Java Cafe / Viva Computers Warning

[ShockG]

You seem to misread people's posts which then leads to unnecessary arguments when instead we could be coming up with solutions. And just fyi I have been overclocking cpus since the 386's and 486's which I ran at 150% of their rated speed if I remember correctly without any problems and with no cooling before overclocking them was even mentioned in any magazines I saw so I think I have earned the right to call myself a real overclocker despite what you may think. Just because I can't afford to get every cpu there is doesn't make me a noob as you seem to think.

hmm, 150% you say.
SLowest 386SX I knew and was home PC was 16MHz. so at the least yo'd have been running it at 40MHz no? So what did you use crystal? changed FSB to what? Since you couldn't change muliplier back then on any available board.
Alas, lets assume you had the 486 SX and maybe the 487-co-cpu you bought for your birthday.
Lets say you had the 25MHz. +150% so you ran it at 75MHz ala DX4 speeds?
Nah that can't be. All 486SX supporiting motherbards only had Dip-switches enough to select the available cpu speeds only. at that time there wasn't even a DX2 so the only 486 PCU's that were areound were the 25/33MHz. So the muli's you could pick would have ben 2x and 1.5X. Maybe you could use FSB, but you cold not easily change FSB, but mayeb you could select or add 8MHz from the 16MHz bus. So that would put you at 48MHz tops (2x24MHz).
But seeing as that the motherboards were ISA/EISA maybe VLB slot motherboards and had no dividers for the bus\expansion slot\CPU internal clock. You would have had to run the ISA/EISA bus at 24MHz!
Maybe I'm wrong no?

486 SX came with a heatsink, despite the previous 386 not needing one. So with your +150% OC you didn't use a heatsink?

Okay, You really are the master!
=======================================

Dude I never said yo're not an overclocker, Clearly by having internet access and knowing how to find OC-inside makes you top tier overclocker. I mean,
how would I know. Despite being the first to run DryIce in the country on an M-AthlonXp 2500+ with a DFI board @ 3.57GHz, I am clearly just givig you grief because I'm mean spirited.
proof:
http://www.systemshock.co.za/forums/index.php?showtopic=559#

 

[me]

Wow ShockG, I hope you didn't do allllllllll that typing for just this little thread!

My opinion is still the same because my experiences are still the same.

The latest nVidia IDE driver has a bug when applied to 430-based boards. Windows will not autosuspend into S3, and returning from forced suspend locks up the CPU with interrupts.

I also hear the firewall software nVidia-chipset NICs come with is full of crap.

Forget Hollywood -- if I was a home-market PC manufacturer I'd stay away from nVidia.

 

[me]

Bus? the Barton AthlonXP's operated on 400MHz bus including the Barton AXP 3200+ (11x200) and the late and very short lived AXP T-BredB 3000 (10.5x200)

Isn't that 200MHz, double-pumped to effective 400? Prometheus mentioned overclocking by raising HT bus speeds from 200 to 400MHz. I didn't know this was possible with modern RAM.

 

[ShockG]

Bah. Ha Ha!
Nah I didn't, those who know the G know that this stuff is all straight from the head. And it doesn't help that i usedto be a full on nvidiot gas head!

IDE driver is a driver bug and not a hardware one.
selling the 430 sweries is not nvidia's focus when it comes to mobos.
Yes they should work, but that's OEm/low end and that's exactly why they bought ULI.

NV gigabit NIC. I've had Zero problems with it since 2000 and never.

nvidia is a name all oem's want to get into their rigs. the logo is becoming a powerful drawing card. Sorry about your experiences, but out of all the chipsaets VIA has, how many do not have problems? None.

 

[ShockG]

Yes it is double pumped. Like I said 11x200 not 11x400
Yes you can raise HT speeds from 200 - 400, but yu'll need to use a divider as no DDR ram will make 400MHz. so you'd use 180/166/150/140/133/120/100 divider.

 

[hj2k_x]

talk about hijacking a thread

 

[Prometheus]

SLowest 386SX I knew and was home PC was 16MHz. so at the least yo'd have been running it at 40MHz no? So what did you use crystal? changed FSB to what? Since you couldn't change muliplier back then on any available board.
Alas, lets assume you had the 486 SX and maybe the 487-co-cpu you bought for your birthday.
Lets say you had the 25MHz. +150% so you ran it at 75MHz ala DX4 speeds?

Actually, according to my post that would be 16MHz -> 24MHz and 25MHz -> 37.5MHz.
16/100*150=24
25/100*150=37.5

16/100*250=40
which is something I again never stated. You might have mistaken 150% to be 150% above normal speed. 100% is normal speed where 250% is 150% above the normal speed as you used in your calculations. All you seem to be showing people is that you can't interpret posts.

 

[Beta]

....hj2k,More like dropping Heavy knowledge....best keep quiet and soak it up.

 

[mccrack]

those who know the G

When you start talking about yourself in 3rd person you got issues

 

[hj2k_x]

....hj2k,More like dropping Heavy knowledge....best keep quiet and soak it up.

* keeps quiet and starts soaking....

 

[ShockG]

which I ran at 150%

Can you blame me?

All you seem to be showing people is that you can't interpret posts

Maybe my maths is terrible, maybe I downright suck at mathematics...
Either way you still have not proved anything you have claimed at all. You're the one calling yourself a guru overclocker. Show and tell, I've posted my proof and I've shown I know that while I dont know the gritty about the AthlonXP, I know enough about it to prove that you were taking chances are have been caught.

You were the one coming of all "12yrs" of electronics. So show and tell.
Let us go point for point and see how much overclocker you are?

 

[ShockG]

Nah it's not issues McCrack its that I know that I used to come of as an nvidiot on other forums and still do sometimes. I could have easily said, those who know me know that I'm a die hard nvidia fan or used to be and I tend to read about them alot.

And the "G" is not the "G" for "ShockG"

 

[Prometheus]

Yes it is double pumped. Like I said 11x200 not 11x400
Yes you can raise HT speeds from 200 - 400, but yu'll need to use a divider as no DDR ram will make 400MHz. so you'd use 180/166/150/140/133/120/100 divider.

Unless you use the Max Memclock like on the ASUS A8V-E Deluxe which according to quote from BIOS:

Place an artificial memory clock limit on the system. Memory is prevented from running faster than this frequency.

So you can set your cpu to 400MHz which would give you effectively 800MHz and still run your memory on DDR200, 266, 333 or 400. I hear that the nForce actually has better memory adjustments. Too bad I don't have one.

 

[me]

Does doing that actually improve performance?

 

[Prometheus]

Does doing that actually improve performance?

Will increase your cpu's performance but could also fry it as it means double the amount of short-circuits and therefore double the heat. Then in order for it to work you must increase the core voltage which adds even more heat. Then your cpu also generates more heat the hotter it actually gets if I'm not mistaken. You'll need to use liquid nitrogen or a lot of peltiers and anti-freeze to run it sub-zero and even then you might not be able to get the heat away fast enough from the cpu core so it doesn't fry itself. So it will increase performance if you know what you're doing.

 

[Prometheus]

Either way you still have not proved anything you have claimed at all. You're the one calling yourself a guru overclocker.

I never said I was a guru overclocker, but I'm not the idiot you make me out to be. I'm more a Jack of all trades.

Show and tell, I've posted my proof and I've shown I know that while I dont know the gritty about the AthlonXP, I know enough about it to prove that you were taking chances are have been caught.

Show and tell what, not everything has to be documented in life in order for it to be true. I'm still wondering where I was caught out and with what. I never said I know how to overclock a A64. I said I was planning on doing it and I still am. How exactly was I caught out about anything?

You were the one coming of all "12yrs" of electronics. So show and tell.

Just did that previous post.

 

[ShockG]

Just did that previous post.

If this is what you're talking about then you're royally off the mark! Once again you've put your foot and entire body in it.

Will increase your cpu's performance but could also fry it as it means double the amount of short-circuits and therefore double the heat.

Then in order for it to work you must increase the core voltage which adds even more heat. Then your cpu also generates more heat the hotter it actually gets if I'm not mistaken. You'll need to use liquid nitrogen or a lot of peltiers and anti-freeze to run it sub-zero and even then you might not be able to get the heat away fast enough from the cpu core so it doesn't fry itself. So it will increase performance if you know what you're doing.

1. What short circuit?
WHat generates heat in a CPu are the electrons bumping against the carrying medium, ie silicon/copper traces. The higher the speed the faster the electron transfer and more collisions you'll get. when the electron (1 or 0) gets to the gate it may have changed value because of these collisions and that leads to errors lock ups or reboots.
How Vcore cirmumvents this, is it increseas the signal strength, that is every eletron assigned a 1 or 0, hence it keeps its value despite the collisions.
This is similar to drive strength in ram but not exactly the same.

Voltage however at the same time through the same mechanism increases the heat generated for the thermal dessipation, so overall temeprature will increase. Changing cooling to water\\dice\phase\cascade helps but it can't take you to an infinite clock speed.

There are no short circuits that take place unless there are water particles in the core causing them.

As for speed. MrMe increasing Ht speed doesn't help CPU performance at all. You increase it because the CPu clock is a muliplication of the HT clock and that also governs the ram clock so it is in your best interest to increase the HT speed (not the HT multiplier as that's different).
So you actually use the HT speed tio increase CPU and DRAM clock.

the Max Memclock like on the ASUS A8V-E Deluxe which according to quote from BIOS:

Yes like I said, its a memory divider like all nf4 boards have. Some only have 200/166/133/100, but other like the Asus Deluxe or DFI or Abit you can use the odd ram "dividers" like 180 (183* in A64 tweaker)/150/140 and 120.

 

[me]

Umm.... let's be precise here, electrons have only one charge and it's not 1 or 0. What arrives at the gates are voltage pulses carried by loads of slow-moving electrons. The heat comes from the electrons bumping into _each other_, since the carrying medium is made up of electrons itself.

I wondered about that concept of overclocking a CPU by increasing just its internal frequency and dividing frequency for the interface to RAM and the rest of the system. It would be wasting most of its new-found cycles waiting for data, since thanks to pipelining it can do a hell of a lot in a single cycle already (especially true for AMD's chips which run at lower effective frequencies but get the same amount of work done).

 

[ShockG]

The heat comes from the electrons bumping into _each other_, since the carrying medium is made up of electrons itself.

yes that's what I said.

WHat generates heat in a CPu are the electrons bumping against the carrying medium, ie silicon/copper traces

Indeed no single electron carries a 1 or 0 and it was wrong for me to say that.

You don't waste any cycles at all using a divider since on the A64 the controller is internal so the ram is directly connected to the CPU. The latencies and wait states take place inside the CPU and not on the northbridge. Since HT speed doesn't really make a difference in performance (You can see it when you get higher memory data-rate than the HT data rate actually allows.)

Waiting for Data would be from the Ram and not really connected to the HT speed or data-rate.
ie Intel Conroe\Yonah\Merom have lower memory data-rates even though they use DDR2 ram than A64 systems, but perform better. The delay is internal in the maths units/s and not entirely in the data getting to the CPU.

Much like the gfx cards of today. 7800GTX 512MB has higher memory data-rate than 7900GTX, but even at the same clock speed, the 7900 comes out faster. A fully optimized Processor would scale at a 1:1 ratio in performance. ie 1MHz on CPU is matched by 1Mhz in memory speed.
But this isn't the case.

 

[Prometheus]

Umm.... let's be precise here, electrons have only one charge and it's not 1 or 0. What arrives at the gates are voltage pulses carried by loads of slow-moving electrons. The heat comes from the electrons bumping into _each other_, since the carrying medium is made up of electrons itself.

Yes that is correct. Electrons don't have a value of 0 or 1 except perhaps in quantum processors where they actually have both values simultaneously and afaik all quantum processors work with atom spins rather than electron spins. Electrons have a charge. What causes the signal loss is a quantum effect based on the fact that an electron doesn't have a fixed location but rather an approximate location where it will be found. An electron can theoretically jump to any point in space thereby causing a higher or lower voltage at different locations in a conductor. This difference of the voltage is what is known as noise. More noise is introduced the higher the frequency of the circuit. To counter the noise the signal strength (aka voltage) gets upped. The noise is also increased but not so much as the signal strength so the ratio of noise over signal strength is lower and causes less errors. None of that however explains where the extra heat comes from when the frequency is increased or how the electrons start moving in the first place (more on that in part two).

I wondered about that concept of overclocking a CPU by increasing just its internal frequency and dividing frequency for the interface to RAM and the rest of the system. It would be wasting most of its new-found cycles waiting for data, since thanks to pipelining it can do a hell of a lot in a single cycle already (especially true for AMD's chips which run at lower effective frequencies but get the same amount of work done).

If increasing the cpu's internal speed didn't increase it's performance then it would be pointless to sell processors running on a higher frequency. Don't forget about the prefetch stages and the cache either. If these two things are effective like they are in modern processors then a cpu can do a lot of work before it becomes necessary to fetch more data from memory in order to continue execution.

Part Two:
CMOS (Complementary Metal Oxide Silicon), the technology used in the manufacturing of cpus and which is derived from MOSFETs (Metal Oxide Silicon Field Effect Transistor) consist of an N-channel and a P-channel FET connected in series. The input resistance of a MOSFET is often more than a million megaohms (1 000 000 000 000 ohms) and the current flowing into the gate is therefore negligible and even impossible to measure using standard test equipment. This extremely high input resistance gives them a very low current consumption at low operating speeds. The output stages of a CMOS gate consists of complementary (PMOS/NMOS) transistors acting as switches. When one of them is switched on it typically has a resistance of around 400 ohms and when it is switched on it has a resistance that is normally at least a thousand megaohms. The outputs have these circuit equivalents:

|---V+
|300R
|---Output (High)
|1000M
|---0V

|---V+
|1000M
|---Output (Low)
|300R
|---0V

In both cases the total resistance across the supply lines is extremely high and the current flow insignificant. So the current consumption should be negligible. But here enters reality. The first point is that any current flowing into a load connected at an output will be added to the normal current consumption. The second point is that a pulse of current gets consumed each time an output changes state. One transistor gets turned on to a significant degree before the other switches off properly. This brief short circuit between the two supply lines is what causes most of the current to be consumed and it's at this stage where most of the heat is produced. The faster the circuit operates at the more short circuits is produced and the more heat gets generated. It's also this ultra-high input resistance that makes them vulnerable to high static voltages.