why not mb support for 2-8 cpu`s?

[slvr_phoenix]

i don`t mean to improve it this way, that is the point i don`t think sound can get much better that audigy 2zs quality (albeit extra channels) but rather to decrease the parts count, in the same way you said multi cpus couldn`t work because of high costs, intergration into the cpu would decrease the parts count and also the price.

But 1) Sound quality <i>can</i> get better than that.
2) The 3D effects engines of sound cards can definately improve.
3) This technology is already being built into motherboards anyway, which is a much better solution than integrating it into the CPU.

its funny that you mention that because a company just announced a physics chip, which the article alluded to becoming intergrated into the cpu sometime in the future and i imagine it could also be intergrated into the video card as you suggested.

Yes, I've heard about the physics chip. It's an interesting way to start, but I don't believe that it will ever be integrated into a PC's CPU. The only CPU that it will likely ever make its way into is that for a console platform.

is this why amd`s are faster?

**ROFL** Sorry, but no. AMD's CPUs are faster right now because Intel hasn't gone to SoI yet, so they're running into leakage problems that restrict how fast they can push their CPUs. In a loose way it's at best a partial explanation for why AMD's CPUs do more work per clock than an Intel P4.

you don`t think certain albeit possibly limiting functions could be set in hardware?

Do I think that it could be done? Sure. Do I think that anyone would benefit from it? Nope. The types of AI components that you're describing could easily be handled by basic programming. So not only would you not gain from developing this low-grade AI API into a CPU, but as I said, the rest of the CPU's performance would suffer slightly for even trying. It'd be a trade off that's just not worth making.

i guess this can all be written in the software

Exactly. Not only can it be written in the software, but it'd perform at the same speed as it would if you integrated some new logic.

i guess this is as bad as my idea i had for video cards; to have textures built into rom - another words have many megabytes of textures loaded into roms on the video card and have the programmers raytrace the scene and like paint by numbers have the video card apply the textures the programmer designated into those spots.

:\ Other than the fact that applications have to load their textures into the video card manually when the application needs them, this is how video cards already work. The loading sequence that you do this in before play starts isn't so bad. And likely not many developers would use any of the pre-loaded textures anyway as they each tailor their own for a reason.

One good idea i had years ago was to put music on chips and i wrote to a company online with this idea. My idea came about because i was annoyed with cd`s and that they skipped even though people said this wouldn`t happen that cd`s were perfect.
...
well tell me what you think i had to have written to them around 1991-92

I hate to break it to you, but that was far from an original idea. Memory-based MP3 players were just a natural progression of technology. And, as you noted, music files on computers were hardly a new concept even in '91. It was really just a matter of time before a specialized computer could be built affordably to hold music files that rivaled a CD player in size and quality. And before those existed, I knew people who were doing it with 386 laptops and low-quality WAV files. It wasn't until about the MP3 on a Pentium 133 era that this really worked all that well. Man was it slow to compress an MP3 on a P133 though. :O Still, it was worth it.

<pre>Antec Sonata 2x120mm
P4C 2.6
Asus P4P800Dlx
2x512MB CorsairXMS3200C2
Leadtek A6600GT TDH
RAID1 2xHitachi 60GB
BENQ 16X DVD+/-RW
Altec Lansing 251
NEC FE990 19"CRT</pre><p>

 

[juin]

A share bus is easier for multi socket mothersboards.


i need to change useur name.

 

[endyen]

Tim Sweeney from Unreal is trying to make you look like a genious <A HREF="http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2377&p=3" target="_new">http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2377&p=3</A>
Credit to Rugger for finding the article

 

[P4Man]

>i need to change useur name.

No, you need to get a <A HREF="http://www.eetasia.com/ARTICLES/2004JUN/B/2004JUN01_NTEK_ID_TA.PDF" target="_new">clue</A>.

>A share bus is easier for multi socket mothersboards.

I'm not sure if I'm more amazed by the arrogance or stupidity of this claim.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[arnold873]

nods and bows

 

[morbidangel]

wow it seems you were right after all arnold. Especially seeing the xbox 360 and ps3 have so many cores. That pheonix guy said it was impossible to run so many threads at once, i guess he has a lot to learn.

 

[mozzartusm]

Well ill just be damned. Your alive and still doing Meth!

ASUS P5WD2 Premium
Intel 3.73 EE @ 5.6Ghz validated.
Corsair XMS2 DDR2 8000UL @ DDR2 <font color=red>1068</font color=red> 5-3-2-9
SuperPI in 25 Seconds and a verified bandwidth of 9267 MB/s

 

[morbidangel]

only on the weekends

😉

 

[morbidangel]

I hate to break it to you, but that was far from an original idea. Memory-based MP3 players were just a natural progression of technology. And, as you noted, music files on computers were hardly a new concept even in '91. It was really just a matter of time before a specialized computer could be built affordably to hold music files that rivaled a CD player in size and quality. And before those existed, I knew people who were doing it with 386 laptops and low-quality WAV files. It wasn't until about the MP3 on a Pentium 133 era that this really worked all that well. Man was it slow to compress an MP3 on a P133 though. :O Still, it was worth it.

i originally had this idea during the time of the amiga 500
which had superior sound to anything pc`s had at the time.
And the amiga was only 7.4mhz. you can argue whether or not this was an original idea but i came up with it independently and probably a year after i wrote to diamond multimedia they came out with a mp3 player.

<P ID="edit"><FONT SIZE=-1><EM>Edited by morbidangel on 06/24/05 02:33 AM.</EM></FONT></P>

 

[Stimpy]

I have just had a thought were a parallel computing may well help in a game. Pre-calculating future events, i.e.
your character may have the choice of turning left or right in x number of frames time. Why not do the calculation of both possibilites up front in parallel, so when the decision is made the result is already ready, yes there would be redundent work done, but this is pretty much what we do with our brains (in a uber-massive way).
Off load any physics stuff that can be calculated in parallel to the AI, to multiple physics engines.
This is pretty much how a chess program works with much simpler rules.
My only concern with this type of model is that it would only work with massivly parallel systems, a simple 4 or 8 way would not be enough.

 

[P4Man]

sounds like an ingenious way to waste almost infinite processing power while bringing no performance advantage whatsoever.

You brought up the chess example yourself; even though there only 32 pieces on a chessboard, and the chessboard is only 64 squares big.. (compared to hundres of NPCs in a game, with a map that contains nearly limitless locations) with chess rules severly limiting the number of valid moves (like maybe 1 or 2 per piece on average, compared to nearly infinite number for a PC game), all in a completely linear way (no 2 pieces can move simultaneously, unlike in a game), yet the fastest supercomputers out there can only look ahead, what, 5 ? moves per second ?

>My only concern with this type of model is that it would
>only work with massivly parallel systems, a simple 4 or 8
>way would not be enough.

No kidding. My WAG is all processing power on the planet combined would not be sufficient to calculate every possible permutation on say, a typical Q3 map with 50 characters for furter than a single second ahead (assuming 60FPS).

Don't ever underestimate the power of 2. You know the story about the guy that supposedly invented chess, and his king wanted to reward him ? He only asked for a chess board with a single rice grain on the first square, 2 grains on the second one, 4 on the next, 8, 16, etc.

Now you can try and make a guess how many kg of rice that would end up to. Don't take a calculator, just use your gut feeling. The "correct" answer is posted below (spoiler).



































































-> it would be more rice than a cargo train that stretches from the earth to the moon can hold. At least that is what I recall ending up with when we did the exercise in high school a decade ago.
edit: just redid the math.
Assumed 1000 grains of rice to weigh 25 grams <A HREF="http://www.ricecrc.org/reader/rice-crc/tg_Size_and_Weight.htm" target="_new"> source </A>
Assumed a traincar to hold 50 tons, and have a length of 15m
Then you get:
On the last square there would be 2^64 grains of rice. On all the other squares combined, there would be an equal ammount (minus one) , so you get 2^65 -1 grains of rice:
3,68935E+19 Grains
9,22337E+17 gram
9,22337E+14 Kg
9,22337E+11 ton
18446744074 railcars
1229782938 m
1229782 km

Which is roughly 3x the distance earth-moon.

= The views stated herein are my personal views, and not necessarily the views of my wife. =<P ID="edit"><FONT SIZE=-1><EM>Edited by P4Man on 06/24/05 10:03 AM.</EM></FONT></P>

 

[Stimpy]

OK let make is a little simpler at any point where you have a choice (again left or right), you purly load up whatever information is required. I.e. pre-load all of the textures, maps, bots, etc, into memory ready for the time you make the decision. You could do this way upfront when you enter a room with two exists, buy the time you've crossed the room all that pre-loading and pre-calculation has been done. I would not advocate doing it more than one choice in advance. Yes memory would be wasted and CPU cyles (heat, etc,etc), but it could mean a much smoother game play.
Although I also like the idea of enemy characters being independent threads, with their own more advanced AI.
I know there are massive issues with interthread communication, but these are programming issed that are likley to be overcome at some point (next level of object orientation)

 

[P4Man]

>OK let make is a little simpler at any point where you have a
> choice (again left or right), you purly load up whatever
>information is required. I.e. pre-load all of the textures,
>maps, bots, etc, into memory ready for the time you make the
>decision

All of that is done already. Doesn't require massive parellel resources either, just enough RAM.

Besides, what you are suggesting simply makes no sense. The calculations being done know, would still have to be done, but in your theory would only serve to select the correct "state" (machine/cpu/thread/). You're just wasting everything else, and not bringing *any* benefit, the game would still be exactly as fast/slow assuming you could even switch states infinately fast.

Not only does it bring no benefit, it would be completely impossibly for anything remotely more complex than Pacman.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[slvr_phoenix]

On the last square there would be 2^64 grains of rice. On all the other squares combined, there would be an equal ammount (minus one) , so you get 2^65 -1 grains of rice:
3,68935E+19 Grains
9,22337E+17 gram
9,22337E+14 Kg
9,22337E+11 ton
18446744074 railcars
1229782938 m
1229782 km

Which is roughly 3x the distance earth-moon.

Wow! You came up with a <i>totally</i> different number than I did.

First off, it should be 2^<i>64</i> - 1, because this is a zero-based index. (2^1 = 2, 2^0 = 1)

That gives you 18,446,744,073,709,551,615 grains.

I used the actual grain weight unit (though I think that's based on something like a grain of barley, not rice) which is 64.79891 milligrams. So our big lot 'o rice weighs 1,195,349,015,976,379 kilograms, or 1,317,646,740,813.0993 US tons.

The average stat that I got for 100-ton grain hopper trian cars is 55 feet. At 13,176,467,409 cars, that would take about 137,254,869 miles worth of train cars.

A quick search placed the distance to the moon at approx. 236,000 miles. (Since this distance is constantly varying, accuracy must suffer.)

So the train (not counting engines) would have to stretch about 290 times to the moon <i>and back</i>.

Of course, if you use a weight of 25 grams per grain according to your link, then you get a weight of 18,446,744,073,709,552 kilograms, or 20,334,054,642,177.46 US tons, requiring 203,340,546,422 cars, at 55 feet each, for 11,183,730,053,210 feet, or 2,118,130,692 miles. So that's a train that stretches to the moon <i>and back</i> 4487 times.

😱 <font color=purple>یί∫υєг ρђœŋίχ</font color=purple> 😱
@ 191K -> 200,000 miles or bust!

 

[slvr_phoenix]

All of that is done already.

Could you imagine a game that <i>didn't</i> pre-load textures and maps? It'd be like the wait periods between HL2 zones, but a heck of a lot more frequently! Ouch.

Besides the fact that it's all completely impossible to pre-determine paths for free-moving objects in 3D space anyway. Even if you limit it to 2D space, you're still looking at 360 degrees of possible rotation at any given moment. Oh, okay, so maybe you could limit that to something like a 30 degree rotation per frame, but with accuracy involved that's still thousands of possible variations on a turn per frame.

Not only does it bring no benefit, it would be completely impossibly for anything remotely more complex than Pacman.

We really do agree on occasion. :O

😱 <font color=purple>یί∫υєг ρђœŋίχ</font color=purple> 😱
@ 191K -> 200,000 miles or bust!

 

[P4Man]

I did indeed make some errors; first I divided the number of cars by their length, instead of multiplying them.. oops. Then:

>First off, it should be 2^64 - 1, because this is a
>zero-based index. (2^1 = 2, 2^0 = 1)

Yes, you are correct, the first square will have 2^0 grains, the last square will contain 2^63. However, I combined all the grain on the chess board, not only the ammount on the last one, so the correct number is 2^0 + 2^61 +... + 2^63 which is equal to 2^64 - 1. I guess its safe to ignore that one rice grain

>I used the actual grain weight unit (though I think that's
>based on something like a grain of barley, not rice) which
>is 64.79891 milligrams

I used rice grain, not corn or whatever that weight unit is based upon. Which one is correct depends how you tell the story Since chess was most likely invented in the far East, india or china, using rice grains seems more appropriate as a reward for the inventor

>The average stat that I got for 100-ton grain hopper trian
>cars is 55 feet

Couldn't find any stats, so I just made a guess there.

>Of course, if you use a weight of 25 grams per grain
>according to your link,

25gr per grain ? ROFL, no per ONE THOUSAND grains ! You wouldn't expect 40 grains to weigh a Kg, would you ?

Anyway, lets use your numbers for the train, since I only guessed them, my numbers for the grain, since these seem correct, and the correct number of grains which is 2^64 ( -1).
So we have:
1,84467E+19 grains
4,61169E+17 gram (x25gram/1000grains)
4,61169E+11 tons (1 million grams in a ton)
4611686018 train cars of 100 ton each
84547577005 meters (used 55/3m per car)
84547577 km

Earth moon distance is roughly 385000 km, so its 220x the distance earth moon. Ok, I was off by two orders of magnitude, which isn't <b>that</b> bad in this context. Afer all my result would have been about correct if the chess board had only 57 squares

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[P4Man]

>We really do agree on occasion. :O

If you would be man enough to admit it when you're wrong, I don't see why agreeing couldn't be the norm.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[Starfishy]

NO!! Don't start being nice to eachother and agreeing and crap like that. What will the rest of us do instead of sitting here reading all your techno babble about who is right and who is wrong. Our lives will be empty. Think of the little people!!

 

[slvr_phoenix]

>First off, it should be 2^64 - 1, because this is a
>zero-based index. (2^1 = 2, 2^0 = 1)

Yes, you are correct, the first square will have 2^0 grains, the last square will contain 2^63. However, I combined all the grain on the chess board, not only the ammount on the last one, so the correct number is 2^0 + 2^61 +... + 2^63 which is equal to 2^64 - 1. I guess its safe to ignore that one rice grain

[confused]Umm ... as I said, "2^64 - 1".[/confused]

Since chess was most likely invented in the far East, india or china, using rice grains seems more appropriate as a reward for the inventor

I won't argue that any. Rice does make more sense. I just used the grain unit of weight because I could actually <i>find</i> that statistic. **ROFL** Until you provided your link to rice's weight I had nothing for an actual value for rice.

>The average stat that I got for 100-ton grain hopper trian
>cars is 55 feet

Couldn't find any stats, so I just made a guess there.

In truth, I couldn't find any <i>real</i> stats either. 🙁 You'd think they'd be easier to find than that. What I <i>did</i> find however were model train car stats and the model scale. I compared a few and decided that the figure should be close enough to the real thing.

>Of course, if you use a weight of 25 grams per grain
>according to your link,

25gr per grain ? ROFL, no per ONE THOUSAND grains ! You wouldn't expect 40 grains to weigh a Kg, would you ?

Well slap me silly and color me red. You're right. I have to admit having been a bit confused about that at the time. It seemed awfully heavy to me for rice, but I didn't go so far as to put it in the context of 1Kg. Geeze. That would be nuts. I didn't notice the asterisk footnote. Silly me.

So we have:
1,84467E+19 grains
4,61169E+17 gram (x25gram/1000grains)
4,61169E+11 tons (1 million grams in a ton)
4611686018 train cars of 100 ton each
84547577005 meters (used 55/3m per car)
84547577 km

Earth moon distance is roughly 385000 km, so its 220x the distance earth moon.

Okay, yeah. So adjusting my rice weight yields 461,168,601,842,738 Kg, or 508,351,366,054 US tons for the lot. That means 5,083,513,661 cars for a length of 279,593,251,355 feet, or 52,953,267 miles. That makes 224 trips one way, or to my previous scale, the train would have to stretch 112 times there and back. So now even though I'm using mostly US units we've got approximately the same results. (I'm guessing that most of the variation comes from rounding and the distance to the moon that we each used.)

I guess that means at this point we've validated each other's results and now we can safely say that'd have to be a damn long train! **ROFL**

Of course if one really wanted to throw a spanner into the gears, one could argue that a semi-wise king would have argued that the number of grains of rice could actually fit into a chessboard square be a limit. 😉 (Note, not a wise king, because a wise king wouldn't have fallen for it in the first place.) But this, of course, would be no fun.

😱 <font color=purple>یί∫υєг ρђœŋίχ</font color=purple> 😱
@ 191K -> 200,000 miles or bust!

 

[Starfishy]

Ok, if you guys are going to agree on this one that's fine, but you guys had better put those gloves back on and get back into that other thread with arms flailing.

 

[HansGruber]

4-8 way systems are really expensive even if you can have CPU's for free.


<font color=red>"We can't solve problems by using the same kind of thinking we used when we created them."
- Albert Einstein</font color=red>

 

[slvr_phoenix]

If you would be man enough to admit it when you're wrong, I don't see why agreeing couldn't be the norm.

No offense, but in case you haven't noticed, when I <i>am</i> wrong I <i>do</i> freely admit it. There are examples of such scattered about the threads here at THGC.

I have no false pride. As a computer programmer and an author I kind of had to lose that since I'm pretty much wrong daily. **ROFL** (Bugs, typos, code peer reviews, critiquing, etc.) The difference is, when (as far as I know) I <i>am</i> right, or when it's just a difference of opinion, then I don't back down.

😱 <font color=purple>یί∫υєг ρђœŋίχ</font color=purple> 😱
@ 191K -> 200,000 miles or bust!

 

[slvr_phoenix]

Ok, if you guys are going to agree on this one that's fine, but you guys had better put those gloves back on and get back into that other thread with arms flailing.

**ROFL** I've left that thread. I've got way too many other things going on in my life right now to continue investing energies into something that unimportant. Heck, if it weren't for the fair stretches of time that it takes to test my code at the moment I wouldn't even be posting. :O

It does kind of make me miss the old days of THGC though when I absolutely avoided the CPU section for months because of all of the absurd arguments and flaming. It wasn't worth posting then, but on occasion it was fun to read.

😱 <font color=purple>یί∫υєг ρђœŋίχ</font color=purple> 😱
@ 191K -> 200,000 miles or bust!

 

[P4Man]

>No offense, but in case you haven't noticed, when I am wrong
>I do freely admit it.

Unless you previously spent 5 or more long posts digging a hole for yourself, in which case, no, you won't.

0.9r still equals 1 and the branch misprediction penalty of a netburst core still is at least 19 cycles, no matter how hard you continue to spin fables around it. And btw, neither of them has anything to do with opinions, both of them are well documented, uncontentested, simple plain facts.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[slvr_phoenix]

both of them are well documented, uncontentested, simple plain facts

I've already made my points on where you're wrong on that and as far as I'm concerned both conversations are pointless to continue at this point, so I'm not going to continue either of those conversations here any more than I am there.

😱 <font color=purple>یί∫υєг ρђœŋίχ</font color=purple> 😱
@ 191K -> 200,000 miles or bust!