Intel taking backward step...

[Reverand]

Apparantly, they have had enough with all this powerful processor stuff, and instead are going to work on making really small, really cheap, 1ghz processors, so you can put tons of these little processors in the slot where previously only 1 processor could fit.

I don't really see the point in this because of course the price of a 3.8 ghz northwood would probably be the same as 4 of these 1 ghz processors. Also, to handle a job together, they will have to split the task up into lots of little bits - and I seriously doubt that this is possible.

Unless you like to run 8 programs at once, I don't ever see this as being an advantage at all. More like a backward step for those of us who like to run one big program and have it all go in the processor and out again as a single task.

I am loathed to say it, but it looks like I'm going to have to go for AMD if I need some real computing power say 2 to 5 years down the line.

[Afn]

Cluster computing is faster than a single processor on a single computer. Supercomputers use multiple processors to achieve blinding speed.

If you had a 3 1 GHZ processors, you could have three independent decisions and one action.

If all you want to do is number crunch reality (super fast processing of information) then 16,32,128 or 256 1 GHZ processors on a single plug-in card or on the motherboard die could create some very realistic applications like human intellegence, realistic raytraced games, super fast applications. Using brute force computer processing with 256 1ghz processors on a single chip/mobo in a single PC would be awesome.

[Reverand]

Is it actually possible to split a memory process' tasks between multiple processors on a windows machine though?

If it is, and what you say is true, then this is very good news indeed. With expanded surface hard drives on the horizon, the advent of 1ghz memory, and various other progress in the world of computing, this will make supercomputing a household ability. (well, almost)

[shawners]

I doubt if it would be on windows operating software. I think the software in general will have designated processors to go to.

[Afn]

Is it actually possible to split a memory process' tasks between multiple processors on a windows machine though? 1.

If it is, and what you say is true, then this is very good news indeed. With expanded surface hard drives on the horizon, the advent of 1ghz memory, and various other progress in the world of computing, this will make supercomputing a household ability. 2. (well, almost) 3.

1. single cpu systems are going to become obsolete. If you use an innovative chip design in theory you could parellel process in the processor. If you could nanoscale 1 GHZ processors, and reduce the size by a factor of 8,16,32, You could have 32 times the processing power in the same 1 GHZ slot.


2. The internet is a supercomputer, not in the classic sense of a supercomputer, but combined resources on a global scale is "supercomputation" and by linking millions of computers together you get a "net" effect.

3. We know how to raytrace photorealistic images, now we need the hardware to process information much faster. If you can upgrade your 1 Ghz box and get x32 times a faster processor, using the same hardware, productivity would increase X32 times per pc.

I would love to see a processor motherboard design of four CPU slots with 256 1Ghz processors per slot in a single machine for about $1,000 dollars, and $250 to $500 dollars per additional processor.

My guess is that the 256 1 GHZ processor design would have to have memory on the processor die, in the processor to hold the operations of billions of flops. since the processor would be a single processor from the outside, but // process in the cpu, the result would be awesome performance. 4 chips (256 * 1ghz ) =1,024 times faster than a single 1GHz chip.

1. Motherboards that have 4 chip processor designs exist.
2. Nanotechnologies could be applied to allow the known 1Ghz chip design to be nanosized down.


It would be awesome if it could be done.

[Reverand]

I can see a day now when the need for memory is nonexistant, and all the processes spend all their time in the processor chipset. By the sound of this new technology, that may not be too far off.

If intel could make 2 ghz chips, of which you could fit around 10 in a single standard CPU slot, then that would be awesome. 20ghz of processing power certainly beats 3.8ghz!

And like I was saying, with terrabyte hard drive technology on the horizon, this will be one step closer to the perfect computer.

[Afn]

I can see a day now when the need for memory is nonexistant, and all the processes spend all their time in the processor chipset. By the sound of this new technology, that may not be too far off.

If intel could make 2 ghz chips, of which you could fit around 10 in a single standard CPU slot, then that would be awesome. 20ghz of processing power certainly beats 3.8ghz!

And like I was saying, with terrabyte hard drive technology on the horizon, this will be one step closer to the perfect computer.Greed and money.

4 computers at 1 GHZ is better than 1 machine at 4 GHZ.

If PC's were designed to allow for 10 processor cores per PC, manufacturers would sell 9 less pc's. The power to have 10 processor slots on a single mother or daughterboard would create raw computational power to do very interesting computations that could mean photorealistic VR, accurate speech recogition and (perhaps) smart agents that do work, provided the software exists.

It is a no-brainer that in the future a single pc will be able to have cluster power. Chip prices fall all of the time.

If there was an innovative motherboard design that had 10 processor cores and ran windows, I would want to buy one.

[DampCold]

They are also working on a dual core chip, which is simular to the multiprocessor idea.

I believe Windows 2000 and 2003 servers utilize the advantages of MPS.

There are many motherboards that offer mulitple Intel processors. They can be expensive though. You could check out IBM to get an idea on how much a server would cost.

[Afn]

20ghz of processing power certainly beats 3.8ghz!
I remember when 20mhz copmuiters were being marketed as a great improvement.

Software drives use of computers. Chips are meaningless if the software can not use the new processor designs.

Hyperthreading may make for multiprocessor designs work better under windows. It will be some years before the multiprocessor design is standard on every home PC. It takes a billion dollars to create a chip fab factory. Any radical chip design will take many years to mainstream.

Sad, it would be cool to have 20 cpu's in a single box.

[tamarisk]

Distributed Computing not multiple processing is the future.

At least in business for firms "renting" processor cycles wil be the future.

Cause it will be easier for them to "rent" cpu cycles than build their own supercomputer.

[Afn]

Distributed Computing not multiple processing is the future.

At least in business for firms "renting" processor cycles wil be the future.

Cause it will be easier for them to "rent" cpu cycles than build their own supercomputer.
Applications will drive the design of faster computers. It is cheaper to rent cpu cycles on a million pc's for cancer, drug or materials research. Supercomputing on each desk or in each robot will create (hopefully) pc's that can parse human speech and carry out complex intelectual tasks.

In 10 years we will have all the computers we have today, plus 10 more years of hardware to work with.

[shawners]

I remember when pentium processors came out.. And people said it couldnt do the basic math problem or some sort of thing they claimed it was doing right. I remember seeing a nice 486 33 dx computer for 1200 dollars.

[Reverand]

I keep a 1 year old office server unit here in my study. Pretty beefy machine, my friend asked me to save it so he could take it in and do all his unix/linux niknaks on it.

Anyways, the thing has several processors in it, quite a lot of ram, and a considerable number of entry level ATA hard drives. It's excessive even by my overclocking standards. My 3.8ghz Northwood will run pretty much everything at blindingly fast speed, so this thing is most definately for heavy duty bulk file handling.

Just imagine that one day your common household computer may make this total beast of a machine look like a spectrum! It would turn the industry on its head!

[MegaMog]

speaking of overclocking the new chips they are makeing are locked
http://www.tomshardware.com/motherboard/20040619/

[Reverand]

speaking of overclocking the new chips they are makeing are locked
http://www.tomshardware.com/motherboard/20040619/

Yuck, they're using SDRAM. They should use RDRAM, purely because it's 8 times faster. I can't beleive they'd go to all this effort and then go cheap on the ram. Intel said they were going to license RDRAM for their motherboards and they haven't. Le bastarde.

[Psilaxs]

Cluster computing is faster than a single processor on a single computer. Supercomputers use multiple processors to achieve blinding speed.

If you had a 3 1 GHZ processors, you could have three independent decisions and one action.

If all you want to do is number crunch reality (super fast processing of information) then 16,32,128 or 256 1 GHZ processors on a single plug-in card or on the motherboard die could create some very realistic applications like human intellegence, realistic raytraced games, super fast applications. Using brute force computer processing with 256 1ghz processors on a single chip/mobo in a single PC would be awesome.


For number crunching yes, they are faster, but for real time applications they are not. Games are the perfect example, you will not get any performance gains (maybe minimal ones) from having multi processor set up. real time applications are to time sensitive, and parallel processors have an inherit lag between communications, plus 1 has to act as a master processor, while the others are slaves, resulting in a "Mother May I" type of processing environment.

[Reverand]

For number crunching yes, they are faster, but for real time applications they are not. Games are the perfect example, you will not get any performance gains (maybe minimal ones) from having multi processor set up. real time applications are to time sensitive, and parallel processors have an inherit lag between communications, plus 1 has to act as a master processor, while the others are slaves, resulting in a "Mother May I" type of processing environment.

This is what I was thinking. They can't all collectively process one single thing. It's like compairing building a wall to building a model airplane - building a wall can be done by many people, since each individual can focus all his concentration on building his part of the wall - however building a model airplane can only be done by one person, since the kit demands your full attention to the entire project so you may manipulate it in the "bigger picture" as it were.

I agree with Psilaxs, in that I consider running a computer game to be just like building the airplane kit. However, compiling scripts and programs, or possibly rendering computer graphics IS more like building a wall, and multiple processors would probably do a better job in that area of computing.

So I think this may be AMD's possible future market - brute force computing on single chips, and Intel will opt for this cluster computing. Or intel may find a way to combine all these little individual chips into one single, very high power chip. Which I think is the route they are headed down, but we will have to see.

[tamarisk]

Programs ex games will have to be specially designed to exploit multiprocessor capabilities

[CCSDUDE]

Apparantly, they have had enough with all this powerful processor stuff, and instead are going to work on making really small, really cheap, 1ghz processors, so you can put tons of these little processors in the slot where previously only 1 processor could fit.

I don't really see the point in this because of course the price of a 3.8 ghz northwood would probably be the same as 4 of these 1 ghz processors. Also, to handle a job together, they will have to split the task up into lots of little bits - and I seriously doubt that this is possible.

Unless you like to run 8 programs at once, I don't ever see this as being an advantage at all. More like a backward step for those of us who like to run one big program and have it all go in the processor and out again as a single task.

I am loathed to say it, but it looks like I'm going to have to go for AMD if I need some real computing power say 2 to 5 years down the line.

Have you ever worked with multi-processor systems? 4 or 5 1ghz chips all working in unison on an OS that takes advantage....is much faster then the same mhz speed on 1 chip. Why? Because they're all doing work at once rather then one thing after another.

They know they can't kick AMD's ass in most done in cycles per split second so they choose to do the super computing thing.

Simple reasoning, eh? :)

[Reverand]

Have you ever worked with multi-processor systems? 4 or 5 1ghz chips all working in unison on an OS that takes advantage....is much faster then the same mhz speed on 1 chip. Why? Because they're all doing work at once rather then one thing after another.

They know they can't kick AMD's ass in most done in cycles per split second so they choose to do the super computing thing.

Simple reasoning, eh? :)

My computer = P4 3.8ghz, 1 gigabyte Samsung RIMM RambusDRAM 1066mhz, Seagate Barracuda 80 gig, ATi Radeon 9800XT with 256mb 400mhz DDR watercooled.

The entire system runs on water cooled heatsinks, and has been completely overclocked. It runs everything, and I mean EVERYTHING, you can throw at it. It's a silent beast.

Now, my friend approached me about last month I think it was, and asked for my advice on building good systems - so I told him, naturally, to build a system like mine. He said he wanted something a bit beefier, so I showed him a rambus compatible mobo which allowed him 4 (I think) gigabytes of RAM, and 2 processors. It cost him a ton, but he bought it.

His computer = Twin P4 3.8ghz, 4 gigabyte Samsung RIMM RambusDRAM 1066mhz, 2 Western Digital Caviar 7200RPM 250GB 8MB SATA OEM HDD, ATi Radeon 9800XT with 256mb 400mhz DDR watercooled.

He bought a load of watercooling kits and got nearly every single chip in the system sorted with cooling sinks. It's real POWER computing.

Obviously, he doesn't use it to play computer games. He does everything to do with 3d and 2d computer art, graphics, and animation. So naturally we never got a chance to compare systems for gaming abilities. So, I installed a game on the machine, then I started up my machine, ran the same game, and played it first on mine, then on his.

There was NO difference. Both computers ran the game at optimum efficiency. The extra resources of his computer were not being used, because they simply were not needed.

In fact, I estimate it will take roughly 2 years before the games become so advanced that they actually need that kind of power. But that's fine with me, it's why I invested so much money in such a good computer - I won't need to buy upgrades for a long time yet.

So it just goes to show, we may not need all this power for quite some time.

[CCSDUDE]

Rev, just because certain things not optimized for multi-processor systems ran the same stuff at a similar speeds means nothing. lol

Once they start coding games an such for these systems you'll notice either a huge bump in FPS or polygon counts and texture size. Basically they'll use one cpu for number cruching on real time wire framing and the other to push and bend textures, maybe a few more working out AI and more doing physics calculations to make the hair an grass an such move as it would in the real world.. Add to that GPU's are speeding up at a steady rate...we're gonna have true to life virtual reality soon enough.

[Psilaxs]

Rev, just because certain things not optimized for multi-processor systems ran the same stuff at a similar speeds means nothing. lol

Once they start coding games an such for these systems you'll notice either a huge bump in FPS or polygon counts and texture size. Basically they'll use one cpu for number cruching on real time wire framing and the other to push and bend textures, maybe a few more working out AI and more doing physics calculations to make the hair an grass an such move as it would in the real world.. Add to that GPU's are speeding up at a steady rate...we're gonna have true to life virtual reality soon enough.

We all know how multi processor games worked out on the sega saturn dont we?

Diferent era, yes, still have the same problems with tandum processing for real time application? YES.

[CCSDUDE]

We all know how multi processor games worked out on the sega saturn dont we?

Diferent era, yes, still have the same problems with tandum processing for real time application? YES.

lol true

I'm sure if enough egg heads work on it though...many geeks with be spanking it to virtual real time porn soon enough. Sexy Beach XXX, anyone?

[Afn]

We all know how multi processor games worked out on the sega saturn dont we?

Diferent era, yes, still have the same problems with tandum processing for real time application? YES.
Games have used multiprocessors for years, sound chips for sound, video processors for video and cpu's in routers to send ip packets. New hardware and software engineering is needed to create computers optimized for processing reality. (understand speech and perform real work from spoken or visual input.)

AI and true VR will need much more processing power. IBM has a goal in about 8 years to make voice recognition with near perfect parsing of real speech a reality.

Capacity or technique? A dumb design on utra fast computers is just as good as an optimized design on existing hardware. Someday, perfect voice recognition will be reduced to a single chip, visual recogition on a single chip with an array of cpu's to thread various thoughts and parse threads to make decisions. VR needs alot of processing power to do serious work.

How fast we get there is anyone's guess. I think it will happen much faster than we expect.

[Reverand]

Rev, just because certain things not optimized for multi-processor systems ran the same stuff at a similar speeds means nothing. lol

I don't like your tone.

The game was optimized for multi-processor systems. It has an option you choose at installation which asks you how many processors the computer has.

The point I was making, was that games do not need this sort of computing power yet. And probably won't need it for another 2 years. My computer runs probably every single PC game on the market as fast as it can go...

There is nothing wrong with multi-processor systems, what I'm saying is, for gaming it's just not needed yet.

[Afn]

There is nothing wrong with multi-processor systems, what I'm saying is, for gaming it's just not needed yet.
Playstation 3 is to be multiproccessor. So the games industry is moving to multiprocessor systems.

[Reverand]

Playstation 3 is to be multiproccessor. So the games industry is moving to multiprocessor systems.

Yes but playstation 3 is a consoletoy. :;)

[Afn]

Yes but playstation 3 is a consoletoy. :;)
consoles have to be better than home systems. Why buy a PSX if your PC does graphics better than a console. As for a console being a toy, a console is a dedicated COMPUTER machine for playing games.

When the sony oligarchy decides to put multiprocessors in it's game machines, it signifies a real change and a need to break away from single processor game machines.

[Reverand]

consoles have to be better than home systems. Why buy a PSX if your PC does graphics better than a console. As for a console being a toy, a console is a dedicated COMPUTER machine for playing games.

When the sony oligarchy decides to put multiprocessors in it's game machines, it signifies a real change and a need to break away from single processor game machines.

The reason I said that was because we are now likely to digress into a conversation about console v pc, and that is not what this discussion is about.

It should probably be noted that consoles are a completely different ball game when compared to PCs. The direction a console system may take in terms of hardware is not necessarily indicative of the direction PC systems may take, because PC games need to work on most systems, whereas console games are developed within the limitations of the console.

[Afn]

The direction a console system may take in terms of hardware is not necessarily indicative of the direction PC systems may take, because PC games need to work on most systems, whereas console games are developed within the limitations of the console.
The whole reason you buy a console is because of the hardware. If we just bought the games, then they would still be selling snes consoles and playing nes games.

As a general rule we do not play nes games, and there is no money in making snes carts.

[Reverand]

The whole reason you buy a console is because of the hardware. If we just bought the games, then they would still be selling snes consoles and playing nes games.

As a general rule we do not play nes games, and there is no money in making snes carts.

When a console company decides to make a new console, they stop making games for their old console, so in order to play new games, you must buy the new console. Which is really the only reason people buy them.

The only other reason is because they want to play the games which will be/have been released on that console.

So maybe PS3 will be using more than one processor, so what? Just because one console goes for multiple processors doesn't mean to say it's where the entire gaming community is headed. PC software is very different to console software, which is essentially why it takes so long for console to PC ports to be released, because a lot of re-engineering has to be done. So data processing systems which are effective for console games may be very ineffective for PC games.

This means that while PS3 may be able to reap the rewards that multiple processor systems provide, a PC is a very different matter due to the type of software it runs.

[Afn]

. So data processing systems which are effective for console games may be very ineffective for PC games.

This means that while PS3 may be able to reap the rewards that multiple processor systems provide, a PC is a very different matter due to the type of software it runs.windows was not designed for multiprocessor systems, there is one version of windows that does allow for multiple processors on some high end servers. As PC's need more processing overhead for VR and cognative processing, so will be the need for multiprocessor PC's and the software re-engineering to make the software work.

The good news is multiprocessor systems and software exist. More powerful mass produced hardware will produce software in time that exploits our need for computation.

[isus]

my lord. this whole thread is completely retarded.

apparently a few of you have never heard of the mhz myth... a processor's power is not made up solely by clockspeed.

[Psilaxs]

Games have used multiprocessors for years, sound chips for sound, video processors for video and cpu's in routers to send ip packets. New hardware and software engineering is needed to create computers optimized for processing reality. (understand speech and perform real work from spoken or visual input.)

Apples and oranges, different chips for different tasks, what you described is not parallel processing. 2 chips crunching the same data in real time (well, attempting to anyway) IS.

[Malakai1911]

Reverand - I only have a few points to make.

- RDRAM is dead. There is a reason Intel stopped using it and AMD never used it. It blows. Dual channel DDR400 2-2-2-5 T1 is just about as fast as you are gonna find ram (unless you move up to unsupported DDR466/533/whatever. But I'm talkin currently "standard" technologies here.) DDR2 has a little while before it starts to shine, give it time.

- P4 3.8 Is nothing to tout. Even the 3.4EE beats it in alot of benchmarks, and the FX-53 dominates all of them.

- 9800? Why not the X800.

At least you watercooled it.

[shawners]

where you been isus?? your missing the debate.

[Reverand]

Reverand - I only have a few points to make.

- RDRAM is dead. There is a reason Intel stopped using it and AMD never used it. It blows. Dual channel DDR400 2-2-2-5 T1 is just about as fast as you are gonna find ram (unless you move up to unsupported DDR466/533/whatever. But I'm talkin currently "standard" technologies here.) DDR2 has a little while before it starts to shine, give it time.

- P4 3.8 Is nothing to tout. Even the 3.4EE beats it in alot of benchmarks, and the FX-53 dominates all of them.

- 9800? Why not the X800.

At least you watercooled it.

Actually, you couldn't be more wrong. The most that DDR or any kind of SDRAM can manage is 533mhz. However, the fastest speeds that RDRAM can manage is with Samsung RIMM RDRAM, the worlds FASTEST RAM. It clocks in at a massive 1066mhz! That's twice as fast as the fastest DDR dual channel RAM, and don't try to tell me otherwise, because it's FACT.

If you really knew what you were talking about, then you'd know that rambus have been pushing to become the RAM standard for a while now. Several companies, who previously manufactured mostly DDR, have been looking at moving to RDRAM or creating competitive variants. In fact, Kingston, one of the world leaders in memory production, now sell 800mhz+ RDRAM.

RDRAM is definately not dead.

Also, the graphics card I use was at the time the worlds most powerful. It is NOT just the 9800, it is the 9800XT, which has the 256mb of ram. It smooths out graphics as much as they need to be smoothed out. The X800 is a good card, but compaired to the 9800XT, there is little improvement except for a slightly more powerful VPU.

[Malakai1911]

Alrighty, you couldn't be more wrong.

You are looking at MHz numbers, which is fine and dandy. Now lets look at data transmission rate over the bus, comparing DDR to RDRAM, shall we?

RDRAM 1066 is 1,066,000,000 * 2 = 2,132,000,000 or 2.13 GB/s

Paired in dual-channel mode (because RDRAM is dual-channel only, afaik), that doubles peak theoretical bandwith to a total of 4.26GB/s. Not too shabby.

Now let us take a look at DDR400, with an operating speed of 400MHz.

400,000,000 * 8 = 3,200,000,000 or 3.2GB/s

Paired in dual-channel mode, that doubles peak theoretical bandwith to a total of 6.4GB/s. Not too shabby.


Now lets compare the bigger number, 4.2GB/s or 6.4GB/s? Which one looks larger to you.


And btw, the X800 dominates the 9800XT, if you didnt notice. Particularly with high AA/AF settings, and in high resolutions with such settings.

[Afn]

And btw, the X800 dominates the 9800XT, if you didnt notice. Particularly with high AA/AF settings, and in high resolutions with such settings.
Faster ram speeds are meaningless if the software does not exist to run on the computer or game machine.

[Malakai1911]

Afn, did you fall down and bump your head?

All computers can benefit by using faster ram, otherwise PC100 would be just fine. Especially for popular things, you know, like gaming and video encoding.

[Reverand]

8mb databus, yes that would be useful if it actually worked.

SDRAM might as well be made of cockroach shit. That 8mb databus is completely useless if all the thing can manage is 400mhz.

But actually in case you haven't noticed, Samsung has already made 32 bit RIMM. So your calculations are way off. It's still a 4mb databus, which is half the databus of ShitDRAM, but considering that it's twice as fast anyway, that shouldn't, and doesn't even matter.

64 bit RIMM is only a little step around the corner, and when it comes, all our investments into being RDRAM compatible will have paid off, while you lesser SDRAM users sit there wishing you had opted for RDRAM compatibility.

We're already ahead of you in the mhz game, and we're catching up FAST with databus speed. So what are you gunna do? Sit and spin probably.

[CCSDUDE]

Alrighty, you couldn't be more wrong.

You are looking at MHz numbers, which is fine and dandy. Now lets look at data transmission rate over the bus, comparing DDR to RDRAM, shall we?

RDRAM 1066 is 1,066,000,000 * 2 = 2,132,000,000 or 2.13 GB/s

Paired in dual-channel mode (because RDRAM is dual-channel only, afaik), that doubles peak theoretical bandwith to a total of 4.26GB/s. Not too shabby.

Now let us take a look at DDR400, with an operating speed of 400MHz.

400,000,000 * 8 = 3,200,000,000 or 3.2GB/s

Paired in dual-channel mode, that doubles peak theoretical bandwith to a total of 6.4GB/s. Not too shabby.


Now lets compare the bigger number, 4.2GB/s or 6.4GB/s? Which one looks larger to you.


And btw, the X800 dominates the 9800XT, if you didnt notice. Particularly with high AA/AF settings, and in high resolutions with such settings.

HAHAHA....you took the words right out of my mouth. I run a pair of Corsiar DDR400 512mb dimms.....my bios allow me to tweak voltage as well as mhz in 1mhz steps.

I've pushed this stuff to the limit and beyond, it makes Rambus look like shit in dual channel mode.

[Reverand]

HAHAHA....you took the words right out of my mouth. I run a pair of Corsiar DDR400 512mb dimms.....my bios allow me to tweak voltage as well as mhz in 1mhz steps.

I've pushed this stuff to the limit and beyond, it makes Rambus look like shit in dual channel mode.

My system = RIMM4200 RDRAM * 32-bit (4 bytes) 1066 MHz 4.2 GB/s

Your system = DDR-400 SDRAM * 64-bit (8 bytes) 200 MHz x2 DDR 3.2 GB/s

Rambus is the way forward. You WILL be left in the dust by the time 64bit RIMM comes out, which is apparantly just about to come out!

[Malakai1911]

Now, I am by no means an expert, but where did you get "8mb" from.

Also, those calculations were by using a 32 bit data bus (sort of that "shows how much you know" kind of thing).

32 bit data bus, utilized by 2 memory chips. 32/2=16. 16 bits, divided by 8 = 2 bytes.
1066MHz x 2bytes = 2.13GB/s peak transfer rate.

It's pretty simple. RAMBUS moves FOUR TIMES LESS data THREE TIMES as fast. Quite an amazing feat of modern technology if I've ever seen it.

Lets use a benchmark to test my speed theory.
http://www.hothardware.com/viewarticle.cfm?page=3&articleid=531&cid=1

In particular "Athlon 64 FX-53, Memory Test @ 2.4GHz". It shows the demo system, using DDR400 benchmark at 5968MB/s (5.9GB/s, which isnt quite 6.4 peak theoretical, but close to it.)

Go ahead and benchmark your own monster machine with SiSoft 2004, I bet you'll be plesantly disappointed.

[Malakai1911]

You incorrectly state memory speed at 200MHz to try and discount the memory as being less than it actually is.

Lets take a look at your precious without using QDR signaling, shall we? Lets.

The following is untrue data, using a mathmatical scheme divised by an idiot. Please do not use it as a statement in which you base a critical decision on, such as the purchasing of a new computer.

1066 / 4 = 266.5MHz "true" speed. 266 * 2 = 532,000,000 or 0.53GB/s per module (1.06 GB/s in dual channel mode)

Thank you for bearing with me, I will now be returning to "factual mathmatics" mode

Edit: Btw ccsdude, you are one of the few people here I like.

[Reverand]

Malakai, watch this, I can do what you're doing too:

PC400 DDR 64 bit * 1000000000 = 13373r than thou.

RIMM divided by 2387492873487 = teh shitteh.

RIMM is better than DDR, face the facts. Now get the fuck back on topic. We're talking about multiple processors not your precious roach-shit SDRAM.

[CCSDUDE]

I've pushed it past 300mhz without a problem..... over volting does wonders for stability. You're also not calculating in that the 512mb pair is working in dual channel mode....

And you're also not calculating that my board happens to have 4 slots for a set of dual channels....

Try some standard benchmarks on P4 Rimm systems and AMD DDR systems....you'll no doubt see the Rimm get trounced at cas2 with the cpu and DDR bus over clocked along side upping the voltage.

[Malakai1911]

No Rev, what I'm doing is proving you are an idiot who cant follow simple multiplication. Why would anyone take advice from you?

[Reverand]

Obviously you did not catch me the first time round.

TAKE THIS DISCUSSION ON RAM TO A NEW THREAD PLEASE. This thread is dealing with processors, not RAM.

[CCSDUDE]

Ok.....so when you want to go off topic, it's totally cool, but when someone else wishes to do the same and are directly making you look bad.....it's all hands off?

Start a RAM thread right now if it's that important to prove you're right. ;)

[Reverand]

CCSDUDE, stop inciting arguments.

[Malakai1911]

I will reiterate.
Why should anyone take you seriously, Reverand.

[CCSDUDE]

CCSDUDE, stop inciting arguments.

You're kidding, right?

[cjules13]

CCSDUDE, stop inciting arguments.
lol Rev - take your own advice.

Stop being a sore loser and wanting to change the subject only when you are feeling the pressure. Last page all you wanted to do was talk about RAM, but when you got sprayed, now you want to run away and blame the inciting on CCSDUDE?

You are kidding right?

[Afn]

Afn, did you fall down and bump your head?

All computers can benefit by using faster ram, otherwise PC100 would be just fine. Especially for popular things, you know, like gaming and video encoding.
My point was that faster hardware is a moot point. It takes a combination of hardware, software and market to make a product work. We all know of products that had superior hardware and failed in the marketplace. Fast ram speed alone does not make a marketable product.


If we want a supercomputer cluster on our desktop, new chip and memory designs are needed. I think it is a waste of resources to place a single chip in a motherboard when a more intellegent open motherboard design could reduce less cpu and pc turnover.

Unless your wealthy and live with the latest technology, most of us are using machines and technology from the last 5 or so years. In my case the over-all effect of 6 computers to do work is more powerful than if I had a single brand new 3.8 Ghz P4.

[Malakai1911]

This isnt business 101, I've taken that course.

In fact, your faster hardware being a moot point is udder bullshit, considering there is a market for said product. Performance desktop hardware is marketed to enthusiasts, otherwise you wouldn't have Corsair XMS memory, or OCZ High Performance, Low Latency ram. If there weren't enthusiasts, you wouldn't have varying grades of high end desktop processor ranging from the new S939 3500+ to the FX-53 (or EE processors, for you Intel folk). If there weren't enthusiasts, there wouldn't be a need for peltier, liquid, or heatpipe cooling.

But the problem with your theory is that there are enthusiasts. And they aren't buying 6 shitty PC's to build one dream machine.

And even technology from this year is boiling down to the value segment, with the introduction of the nVidia nForce2 Ultra 400Gb, for about $200 I can have a motherboard with GbE, Integrated TCP Accelerated Firewall, 6 channel audio, RAID 0/1/0+1, SATA (all on the Epox 8RDA6+, to be released soon); Coupled with an Athlon XP.

And unless you are creating a renderfarm, I hardly doubt that clustered computing will make your desktop, internet surfing, or gaming experience any more enjoyable. At the current moment, almost no desktops need more than dual processors.

As you said yourself, software is needed for a product to work. The only case where 6 lesser computers will do as much work as a FX-53 or P4EE will be rendering highly complex scenework, or searching for extraterrestrial life maybe, if SETI supports that (i dunno). Maybe complex mathematical work also.

But for real life scenarios where enthusiasts are concerned? Dual processors are just fine right now.

[Afn]

As you said yourself, software is needed for a product to work. The only case where 6 lesser computers will do as much work
But for real life scenarios where enthusiasts are concerned?
I used to follow the cutting edge hardware. I would drool at faster processors and dream of owning black boxes that could improve my life. About 5 years ago I got wise to the marketing of technology and only focus on lower end, mainstream hardware.

Why? Most hardware is obsolete in a few years, hard to find parts for exotic configurations and cutting edge today is mainstream in 2 years and junk in a few years after that.

Without software hardware is worthless.

Can you model the human brain or create inteligent design with a single p4 cpu system?

You can not... well at least not yet. :)


Faster multiprocessor systems will be able to model complex systems like the human mind. I think it will be done with a supercomputing cluster on your desk.

TODAY's software is built for single proc systems.

Create a fast enough computer with some basic AI and speech to text interface that is accurate, and a self coding backend, I doubt that a P4 single chip system could do that kind of automation.

It will take a new generation of machines to get realistic VR, agent technology and cognative computers that can automate intellectual tasks, including self modification.

Why pay for software when the computer can self modify?

Single core cpu's or multiple core cpu's does not matter. What does matter is what the machine is doing something useful for you or your company. I think realistic agent technology will need supercomputer power in a desktop type configuration.

[Psilaxs]

I've pushed it past 300mhz without a problem..... over volting does wonders for stability. You're also not calculating in that the 512mb pair is working in dual channel mode....

And you're also not calculating that my board happens to have 4 slots for a set of dual channels....

Try some standard benchmarks on P4 Rimm systems and AMD DDR systems....you'll no doubt see the Rimm get trounced at cas2 with the cpu and DDR bus over clocked along side upping the voltage.


AMEN Brotha, My AMD system with Corsair XMS DDR400 absolutely screams. it will completey stomp most p4 systems in real world performance (though, intel owns the synthetic benchmarks :bk )