is, in general, untrue for any communication medium using electrical signals.
I don't think it is possible to predict, accurately, how much a signal
of a given frequency will be distorted (phase-shifted, attenuated,
etc).

I thought dispersion was only an issue in optical media.

Nope, I mean dispersion. At high enough frequencies, it becomes an issue for electrical signals, too. I'm not sure it's the limiting factor in the case of CPUs, but it will be an issue for ~>1 GHz signals.

I used to do research on high-frequency modulators for optical
telecoms; not only did we have to worry about optical dispersion in the
fibre, but also dispersion of the electrical signals driving the
modulator. Co-ax cable suitable for 20GHz signals is a right pain to
use, it's more like plumbing than anything.

Cool. I'm always happy to learn something new. I really wish I had been able to take a waves and optics class while I was in school. I know more about quantum mechanics (which isn't very much) than classical waves, although I suppose there are many parallels as this topic points out.

never fails, you guys do this to my threads every time :biggrin:

just once, i would like to be able to follow it from beginning to end. actually, the only time i see the end of a thread and be sure i fully comprehend its contents is when lep locks it after my first post.

you kids are way to smart to associate with dumb people like me. :/

Nope, I mean dispersion. At high enough frequencies, it becomes an issue for electrical signals, too. I'm not sure it's the limiting factor in the case of CPUs, but it will be an issue for ~>1 GHz signals.

Something new each day :smile: I hadn't heard that term applied to
electrical signals in my course work before. A bit of googling tells me
that delay distortion is also called dispersion, so I was wrong.

never fails, you guys do this to my threads every time :biggrin:
just once, i would like to be able to follow it from beginning to end. actually, the only time i see the end of a thread and be sure i fully comprehend its contents is when lep locks it after my first post.
you kids are way to smart to associate with dumb people like me. :/

Don't worry Orph, when the Snarkpit Science Club start it sends me running for cover too :razz: I'm going to go and have a game of marbles if you fancy it :biggrin: