please check my math correct me if wrong

dumbfish · Mar 14, 2004

from a different post

if this were possible.

say we had the same gfx cards.but say mine has a 1.8 ghz 333 barton.and if they made one she had a 2.7ghz 266 athlon.neither was overclocked .both ram settings were 2.5/2/2/2.we both have pc3200 ddr even thought her board is only 266max so it shows 266 when she boots up.
WHO IS FASTER?

-------------------------------------------------------
answer quoted from (The coolest)
2.7GHz.
The FSB matters but not as much as CPU speed. As simple as that.
FSB is just the speed that the northbridge can talk to the CPU, the CPU speed is HOW FAST data is actually being processed. Although higher FSB will make a machine run faster at a given CPU MHz, it won't make that much of a difference if you have a 1.5GHz and a 2GHz CPUs. In most if not all cases the 2Ghz will be faster.(end of quoted)
----------------------------------------------------

MY NEW QUESTION
my math to prove this:
1.8 / 333
1.8 on 32 bit process or 4 byte
1.8 x 1800mhz=7200 mb/s or 7.2 gb/s
64 bits (ddr) or 8 bytes x 333mhz=2664 mb/s or 2.6gb/s (why the say pc 2700)
to get avg thruput max
(7.2 gb/s + 2.6 gb/s ) /2 = AVERAGE 4.9

2.7/266
2.7 on 32 bit process or 4 byte
2.7 x 2700mhz=10800 mb/s or 10.8 gb/s
64 bits (ddr) or 8 bytes x 266=2128 mb/s or 2.1 gb/s (why its pc 2100)
to get avg thruput max
(10.8 gb/s + 2.1 gb/s) /2 =AVERAGE 6.4

The Coolest · Mar 14, 2004

huh?
PC 2100 means it can transfer data at about 2.1GB/s PC 2700 means 2.7GB/s.

Processors PROCESS data, they don't have a "Throughput". Only memory does. on 333FSB bus your processor can GET and SEND data at 2.7gb/s, at 266 it can GET and SEND it at 2,1gb/s, it will not make the calculation of the data itself any faster or slower.
meaning that with a higher FSB and a certain CPU speed the whole procedure will be faster, but the time it takes to process the data will be the same.

Gnufsh · Mar 14, 2004

dumbfish said:
from a different post

if this were possible.

say we had the same gfx cards.but say mine has a 1.8 ghz 333 barton.and if they made one she had a 2.7ghz 266 athlon.neither was overclocked .both ram settings were 2.5/2/2/2.we both have pc3200 ddr even thought her board is only 266max so it shows 266 when she boots up.
WHO IS FASTER?

MY NEW QUESTION
my math to prove this:
1.8 / 333
1.8 on 32 bit process or 4 byte
1.8 x 1800mhz=7200 mb/s or 7.2 gb/s
64 bits (ddr) or 8 bytes x 333mhz=2664 mb/s or 2.6gb/s (why the say pc 2700)
to get avg thruput max
(7.2 gb/s + 2.6 gb/s ) /2 = AVERAGE 4.9

2.7/266
2.7 on 32 bit process or 4 byte
2.7 x 2700mhz=10800 mb/s or 10.8 gb/s
64 bits (ddr) or 8 bytes x 266=2128 mb/s or 2.1 gb/s (why its pc 2100)
to get avg thruput max
(10.8 gb/s + 2.1 gb/s) /2 =AVERAGE 6.4

I can't understand why your applying calculations for the maximum bandwidth of a bus to the core speed of a CPU. What are you trying to do here?

I'm going to quote myself and hope it explains a little:

The speed of the CPU is generally much fater than the speed of the BUS. This is for several reasons. The first of which is that one operation may take several cpu clock cycles, even if it is on a small amount of data. This is the reasons for pipelines in the cpu, so that multiple instructions can be worked on at once, giving the appearence of executing the instructions faster. The second reason ius that some data is stored on the cpu in caches, and does not need to be fetched from main memory. I don't really have the time or expertise to go into a full explanation of how a microprocessor works and give you an exact answer to your question, so I'll just give you some good links:

arstechnica.com had an understanding the microprocessor series that delt with things like pipelining, superscalar execution, bandwidth and latentcy. I'll just post some links to the articles.
caching: http://arstechnica.com/paedia/c/caching/caching-1.html
multi-threading: http://arstechnica.com/paedia/h/hyperthreading/hyperthreading-1.html
bandwidth/latency: http://www.arstechnica.com/paedia/b/bandwidth-latency/bandwidth-latency-1.html
Understanding the microprocessor: http://arstechnica.com/paedia/c/cpu/part-1/cpu1-1.html
pipelining/superscalar execution: http://arstechnica.com/paedia/c/cpu/part-2/cpu2-1.html

and just for fun Moore's law: http://arstechnica.com/paedia/m/moore/moore-1.html

The ars article comparing the G4 and P4 talks about work per cycle I think, but I haven't read it in a long time:
http://arstechnica.com/cpu/01q2/p4andg4e/p4andg4e-1.html
http://arstechnica.com/cpu/01q4/p4andg4e2/p4andg4e2-1.html

Most of the other cpu theory and praxis articles are very good.

If you go to aceshardware.com and click on the cpu arch & tech link, you can find some very nice articles on the differences and performance of various cpus (some of the older articles from when the P4 first came out are very interesting).

http://www.chip-architect.com/ has some very in-depth exploration of the architectures of different chips.

dumbfish · Mar 14, 2004

so the only way to know what a rig should do is to put it together and bench it? there should be a way to predict it through math.

how can i compute the end result speed of a computer using specs of processor and bus?

The Coolest · Mar 14, 2004

The only way to know what a rig can do in what meaning? in performance? in what application?
You can look for benchmarks of different systems at different FSB and CPU speeds
you can also ask any of us to run some tests for you if you see a certain CPU speed/FSB speed that interests you.

dumbfish · Mar 14, 2004

according to www.numion.com my webpage loading thruput went from 800mbit avg to 1000mbit avg after i overclocked my gfx card.
this was interesting because at 3.5 mbit comcast cable isnt even 1mbyte /sec.so i always thought and still think cable is not "blazing fast".as "blazing fast" would mean my computer can't keep up.i would say that i shouldnt get any online improvement switching from a amd k-5 to athlon.
that is another subject altogether though.

what would be great would be if i could know what im buying before i buy it.looking at specs of cpu's and mobo's i would think i could know a systems basic speed.but when you have so many numbers to look at ...mhz cache fsb ....this has gotten to be a crazy time when you want the best (fastest) for you money.the worst part for me is trying to get this into a order of priority list figured out.i sure dont need to get a ata 133 drive over ata 100 if it doesnt have much impact on the bottom line.looking at the new 64bit cpu 1600bus makes me think what a deal .but after this thread i think but look its how many mhz? maybe i can stay with a 333 2500 barton.but to me its not about 64 vs the 2500.
its about knowing the difference.which i still cant grab a calculator and figure out.maybe i should just use the math i have at the top of this thread sure its not dead accurate but i need a point system.some way to get a basic idea here.

The Coolest · Mar 14, 2004

You can't use what you put up there as a real measure of a systems' performance.
tell us what you're gonna be using your computer for and what is your budget and we'll help you get the best bang for the buck for your needs.

dumbfish · Mar 14, 2004

your right cause now i have a different problem.
it awakens me at night.it makes me turn from naked women with perfect bodies.i just have this urge to OVERCLOCK.
and using the math i suggested has nothing to do with how overclockable something is. !!

filip04 · Mar 14, 2004

Gnufish[/i] [B] I can't understand why your applying calculations for the maximum bandwidth of a bus to the core speed of a CPU. What are you trying to do here? [/B][/QUOTE] [QUOTE][i]Originally posted by The Coolest said:
tell us what you're gonna be using your computer for and what is your budget and we'll help you get the best bang for the buck for your needs.

its like you guys can read my mind... this is EXACTLY what i was going to say... except, now i don't have to say it. thanx

Gautam · Mar 14, 2004

at night.it makes me turn from naked women with perfect bodies

Flip04's avatar.
Couldn't resist.

You can make rough calculations in regards to the IPC of a processor, and the bandwidth of the buses, and even factor in inefficiency, but all those numbers won't tell you much in the end.

redface · Mar 14, 2004

To Guatam:
What happened to your PC133Ram?

Will it be a similar case?
I used to have 512mb PC133 (2sticks), i forget to locked it to 133 and after i oced to 160FSB and ran 3Dmark01se for the whole nite. The next day I got a computer self-rebooting again and again. I found out that I have busted one of the stick.
Running a PC100 256 and a PC133 256 both AT 100mhz now....hahahahahaha

Gautam · Mar 14, 2004

I've got PC100 memory; this a laptop from 1999 if memory serves.

160MHz is quite a nice oc for PC133 memory, btw.

Lets try to stay on topic, though.

ChillPhatCat · Mar 14, 2004

dumbfish said:
from a different post

if this were possible.

say we had the same gfx cards.but say mine has a 1.8 ghz 333 barton.and if they made one she had a 2.7ghz 266 athlon.neither was overclocked .both ram settings were 2.5/2/2/2.we both have pc3200 ddr even thought her board is only 266max so it shows 266 when she boots up.
WHO IS FASTER?

MY NEW QUESTION
my math to prove this:
1.8 / 333
1.8 on 32 bit process or 4 byte
1.8 x 1800mhz=7200 mb/s or 7.2 gb/s
64 bits (ddr) or 8 bytes x 333mhz=2664 mb/s or 2.6gb/s (why the say pc 2700)
to get avg thruput max
(7.2 gb/s + 2.6 gb/s ) /2 = AVERAGE 4.9

2.7/266
2.7 on 32 bit process or 4 byte
2.7 x 2700mhz=10800 mb/s or 10.8 gb/s
64 bits (ddr) or 8 bytes x 266=2128 mb/s or 2.1 gb/s (why its pc 2100)
to get avg thruput max
(10.8 gb/s + 2.1 gb/s) /2 =AVERAGE 6.4

If neither computer is overclocked then you are talking about a 2700+ t-bred which is a far cry from 2.7 GHz... that processor has a clock speed of 2.17 GHz... so your calculation is way off... basically I'd expect both processors to perform about the same... the extra cache on the barton effectively gives it a boost equivelent to about 100 MHz and the 66MHz more on the FSB feeds data to the CPU faster so most applications will benefit from that...

It's really a combination of the different numbers... The bigger the gap between each component, the bigger the difference in performance... It's a combination of all those factors, so you just have to get a feel for what each number means... FSB and CPU speed do not interrelate.

dumbfish · Mar 14, 2004

but for the record i was talking about a pretend cpu. so i meant 2.7 ghz

if this were possible.

ChillPhatCat · Mar 15, 2004

In that case... If you paired a 2.7GHz 266 chip vs a 1.8GHz 333 chip, considering that both share the same architecture... the 2.7 would utterly destroy the 1.8... 333 vs 266 is negligeable...

This debate really is kinda pointless... the best way to get top performance is to find out where the top speed is of the chip and if you've got an unlocked chip, find the highest front side bus that the motherboard/memory will do... get the closest to that top speed with that front side bus through multiplier manipulation... that's the best performance you will find. if the front side busses are relatively close... like 20-30% then performance gains and losses will probably be unnoticeable.

If there is too much of a difference in CPU clock speeds... well the front side bus will not compensate adequately.

dumbfish · Mar 15, 2004

i have been thinking the 333 would win for so long its hard to think
the right way now.i thought that was why they got past 133mhz because of the "bottleneck" so i would think 2.7 would bottleneck all day unless it was on a 3ghz bus.the bus in my mind should be bigger than the chip.but after reading Gnufsh's post i guess the bus only processes once, while the chip can go thru the same crap again and again. so the chip should be sized what 5 times higher ? or 9 times?

ChillPhatCat · Mar 15, 2004

The bus does no processing... it simply brings data to the CPU to be calculated, then back to the memory when the calculation is done. I don't know how much clearer to make it... it's like a package being delivered to a UPS hub (the CPU) and the speed of the processor is equivelent to how fast the guy can load up a UPS truck and deliver it to your house (the RAM).

Once you get to around 2.5 GHz 266 FSB starts to become a bottleneck, which would be like the UPS truck coming to the Hub... and the guy unloading the truck can't unload it as fast as the other guy is loading the delivery truck.

There always is a theoretical bottleneck when the CPU runs on a multiplier... but in most cases you'll never see it hinder performance.

dumbfish · Mar 15, 2004

that was a nice story.and it was fun to read and helps me alot

jonspd · Mar 15, 2004

dumbfish said:
but for the record i was talking about a pretend cpu. so i meant 2.7 ghz

Whats so pretend about a 2.7ghz cpu? there is pleanty 2.7ghz stable amd chip's out there I'am sure. %100 sure.

dumbfish · Mar 15, 2004

how many run that stock with a 266 bus?

please check my math correct me if wrong

Member

Senior Member

Senior Member

Member

Senior Member

Member

Senior Member

Member

Member

Senior Benchmark Addict

Member

Senior Benchmark Addict

Member

Member

Member

Member

Member

Member

Super Speedy Senior

Member

Similar threads