Possibily stupid idea

S

Scintir

Limp Gawd
Joined
Jul 29, 2008
Messages
475
Okay, I possibly have a retarded idea on my hands, but this is something that has been bothering me for some time.

Why are GPUs on cards?

It seems that the CPU socket interface seems to be working just fine for them, while when CPUs were on a bus interface they were short lived. Why don't we today have a movement towards socket based GPUs?

Socket based GPUs will have the same concept of intergrated GPUs, in which the output is handled by the motherboard (motherboard provides the ports for monitors and other devices such as that to be plugged in). With this would then come in the problem of splitting video more, that I believe could be solved by cards that just duplicate the video and provide more ports (kind of like an add on card of a sort). Is there a reason I seem to be missing as to why there are no socket based GPU systems?

I could think of GPUs as becoming more sophisticated and similar to CPUs. Multiple cores on a single die and an ever evolving socket interface providing faster speeds. It seems to me that the socket GPU solution should have at least been experiment with, but it appears to be an idea that was thrown out or not even thought of. Now I know where are CPU and GPU intergration ideas, but this seems to me to be a waste, once again just use a GPU socket interface. If your GPU appears to be going slow, upgrade to a GPU with more cores on the die and get a faster one, like a CPU per say. Why is it that when we want to upgrade, we look at that slot for our expansion purposes. I don't know if it is just me, but I really thought the idea of a socket based GPU would be interesting. Bridges between the CPU and GPU could possibly be increased greatly as the interface of the GPU has the possibility to become even faster.

I do on the other hand know that there would probably be a large u haul of design to achieve this, but I think it if could be achieved, it could potentially be a good market move. That is of course, if there is nothing I am missing.

Now if everyone would, please at least glance at this and provide anything possible to help me get a better grip of this idea. I would like to have a nice discussion of this if possible. Now on the other hand, if it is completely retarded, please explain why and we can close this thread as fast as it was created :).
 
Bridges between the CPU and GPU could possibly be increased greatly as the interface of the GPU has the possibility to become even faster.
Click to expand...

With that reasoning, why don't we simply have on-die GPU controller on the CPU? Or we can even have on-die GPU ala (vapor now?) AMD's Fusion.

As for socketed GPU, it's a terrible idea for cost and graphic card size. Video card will need to be much larger because you'll need to attach a socket where GPU connects to the card. Also what about memory for the card? Will memory also be socketed? What about power delivery? How do you properly mix these requirements with different grades of GPU? Would you buy a low power socket GPU and a high-end board? Or a high power GPU with low-end board? It doesn't make sense cost-wise.
 
True cost can become a bitch along with other things. Cost wise I know this will be expensive, but as we can see, usually things that are expensive at the beginning will eventually come around to be at least somewhat affordable. Also, I was referring to the GPU being socketed on the motherboard, not onto a card. Next, memory can come in many different ideas, there can be many variance, but of course this will probably be costly as well. Power can also be routed to this socket as well I imagine, the route could carry more power than the standard route to a CPU, this would need to have a more complex design, but I am sure it can be done. Also, the mixing of different grades of GPUs for power can be somewhat represented by the idea that CPUs come in different power ratings as well (65W, 89W, 125W, 140W, etc....). Obstacles can all come down with a few ingenious ideas. Also, the idea that you would buy a high end board and a low power socket GPU combination will all depend on the situation. Some of us don't care about preformance and just want a comp to run. They can just buy a low power board with some low powered GPU designed for power saving along with being able to render what generally is desired with such low power requirements. Then there can be the enthusiast builds that provide a high end CPU socket, GPU socket, and overall a kick ass board with all the features to satisfy everyone. This board can thus be rated to allow more voltage to the GPU socket allowing more power to be pumped to it. There can be many things to get around problems, but I do agree cost will be an utter bitch until a way would be found to make this cheaper (overall probably wouldn't really come in much).

Now onto the on-die GPU controller on the CPU, yes there is the AMD fusion coming out that will do this, but can we honestly say that this will be a good thing? This will be basically for business class machines from what I see. A cheap and easy way to get both computation power and rendering power in a quick and easy package. This can span many other things.

Now comes to the memory part. That is one thing I was kind of stumped on, while the CPU uses the system RAM (cache is used but is a bit too small to do all task), so I was thinking about ways around this. The best thing I could say would be to use the RAM of the system, but that would of course overload the system a bit by putting so much on the RAM and then the speed problems, this could be a problem. Then this idea came to me. I see this every once in a while that there are add on cards for video RAM. now I understand that this may not be optimal, but I think this would be interesting. Have the GPU socketed as the main rendering object, and then use the current PCIE technology to hold RAM cards to add on to the GPU. I could possibly see the idea of the GPU controlling the PCIE bus with the RAM on it, this might all just be an idea that doesn't make sense since the GPU would still in the end use the bus, but who knows. The amount of RAM the GPU could use would then be able to be increase a large amount I believe, while the core can remain the same. While having a, lets say 4850 with 512MB of RAM on the PCIE slot, lets say you want to add more. Replace that with a new add on card providing a whole 2GB of RAM dedicated for the GPU's usage. This would allow RAM for GPUs to increase overwhelmingly, but cost may become a problem there as well. The GPU eating as much of that RAM out of the bus might become a problem along with many other things.

Now all of this falls down to theory. I have seen very few things that actually are comparable to this(although some ideas were brought from current uses of something remotely similar, but I cannot say they are that close). I just in my own mind think that this idea would be something awesome to at least experiment with, now for mass marketing, this probably won't be the best for cost, but I just think this idea would be something incredibly interesting to see how the things are developed and how they work.
 
Aw, I am sorry. I do massive post on accident, some flaw of mine that I cannot seem to rid myself of :(. Thanks for at least saying you will skip it :).
 
Multiple sockets with different pin layouts and attempting to upgrade = epic fail among with other problems.You would have to get a new motherboard when ever you wanted to upgrade either a CPU and/or GPU. There would be so many different socket combinations it would make your head spin.
 
Hehe, combination could become a problem. I doubt it would be enough to make people go insane, but I could see confusion on this. On the other hand you can have the same socket, with just different power ratings. Getting a socket with a high power capacity on it can allow people to upgrade a good distance. Then again things may not be that easy.

It appears that there is no real answer and just theories (what I was afraid of). If you guys wish we can just close this since all we will end up doing is arguing over theories on this ^^.
 
Okay, I'll buy your theory that you can socket the GPU chip onto the motherboard. Now consider this. CPU to northbridge (and thus, rest of the system) are currently about 32-64 bits wide. That's 4-8 lanes. With your GPU socket idea, we need to support at least 256-512 bits with current GPU spec (I have every reason to believe that it'll only get wider in future). That's 32 - 64 lanes. Your motherboard will have to support traces that are 8 times as wide. Imagine the size of that board... Power issue would also be borderline nightmare. You have to essentially build your motherboard to support the highest spec GPU, while keeping cost low enough to make sense for average Joe to pair it with lowest GPU. If you buy a low-spec board, you'll never even have a chance to move upwards. That doesn't make sense at all.

Your RAM thoughts are even more puzzling. You want to stick video RAM onto PCI-E because system RAM is slow? PCI-E at full bandwith (even with 2.0 spec) can barely supply fraction of what a mid-range GPU of today need. PCI-E is not some magical interconnect that can chuck 50GB/s. To implement what you want, you'll need to make the connection between GPU - RAM as wide as demanded by the GPU, so essentially, you'll need to draw the traces on the board itself separate from system bus. That means, you'll run into the same issue as said above. You could make the entire system bus wider than needed by GPU and have every component run off it, but imagine the width of that! Shit, we haven't even discussed about the issue of latency and resource control.

You need to take a class or two in computer hardware before wracking your brain over this. You're definitely overlooking some basic rules about hardware in order to come up with these outlandish theories.
 
You could use a motherboard based on light transmission but those haven't been tested outside of labs. Something like this (Google FTW):

Title:
[SIZE=+1] Signal transmission method and motherboard structure [/SIZE]

Document Type and Number:
United States Patent 6511234


Abstract:
The invention relates to a signal transmission method and a motherboard structure used in a base station of a radio system, for example, and comprises printed circuit board units communicating with the motherboard structure and communicating with one another by signals. The motherboard structure comprises an optical fibre sheet having optical fibres arranged therein, first converter means receiving from the printed circuit board unit electrical low-power and high-frequency signals, such as clock signals and RF signals. The first converter means convert the received signals into optical signals, and each optical signal is conducted into an optical fibre of its own in the fibre sheet. The motherboard structure further comprises second converter means converting the optical signals which have propagated in the fibre sheet back into electrical signals which are conducted to the printed circuit board units after the conversion.
Click to expand...
The only problem with that is we have no idea of how much bandwidth it can handle or how big the motherboard would need to be. Not to mention that the motherboard would likely cost as much as the rest of your components combined or likely more than that.
 
Yes, it is true that I haven't taken any computer hardware classes. I kind of can't in my current position (High school in a country type town in the middle of no where :eek:). This is a place I was hoping to get some information from. Anyways, another way around the RAM thing could be made. Without the use of cards, we can get rid of the PCIE 16X slots, thus making more room and allowing the bridges to have one less things to regulate. This could then open up some area for the north bridge and the motherboard to support the necessary things to support such a GPU. Things would need to be revamped a lot as I said, but I believe they could be done well if people actually planned it out.
 
Scintir, it's an interesting theory, but a bad idea. You're basically talking about reworking a few hundred (if not thousand, just guessing though) industry standards. It's expensive, impractical and would be less efficient, in some form or another.

It's not even worth the time theorizing about it.
 
No doubt there are advantages to the idea of placing a socket on the MOBO for a GPU, with local dimm slots for video ram.

There are also advantages for having a discrete unit that can be pluged into the MOBO at a right angle thus preserving space. Video cards are really a whole computer all by themselves now, processor, ram, power managment etc.

I dont see Nvdia, ATI, Matrox, Intel and all the other makers areeing on power, ram speeds/bit width, or any of all the other many details of a GPU.

I belive it would also increase the complexity and costs of mother board design far more then the benifits.

Now if you were in a closed system, like moterola/PPC Macs, things might be diferant.

I have noticed ove the years I need less and less cards. Sound, ethernet, IDE (SATA), USB, all those things used to need slots... I dont see high end video cards integrating any time soon.
 
Ten years ago, back when memory was expensive, there was talk of a GPU socket. The success of solutions like the PowerVR, which didn't even use local texture memory, pointed to a future where textures were streamed, and GPUs were replacable.

And then memory got cheaper, and the demand for higher resolution, high color depth textures destroyed the concept. Overnight, the size of memory on cards quadruped from 4MB to 16MB, and the types of memory and buses became far too varied: SDRAM, SGRAM, DDR, RDRAM, 64, 128 and 256-bit buses. Just trying to interface a graphics processor with all these varying memory architectures was a foolish waste, because performance was memory-limited thanks to the increasing importance of textures.

Today, we have add-in boards or integrated graphics, with nothing in-between. Eventually, when system memory becomes fast enough and the space becomes available to add the GPU onto a CPU die, we will see them become one. But until then, get used to the expansion card.
 
Hm, that is interesting. Thank you for your input :D. I guess I can be happy with the answers I have now :).
 
Another problem would by physical size, motherboards would be too big to fit most cases.
 
Motherboards and CPUs don't change that frequently except for clockspeeds. You can usually install a new CPU into a mobo that is several years old.. You might not be able to take advantage of all the latest features and the highest bus frequencies etc. but it will work.

I think it would be hard to do the same with GPUs..It seems the videocard (the "motherboard" of the GPU) changes as much as the GPU itself between generations.. From GDDR2 to GDDR3 to GDDR5.. from 256-bit to 512-bit and then back to 256-bit memory bus widths... There have been Ringbus controllers, Crossbar controllers, Switches etc.. and each new generation of videocards bring more changes. Really, what sets the 4870 apart from the 4850, except for clockspeed, is the type of memory used.

I'm afraid socketed GPUs would only mean you need to change mobos even more frequently to keep up. It's good to know that I'll most likely be able to use any videocard released in 2009 with my current mobo by just plugging into the PCI-E slot.
 
razor2050 said:
Okay, I'll buy your theory that you can socket the GPU chip onto the motherboard. Now consider this. CPU to northbridge (and thus, rest of the system) are currently about 32-64 bits wide. That's 4-8 lanes. With your GPU socket idea, we need to support at least 256-512 bits with current GPU spec (I have every reason to believe that it'll only get wider in future). That's 32 - 64 lanes. Your motherboard will have to support traces that are 8 times as wide. Imagine the size of that board... Power issue would also be borderline nightmare. You have to essentially build your motherboard to support the highest spec GPU, while keeping cost low enough to make sense for average Joe to pair it with lowest GPU. If you buy a low-spec board, you'll never even have a chance to move upwards. That doesn't make sense at all.

Your RAM thoughts are even more puzzling. You want to stick video RAM onto PCI-E because system RAM is slow? PCI-E at full bandwith (even with 2.0 spec) can barely supply fraction of what a mid-range GPU of today need. PCI-E is not some magical interconnect that can chuck 50GB/s. To implement what you want, you'll need to make the connection between GPU - RAM as wide as demanded by the GPU, so essentially, you'll need to draw the traces on the board itself separate from system bus. That means, you'll run into the same issue as said above. You could make the entire system bus wider than needed by GPU and have every component run off it, but imagine the width of that! Shit, we haven't even discussed about the issue of latency and resource control.

You need to take a class or two in computer hardware before wracking your brain over this. You're definitely overlooking some basic rules about hardware in order to come up with these outlandish theories.
Click to expand...


What this guy said, watch and learn.
 
razor2050 said:
Okay, I'll buy your theory that you can socket the GPU chip onto the motherboard. Now consider this. CPU to northbridge (and thus, rest of the system) are currently about 32-64 bits wide. That's 4-8 lanes. With your GPU socket idea, we need to support at least 256-512 bits with current GPU spec (I have every reason to believe that it'll only get wider in future). That's 32 - 64 lanes. Your motherboard will have to support traces that are 8 times as wide. Imagine the size of that board... Power issue would also be borderline nightmare. You have to essentially build your motherboard to support the highest spec GPU, while keeping cost low enough to make sense for average Joe to pair it with lowest GPU. If you buy a low-spec board, you'll never even have a chance to move upwards. That doesn't make sense at all.

Your RAM thoughts are even more puzzling. You want to stick video RAM onto PCI-E because system RAM is slow? PCI-E at full bandwith (even with 2.0 spec) can barely supply fraction of what a mid-range GPU of today need. PCI-E is not some magical interconnect that can chuck 50GB/s. To implement what you want, you'll need to make the connection between GPU - RAM as wide as demanded by the GPU, so essentially, you'll need to draw the traces on the board itself separate from system bus. That means, you'll run into the same issue as said above. You could make the entire system bus wider than needed by GPU and have every component run off it, but imagine the width of that! Shit, we haven't even discussed about the issue of latency and resource control.

You need to take a class or two in computer hardware before wracking your brain over this. You're definitely overlooking some basic rules about hardware in order to come up with these outlandish theories.
Click to expand...

You're comparing Apples to oranges when talking about bit-depth here. You compare the interface of a CPU and northbridge, and compare that to a GPU's interface with its memory. Take a Phenom processor for example, it has a memory interface of 2x 64-bit = 128-bit. Take a Geforce 8800GT or Radeon HD 48xx series, they have a comparable width of 256-bit. It is entirely inside the realm of possibility that a 256-bit wide memory bus can be integrated onto a motherboard's PCB.

The way I see it, the GPU "core" would have to be treated as another CPU, with its own memory interface. That would mean you'd have to buy a relatively small amount (512-1GB by today's standards) of very fast memory, and plug them into their own GPU-dedicated memory slots on the motherboard. I think it's an intruiging idea, but a standard (socket type, memory bus interface, etc) would have to be agreed upon by major GPU manufacturers.

Imagine being able to keep your GPU memory, but change between a 4870 core and a GTX 260 core just like changing CPU's. It is possible, but it would very likely make motherboards rather expensive. Additionally, one of the major things a socketed video card has going for it is that it is not constrained by the bus- and memory- technology of the past, as CPU's are. Each time (every few months) a new videocard comes out, the manufacturers have the option to use faster, newer memory bus types, without worrying about adhering to a motherboard standard; they just build the GPU around their new memory. By comparison, Intel cannot say "Oh by the way you need DDR4 with the E6550, and DDR5 with the E8400, and 512-bit DDR3 with the Q9650"... that would force customers to change motherboards and RAM every few months; a socketed motherboard design is a trade-off between upgradeability and "bleeding-edge-ability."

Edit: HOWEVER, do notice that many videocards have been using a tried-and-true 256-bit DDR3 memory standard for a very long time. This intruiges me. Let's say a motherboard has slots dedicated to accepting open-architecture GDDR3. This means you could upgrade your memory speed from a 7900GS-esque 1400mhz (and keep your 256-bit bus) to an 8800GT-esque 1800mhz independently of the GPU core. There could even be a way to allow for and unpgrader to increase bit-depth to 512 bit, akin to running dual-channel memory in motherboards today. It would require a lot of memory slots though.

Combine this with an AMD-like memory controller remaining on the GPU (so that each new GPU could have performance tweaks within a given base architecture, think Athlon X2->Phenom socket AM2 memory controller), and I think this could actually work.

And razor, not too many of us are hardware engineers (I know I'm not), and I highly doubt you are. So do keep the condescending banter under control...
 
Hehe, thanks for that :). I can say that I will hopefully become a computer engineer dedicated to hardware engineering in the future, but for now I am stuck in a shitty high school. Thank you for actually looking at the idea as plausible and stating good reasoning for it instead of just bashing it.

On your ideas, I was actually considering ideas today that are pretty much exactly as you stated. It is nice to see I am not alone in considering these ideas :D.
 
ordovician said:
You're comparing Apples to oranges when talking about bit-depth here. You compare the interface of a CPU and northbridge, and compare that to a GPU's interface with its memory. Take a Phenom processor for example, it has a memory interface of 2x 64-bit = 128-bit. Take a Geforce 8800GT or Radeon HD 48xx series, they have a comparable width of 256-bit. It is entirely inside the realm of possibility that a 256-bit wide memory bus can be integrated onto a motherboard's PCB.
Click to expand...
I don't think "dual-channel" means literally twice the traces. PCB traces can't be miniaturized via improved process. Creating wider traces between CPU-Memory/Northbridge and GPU-memory is prohibitively expensive . Also, you're assuming GPUs of future will not require wider memory interface, which is difficult to swallow.


ordovician said:
The way I see it, the GPU "core" would have to be treated as another CPU, with its own memory interface. That would mean you'd have to buy a relatively small amount (512-1GB by today's standards) of very fast memory, and plug them into their own GPU-dedicated memory slots on the motherboard. I think it's an intruiging idea, but a standard (socket type, memory bus interface, etc) would have to be agreed upon by major GPU manufacturers.
Click to expand...
Such agreement would be next to impossible, as the socket is entirely dependent upon the chip's architecture. Nvidia's just annouced an interesting technology that might make this memory issue entirely moot. Read up on Tesla.

ordovician said:
Edit: HOWEVER, do notice that many videocards have been using a tried-and-true 256-bit DDR3 memory standard for a very long time. This intruiges me. Let's say a motherboard has slots dedicated to accepting open-architecture GDDR3. This means you could upgrade your memory speed from a 7900GS-esque 1400mhz (and keep your 256-bit bus) to an 8800GT-esque 1800mhz independently of the GPU core. There could even be a way to allow for and unpgrader to increase bit-depth to 512 bit, akin to running dual-channel memory in motherboards today. It would require a lot of memory slots though.
Click to expand...
I don't think that'll work very well. Yes, GDDR3 has been quite common in recent GPU, but there are no grantee that it'll be defacto standard. Video memory has to be very fast and very efficient, so it must be custom tailored to the architecture that uses it. This isn't general-purpose memory like DDR2 that you can swap between many different platform.

ordovician said:
Combine this with an AMD-like memory controller remaining on the GPU (so that each new GPU could have performance tweaks within a given base architecture, think Athlon X2->Phenom socket AM2 memory controller), and I think this could actually work.
Click to expand...
AMD Fusion.... Though, Fusion is not a performance orientated GPU, rather a cheaper way of integrating video processing onto the die of CPU.

ordovician said:
And razor, not too many of us are hardware engineers (I know I'm not), and I highly doubt you are. So do keep the condescending banter under control...
Click to expand...
That wasn't my intention. I was merely suggesting Scintir to peruse some hardware articles ( technical tutorial/paper) before he continue arguing his point. His idea isn't that novel or new, I've seen it several times on various forums, and in the end, came to the same conclusion (won't really work now). It's refreshing to see a younger person (I'm not that old, btw) be interested in general architecture of computing devices instead of benchmark squabbling I see so often. However, one needs to learn how to search for information before continuing to argue to same point(s), after being rebutted. Sorry if I sound a bit cynical/jaded to you. That's what happens when you go to engineering school for 5 years.
 
May I ask you what school, major, and possibly your job if you got one that pertains to your major?
 
razor2050 said:
I don't think "dual-channel" means literally twice the traces. PCB traces can't be miniaturized via improved process. Creating wider traces between CPU-Memory/Northbridge and GPU-memory is prohibitively expensive . Also, you're assuming GPUs of future will not require wider memory interface, which is difficult to swallow.



Such agreement would be next to impossible, as the socket is entirely dependent upon the chip's architecture. Nvidia's just annouced an interesting technology that might make this memory issue entirely moot. Read up on Tesla.


I don't think that'll work very well. Yes, GDDR3 has been quite common in recent GPU, but there are no grantee that it'll be defacto standard. Video memory has to be very fast and very efficient, so it must be custom tailored to the architecture that uses it. This isn't general-purpose memory like DDR2 that you can swap between many different platform.


AMD Fusion.... Though, Fusion is not a performance orientated GPU, rather a cheaper way of integrating video processing onto the die of CPU.


That wasn't my intention. I was merely suggesting Scintir to peruse some hardware articles ( technical tutorial/paper) before he continue arguing his point. His idea isn't that novel or new, I've seen it several times on various forums, and in the end, came to the same conclusion (won't really work now). It's refreshing to see a younger person (I'm not that old, btw) be interested in general architecture of computing devices instead of benchmark squabbling I see so often. However, one needs to learn how to search for information before continuing to argue to same point(s), after being rebutted. Sorry if I sound a bit cynical/jaded to you. That's what happens when you go to engineering school for 5 years.
Click to expand...

All the "counter points" you made to me I had already admitted were problems with the idea. Like I said, there are issues with it. I'm not arguing for or against it, I'm just defending the validity and dignity of the OP's ideas. All I'm saying is that it is intriguing.

And yet for some reason you feel the need to compare dick sizes, arrogance maybe? I'm actually a second year graduate student of Geology with thesis work in Micropaleontology, Chronostratigraphy, and radiometric dating. Does that have anything to do with this? No. Does a mechanical engineer (or any other engineer besides a computer/electrical engineer who works for nvidia, ati, etc...) know any more about this than I do? No.

I'm so tired of people on here treating other's ideas like shit, just because they have some sort of sophomoric understanding of a subject.
 
Well we already know that the PCIE bus isn't really all that limiting since when given extra bandwidth cards do get faster, but not that much, so building the card into the motherboard wouldn't be all that necessary.

What about each video card generation coming with a base card which is really a mini video motherboard on which you install the core and memory of your choosing (back in the day some could handle add in memory).

That would be a bit of a mixed bag though. You would pay for support of extra features even if you weren't going to use them and pay a bit extra just for general compatibility being built in. On the plus side you would have a cheaper time upgrading from a low end chip to a high end and could purchase memory based on the resolution you plan to run. You could also possibly keep the board and/or memory between upgrades sometimes.

Something a little like this has happened before. A company sold a mother card that accepted nvidia laptop video cards when they were small and modular. You could install two of them and SLI them in a desktop. Something like that is even easier because the memory and core are already paired in small functional units. No reason that couldn't be expanded and have you add in more to get triple sli, quad sli etc... Imagine the super computers that could be built with to run cuda etc.. if you had a huge motherboard covered in sockets to accept tiny modular mobile gpus. In all honesty though, the companies would have to decide something like this could increase proffits before they would produce it. Since there are only two companies competing, neither is going to jump on something that would only appeal to hardware fanatics (small market demographic) and reduce the amount of money those people have to spend!
 
ordovician said:
And yet for some reason you feel the need to compare dick sizes, arrogance maybe? I'm actually a second year graduate student of Geology with thesis work in Micropaleontology, Chronostratigraphy, and radiometric dating. Does that have anything to do with this? No. Does a mechanical engineer (or any other engineer besides a computer/electrical engineer who works for nvidia, ati, etc...) know any more about this than I do? No.

I'm so tired of people on here treating other's ideas like shit, just because they have some sort of sophomoric understanding of a subject.
Click to expand...

I think you need to calm down and stop taking on issue that doesn't exist. I suggested the OP to take a class in computer hardware, so he may understand WHY his theory will have such issues, instead of going off in random directions trying to justify his idea. That's VERY unproductive. I was not waving my "dick" size. If you get pissed off at this conversation we're having, I don't see how you can handle yourself in professional/technical settings where you're constantly being challenged of your claim.

Scintir said:
May I ask you what school, major, and possibly your job if you got one that pertains to your major?
Click to expand...
I'm on my last semester of Computer Science at GAtech. My co-op job has been programming web services with databases, but I've taken several computer hardware classes as required by my major. If you're interested in this topic, I highly suggest you read up on some basic computer architecture. Real books/papers about designing logic, bus, and systems. Not some popular article on how to build "Uber-1337" PCs. If you can understand them and still have a passion for it, do well in your high-school, do some interesting personal projects with the topic (or form/join a club) to show initiative and apply to a college that's top ranked int the field.

Jakalwarrior said:
Something a little like this has happened before. A company sold a mother card that accepted nvidia laptop video cards when they were small and modular.
Click to expand...

I think you're thinking of "MXM" technology. Those were essentially standardized video boards that allowed you to upgrade you laptop with newer GPU that used the same board specs.
 
razor2050 said:
I think you're thinking of "MXM" technology. Those were essentially standardized video boards that allowed you to upgrade you laptop with newer GPU that used the same board specs.
Click to expand...

Yes but there was a card to allow inserting two of them into a mother card in a desktop system to run them in SLI and make your own X2 card. It was supposed to be carried on and able to accept the next gen of mini laptop cards too etc... but of course nvidia didnt carry the spec on.
 
razor2050 said:
I think you need to calm down and stop taking on issue that doesn't exist. I suggested the OP to take a class in computer hardware, so he may understand WHY his theory will have such issues, instead of going off in random directions trying to justify his idea. That's VERY unproductive. I was not waving my "dick" size. If you get pissed off at this conversation we're having, I don't see how you can handle yourself in professional/technical settings where you're constantly being challenged of your claim.
Click to expand...

Dude, I MADE NO CLAIMS. Re-read my original post. I discussed the pros and cons of the idea. Maybe you should take a literature class so you can learn your alphabet. I hope this thread gets locked.

The OP was obviously self conscious about an idea he had, as evidenced by the title of this post, and YOU PUT HIM DOWN FOR HIS IDEAS to make yourself feel smart. That's not right, and I won't stand for it.
 
You must log in or register to reply here.
Back
Top