There seems to be a new generation of graphics cards and CPUs to emerge this year. I'd be patient for this summer. You might either pay I-need-a-high-end-PC-right-now taxes or end up with a system that is fast but doesn't support some of the coming key features if you're getting impatient now.
Re: New motherboard
Posted by Crono on Tue Mar 20th at 8:29pm 2007
Posted by Crono on Tue Mar 20th at 8:29pm 2007
I didn't say build a machine right now. I was just showing that you shouldn't upgrade to crap and that, if you are indeed impatient, and need to upgrade now, for not much more you can build a decent system.
That's all.
That's all.
Blame it on Microsoft, God does.
Re: New motherboard
Posted by RedWood on Wed Mar 21st at 4:11am 2007
Thats exactly what i got out of what you said. Sorry if it sounded like i was trying to put words in your mouth.
I'll be waiting. Like Reaper said, I'd rather catch the beginning of a wave instead of the end of one.
Posted by RedWood on Wed Mar 21st at 4:11am 2007
? quote:
I didn't say build a machine right now. I was just showing that you shouldn't upgrade to crap and that, if you are indeed impatient, and need to upgrade now, for not much more you can build a decent system.
That's all.
That's all.
I'll be waiting. Like Reaper said, I'd rather catch the beginning of a wave instead of the end of one.
Re: New motherboard
Posted by Crono on Wed Mar 21st at 6:42am 2007
Posted by Crono on Wed Mar 21st at 6:42am 2007
Later this year, Nvidia is likely putting out the G90 cards (possibly SM5 specification as well). It may be another line entirely though.
I believe, ATI is planning on putting something major out by the beginning of 2008 as well (after whatever they're doing now)
I have a feeling that getting a GF8 card could be a premature investment considering this news (it's something I read, search around for some supporting information or something that completely contradicts it, if you wish.)
AMD is also putting out their quad core chips sometime this year.
This all means by the beginning of next year, GF8 cards will be at least as cheap as current GF7 cards and, hopefully, quad core chips will be down too, at the very least the X2 will be far cheaper than it already is.
At least they don't plan on switching sockets again (I hope)
I believe, ATI is planning on putting something major out by the beginning of 2008 as well (after whatever they're doing now)
I have a feeling that getting a GF8 card could be a premature investment considering this news (it's something I read, search around for some supporting information or something that completely contradicts it, if you wish.)
AMD is also putting out their quad core chips sometime this year.
This all means by the beginning of next year, GF8 cards will be at least as cheap as current GF7 cards and, hopefully, quad core chips will be down too, at the very least the X2 will be far cheaper than it already is.
At least they don't plan on switching sockets again (I hope)
Blame it on Microsoft, God does.
Re: New motherboard
Posted by RedWood on Wed Mar 21st at 7:27am 2007
Posted by RedWood on Wed Mar 21st at 7:27am 2007
Anyone have a guess on when they will change the socket again?
Re: New motherboard
Posted by Naklajat on Wed Mar 21st at 7:39am 2007
Naklajat
member
1137 posts
207 snarkmarks
Registered: Nov 15th 2004
Location: Austin, Texas
Occupation: Baron
Posted by Naklajat on Wed Mar 21st at 7:39am 2007
Hopefully not for a while >.> I was gonna get an AMD64 in S754, then AMD announced 939 so I got one of those, then AM2 came out relatively shortly thereafter. Intel has mostly stuck to Socket T in recent years, but man were the prescott Pentium 4's lame. 
I haven't been keeping up with it lately, but AM2 and Socket T should last a while AFAIK. Now it's time for the competition to heat up, quad-core chips to be made widely available, and fast dual-cores to become affordable. At least... that's what I'm hoping.
I haven't been keeping up with it lately, but AM2 and Socket T should last a while AFAIK. Now it's time for the competition to heat up, quad-core chips to be made widely available, and fast dual-cores to become affordable. At least... that's what I'm hoping.
Naklajat
member
1137 posts
207 snarkmarks
Registered: Nov 15th 2004
Location: Austin, Texas
Occupation: Baron
=o
Re: New motherboard
Posted by reaper47 on Wed Mar 21st at 10:19am 2007
Posted by reaper47 on Wed Mar 21st at 10:19am 2007
What I mostly do is looking at what the next big game you consider playing a lot will need. That proved to be helpful for choosing.
Re: New motherboard
Posted by Crono on Wed Mar 21st at 6:37pm 2007
Posted by Crono on Wed Mar 21st at 6:37pm 2007
To be perfectly honest, I'm not seeing the huge difference between 939 and AM2 socket ... physically. They have the same number of pins and pin positions. My guess would be that, generally, the AM2 has the ability to handle a higher FSB.
Because, what really happened wasn't that AMD decided to change sockets. They decided to support faster memory with a new access pattern.
So, I really doubt that a huge socket change will be happening any time soon. (I hear people are pushing GDDR3 to be used for system memory in the upcoming years ... which would make the system BUS unbelievably fast. Though, it would still only transfer 64-bits at any given moment even though that memory can access much higher amounts. Maybe it'll come when multi-core processors are common so the bus will transfer 128-bits or 256-bits (two or four cores respectively) That would be awesome.)
There are other AMD sockets out there alive and kicking, most of them are for incredibly expensive parts (FX, high end Opterons).
Because, what really happened wasn't that AMD decided to change sockets. They decided to support faster memory with a new access pattern.
So, I really doubt that a huge socket change will be happening any time soon. (I hear people are pushing GDDR3 to be used for system memory in the upcoming years ... which would make the system BUS unbelievably fast. Though, it would still only transfer 64-bits at any given moment even though that memory can access much higher amounts. Maybe it'll come when multi-core processors are common so the bus will transfer 128-bits or 256-bits (two or four cores respectively) That would be awesome.)
There are other AMD sockets out there alive and kicking, most of them are for incredibly expensive parts (FX, high end Opterons).
Blame it on Microsoft, God does.
Re: New motherboard
Posted by Junkyard God on Wed Mar 21st at 7:26pm 2007
Junkyard God
member
654 posts
73 snarkmarks
Registered: Oct 27th 2004
Location: The Nether Regions
Occupation: Stoner/mucisian/level design
Posted by Junkyard God on Wed Mar 21st at 7:26pm 2007
AM2 will also probarbly be expanded on more than the older ones.
I'm guessing you can't just slot a 939 proccesor into an am2 slot and make it work correctly.
And I'm to poor to find out![<img src=]()
" SRC="images/smiles/icon_smile.gif">
I'm guessing you can't just slot a 939 proccesor into an am2 slot and make it work correctly.
And I'm to poor to find out
Junkyard God
member
654 posts
73 snarkmarks
Registered: Oct 27th 2004
Location: The Nether Regions
Occupation: Stoner/mucisian/level design
Hell, is an half-filled auditorium
Re: New motherboard
Posted by Naklajat on Wed Mar 21st at 8:40pm 2007
/envisions 32GB GDDR3 system RAM running a 256-bit bus at a clock speed synchronous to the 3GHz quad-core CPU
/drools
Edit:
@rant
I knew AM2 was basically 939 with DDR2 support, my only gripe is that AMD changed sockets three times in a relatively short time span, and the CPUs support one socket exclusively. They should have just cut 754 out of the whole equation IMO, from my point of view it was basically a slap in the face to early-adopters in order to say "We have a 64-bit desktop platform!" before Intel. 64-bit is still mostly hype, as unless you're running *nix or enjoy having no compatible drivers and minimal performance gain in a handful of apps, your 64-bit PC is no better than a 32-bit PC, even with MS's toothless donkey dressed as a stallion that is Vista.
I think it'll take another year or two at least before 64-bit is considered the standard for commercial apps and hardware drivers.
@rant off
Naklajat
member
1137 posts
207 snarkmarks
Registered: Nov 15th 2004
Location: Austin, Texas
Occupation: Baron
Posted by Naklajat on Wed Mar 21st at 8:40pm 2007
? quote:
I hear people are pushing GDDR3 to be used for system memory in the upcoming years ... which would make the system BUS unbelievably fast. Though, it would still only transfer 64-bits at any given moment even though that memory can access much higher amounts. Maybe it'll come when multi-core processors are common so the bus will transfer 128-bits or 256-bits (two or four cores respectively) That would be awesome.
/envisions 32GB GDDR3 system RAM running a 256-bit bus at a clock speed synchronous to the 3GHz quad-core CPU
/drools
Edit:
@rant
I knew AM2 was basically 939 with DDR2 support, my only gripe is that AMD changed sockets three times in a relatively short time span, and the CPUs support one socket exclusively. They should have just cut 754 out of the whole equation IMO, from my point of view it was basically a slap in the face to early-adopters in order to say "We have a 64-bit desktop platform!" before Intel. 64-bit is still mostly hype, as unless you're running *nix or enjoy having no compatible drivers and minimal performance gain in a handful of apps, your 64-bit PC is no better than a 32-bit PC, even with MS's toothless donkey dressed as a stallion that is Vista.
I think it'll take another year or two at least before 64-bit is considered the standard for commercial apps and hardware drivers.
@rant off
Naklajat
member
1137 posts
207 snarkmarks
Registered: Nov 15th 2004
Location: Austin, Texas
Occupation: Baron
=o
Re: New motherboard
Posted by Crono on Wed Mar 21st at 10:36pm 2007
Posted by Crono on Wed Mar 21st at 10:36pm 2007
Well the 754 was really never out by it self. It was around for maybe two to three months before the introduction of socket 939.
If you have a 64-Bit machine now and you're running 32-Bit stuff you should notice a pretty large performance increase. Since, frankly, it can have two sets of 32-bits on the system bus at any given time. The processors are also pipelined so they can operate on more than a single value per cycle (This is not the same thing as having multiple cores). That's why a chip that says 3200+ has an actual clock speed of 2.0GHz or so, but it still has the same throughput as a 3.2 GHz chip.
So there is, in all actuality, an advantage: a chip that has higher throughput and uses less power and produces less heat.
I think the AMD64 was just a stepping stone onto multi-core processors. I think once Quad cores get here, it'll stay that way for awhile. But, who knows, maybe there'll be oct-cores around in the next two years.
To note, I don't think a household system with 32GB of memory would be using the GDDR3 access scheme. I don't think it's fast enough.
Graphics Cards are already on their last leg of using GDDR3 memory. Then again, GPUs are becoming general purpose.
I think in about 15 to 20 years it wouldn't be uncommon to have many general purpose chips in a home machine, especially for games. CPU, GPGPU, AIPU, GPAPU, etc. Companies are already developing AI processors, since AI needs massive speeds the algorithms behind it are usually Combinatorial or NP-C so they have exponential running times.
Really interesting stuff. I think the only processor that really shouldn't exist is a Physics Processor, that really should be done with the graphics, or at the very least ... put the damn thing on the Graphics bus. The PCI bus can only transfer about 32-bits. (This is why people are experiencing slowdown when using current PPU PCI cards)
If you have a 64-Bit machine now and you're running 32-Bit stuff you should notice a pretty large performance increase. Since, frankly, it can have two sets of 32-bits on the system bus at any given time. The processors are also pipelined so they can operate on more than a single value per cycle (This is not the same thing as having multiple cores). That's why a chip that says 3200+ has an actual clock speed of 2.0GHz or so, but it still has the same throughput as a 3.2 GHz chip.
So there is, in all actuality, an advantage: a chip that has higher throughput and uses less power and produces less heat.
I think the AMD64 was just a stepping stone onto multi-core processors. I think once Quad cores get here, it'll stay that way for awhile. But, who knows, maybe there'll be oct-cores around in the next two years.
To note, I don't think a household system with 32GB of memory would be using the GDDR3 access scheme. I don't think it's fast enough.
Graphics Cards are already on their last leg of using GDDR3 memory. Then again, GPUs are becoming general purpose.
I think in about 15 to 20 years it wouldn't be uncommon to have many general purpose chips in a home machine, especially for games. CPU, GPGPU, AIPU, GPAPU, etc. Companies are already developing AI processors, since AI needs massive speeds the algorithms behind it are usually Combinatorial or NP-C so they have exponential running times.
Really interesting stuff. I think the only processor that really shouldn't exist is a Physics Processor, that really should be done with the graphics, or at the very least ... put the damn thing on the Graphics bus. The PCI bus can only transfer about 32-bits. (This is why people are experiencing slowdown when using current PPU PCI cards)
Blame it on Microsoft, God does.
Re: New motherboard
Posted by Riven on Wed Mar 21st at 10:42pm 2007
Riven
super admin
1639 posts
802 snarkmarks
Registered: May 2nd 2005
Location: Austin, Texas, USA
Occupation: Architect
Posted by Riven on Wed Mar 21st at 10:42pm 2007
I agree, those PhysX cards really don't serve much purpose in advantages. To solve that either A.) get a better processor, or B.) better graphics card. Your right, Crono, they (gpu's) are getting pretty general purpose now with the induction of Vista.
Riven
super admin
1639 posts
802 snarkmarks
Registered: May 2nd 2005
Location: Austin, Texas, USA
Occupation: Architect
Re: New motherboard
Posted by reaper47 on Wed Mar 21st at 11:29pm 2007
Posted by reaper47 on Wed Mar 21st at 11:29pm 2007
I heard rumors that TF2 has quad core only graphics features? Maybe just multi-core in general. Valve is quite excited with PUs growing more and more together from what I heard.
Re: New motherboard
Posted by Crono on Thu Mar 22nd at 1:16am 2007
Posted by Crono on Thu Mar 22nd at 1:16am 2007
It has nothing to do with Vista. The step towards "general purpose" on the GPU began (pretty much) with Shader Model 3, since it introduced texture look ups in the vertex processors (SM3 was the first time we could do true displacement mapping)
SM4, however, is drastically different. The processors on the graphics card are no longer vector processors, they're scalar (like your CPU). This means it can't process as much data at a given time, so they upped the number of them and their clock speed. The reason why they did that is because of the limitations on vertex and pixel (fragment) shaders. Now, however, you can do many many more processes on the GPU. This is why they can now to physics on the GPU and we can morph entire polygons and meshes as opposed to individual verticies.
This is a hardware thing ... it has nothing to do with Vista or Microsoft code, even if they helped define the Shader Model standard.
Once you're actually on the GPU, it doesn't care what OS you're running or what 3D API you're using. nvidia's own shader language Cg has direct compatibility to OpenGL and GLSL. The whole thing with "general purpose" is just to give developers more power and choice when using the hardware. But, in all actuality even if you wanted to process audio on the GPU right now you can, it would just take some work in conjunction with the CPU (slow) then the result would have to be passed to the audio chip (slower).
I think what they're doing in the source engine as far as multiple cores go is doing physics on a seperate core. Which would, in turn, effect graphics.
To note, any dedicated processor for some aspect in a game would be leaps and bounds over what we have currently. The issue I was speaking of was their placement of the hardware is a completely ridiculous choice. I think in current machines it's also ridiculous that the audio processors take a long time to get to, even if you have some amazing Creative Labs card, it's still running on the PCI bus. Audio is at least as complex as lighting so it too deserves the processing power.
SM4, however, is drastically different. The processors on the graphics card are no longer vector processors, they're scalar (like your CPU). This means it can't process as much data at a given time, so they upped the number of them and their clock speed. The reason why they did that is because of the limitations on vertex and pixel (fragment) shaders. Now, however, you can do many many more processes on the GPU. This is why they can now to physics on the GPU and we can morph entire polygons and meshes as opposed to individual verticies.
This is a hardware thing ... it has nothing to do with Vista or Microsoft code, even if they helped define the Shader Model standard.
Once you're actually on the GPU, it doesn't care what OS you're running or what 3D API you're using. nvidia's own shader language Cg has direct compatibility to OpenGL and GLSL. The whole thing with "general purpose" is just to give developers more power and choice when using the hardware. But, in all actuality even if you wanted to process audio on the GPU right now you can, it would just take some work in conjunction with the CPU (slow) then the result would have to be passed to the audio chip (slower).
I think what they're doing in the source engine as far as multiple cores go is doing physics on a seperate core. Which would, in turn, effect graphics.
To note, any dedicated processor for some aspect in a game would be leaps and bounds over what we have currently. The issue I was speaking of was their placement of the hardware is a completely ridiculous choice. I think in current machines it's also ridiculous that the audio processors take a long time to get to, even if you have some amazing Creative Labs card, it's still running on the PCI bus. Audio is at least as complex as lighting so it too deserves the processing power.
Blame it on Microsoft, God does.
Re: New motherboard
Posted by Riven on Thu Mar 22nd at 2:10am 2007
Riven
super admin
1639 posts
802 snarkmarks
Registered: May 2nd 2005
Location: Austin, Texas, USA
Occupation: Architect
Posted by Riven on Thu Mar 22nd at 2:10am 2007
Well, I understood that they could do many more processes other than what their name implies, I thought it on a much, MUCH more obvious scale, that is, Vista, with its Aero view would be depending on more than just processor power to render what used to be simple windows and whatnot. But, yea, perhaps one day, there might be a card that can do it all, or processor(s) in cluster without the use of separate buses. (a potentially bad idea I suppose, with the skim of what I know about computers -just enough to build one, which is becoming increasingly easier.)
I probably don't know enough to back up all my opinions, so please bear with me... I'll refrain from commenting again in this topic lol
.
Riven
super admin
1639 posts
802 snarkmarks
Registered: May 2nd 2005
Location: Austin, Texas, USA
Occupation: Architect
Re: New motherboard
Posted by Crono on Thu Mar 22nd at 5:23am 2007
Posted by Crono on Thu Mar 22nd at 5:23am 2007
Opinions are fine, but sadly, a Single END-ALL-BE-ALL processor is not the way to go.
One of the reasons is pretty obvious: complexity for programmers. Time slicing, forking, multi-thread management, for example, are not things most people enjoy implementing. Making a single processor, or even cluster of processors, manageable from a programmer prospect is more difficult depending on the complexity of the hardware. Many processors or Cores in a single processor make things far more complex.
You might have an idea along the lines of, "well what if we just put a whole bunch of processors together and chain them up like a hypercube with n > 3 dimensions". Well, in that case, you have an issue of access. Now only a select few processors can be reached at any given time. Another problem is, once you begin computing things in conjunction (the only way it would work is if, regularly, you could use many of the processors to operate on given values), you may get processors sending data further and further away in the hypercube. So, then, unless precautions are taken, program performance is nothing close to reliable. It will just get slower as time goes on (since processors will be trying to communicate farther and farther away) There are ways around that. But the complexity for a human to program the system is still a very large issue.
If you have a 4D Hyper cube (which is two 3D cubes linked together ... that's 8 processors) you've got 16 processors operating on data. What if it was a higher dimension? Like 12, that's 4096 processors. That's very complicated!
Make no mistake, specially designed machines like this have existed for many years (in the mid-80s lots of schools had machines with 65,000 processors or so). But, the complexity is very high. They're also very special purpose.
One other thing is, you're saying, that we could speed things up by combining the buses ... so instead of having as many as we do now, we would have less that are larger.
The problem is ... that's the current problem. The system bus is slow because ALL memory access go through it. So, quite obviously the solution is not to constrain everything to a single bus (or fewer). If anything, we need more specialized buses. To note, if you had a processor hive you'd have many, thousands, of buses to connect all the processors together. The problem may not be as simple as it seems!
One quick thing about graphics cards and processing 2D data. Modern graphics cards simulate 2D, they can't actually do 2D anymore. They do 3D with no depth values. While that does cut out a lot of calculations, going to 3D only adds "pressure" when doing rotations and things like that. Otherwise, matrix calculations are still being done, the vector is just one value smaller (in all actuality, they may not even do that, they could just make the third value 0, in which case, it's still processed)
I hope all this helps!
One of the reasons is pretty obvious: complexity for programmers. Time slicing, forking, multi-thread management, for example, are not things most people enjoy implementing. Making a single processor, or even cluster of processors, manageable from a programmer prospect is more difficult depending on the complexity of the hardware. Many processors or Cores in a single processor make things far more complex.
You might have an idea along the lines of, "well what if we just put a whole bunch of processors together and chain them up like a hypercube with n > 3 dimensions". Well, in that case, you have an issue of access. Now only a select few processors can be reached at any given time. Another problem is, once you begin computing things in conjunction (the only way it would work is if, regularly, you could use many of the processors to operate on given values), you may get processors sending data further and further away in the hypercube. So, then, unless precautions are taken, program performance is nothing close to reliable. It will just get slower as time goes on (since processors will be trying to communicate farther and farther away) There are ways around that. But the complexity for a human to program the system is still a very large issue.
If you have a 4D Hyper cube (which is two 3D cubes linked together ... that's 8 processors) you've got 16 processors operating on data. What if it was a higher dimension? Like 12, that's 4096 processors. That's very complicated!
Make no mistake, specially designed machines like this have existed for many years (in the mid-80s lots of schools had machines with 65,000 processors or so). But, the complexity is very high. They're also very special purpose.
One other thing is, you're saying, that we could speed things up by combining the buses ... so instead of having as many as we do now, we would have less that are larger.
The problem is ... that's the current problem. The system bus is slow because ALL memory access go through it. So, quite obviously the solution is not to constrain everything to a single bus (or fewer). If anything, we need more specialized buses. To note, if you had a processor hive you'd have many, thousands, of buses to connect all the processors together. The problem may not be as simple as it seems!
One quick thing about graphics cards and processing 2D data. Modern graphics cards simulate 2D, they can't actually do 2D anymore. They do 3D with no depth values. While that does cut out a lot of calculations, going to 3D only adds "pressure" when doing rotations and things like that. Otherwise, matrix calculations are still being done, the vector is just one value smaller (in all actuality, they may not even do that, they could just make the third value 0, in which case, it's still processed)
I hope all this helps!
Blame it on Microsoft, God does.
Re: New motherboard
Posted by Junkyard God on Thu Mar 22nd at 8:07am 2007
Junkyard God
member
654 posts
73 snarkmarks
Registered: Oct 27th 2004
Location: The Nether Regions
Occupation: Stoner/mucisian/level design
Posted by Junkyard God on Thu Mar 22nd at 8:07am 2007
I'm thinking of getting a dual core processor and a new GFX card sometime soon.
I'd need a new mobo, processors and a GFX card.
Now I've got an AM2 3500+ AMD processor on an AM2 socketed motherboard.
And an ATI x1600something graphics card.
What do you guys reccomend me to buy, now we're going on about this kind of stuff anyways, and what price range would those suggested items be in?
Cheers![<img src=]()
" SRC="images/smiles/icon_smile.gif">
I'd need a new mobo, processors and a GFX card.
Now I've got an AM2 3500+ AMD processor on an AM2 socketed motherboard.
And an ATI x1600something graphics card.
What do you guys reccomend me to buy, now we're going on about this kind of stuff anyways, and what price range would those suggested items be in?
Cheers
Junkyard God
member
654 posts
73 snarkmarks
Registered: Oct 27th 2004
Location: The Nether Regions
Occupation: Stoner/mucisian/level design
Hell, is an half-filled auditorium
Re: New motherboard
Posted by Crono on Thu Mar 22nd at 9:03am 2007
Posted by Crono on Thu Mar 22nd at 9:03am 2007
If you've already got an AM2 socket board ... why buy another when upgrading? Just get a new chip and graphics card. (are you simultaneously building another system?)
You can get a 3800+ X2 AM2 socket for under $200 USD (transfer the currency).
What chipset is on the motherboard?
You can get a 3800+ X2 AM2 socket for under $200 USD (transfer the currency).
What chipset is on the motherboard?
Blame it on Microsoft, God does.
Re: New motherboard
Posted by Junkyard God on Thu Mar 22nd at 12:20pm 2007
Junkyard God
member
654 posts
73 snarkmarks
Registered: Oct 27th 2004
Location: The Nether Regions
Occupation: Stoner/mucisian/level design
Posted by Junkyard God on Thu Mar 22nd at 12:20pm 2007
Uhm, it's the crappy type am2 motherboard called ASUS M2NPV-VM, I think there's a VIA 333 chipset on it.
The praphics card slots into a PCI Express slot.
The praphics card slots into a PCI Express slot.
Junkyard God
member
654 posts
73 snarkmarks
Registered: Oct 27th 2004
Location: The Nether Regions
Occupation: Stoner/mucisian/level design
Hell, is an half-filled auditorium
Re: New motherboard
Posted by reaper47 on Thu Mar 22nd at 12:45pm 2007
I doubt many of us had this idea.![<img src=]()
" SRC="images/smiles/icon_biggrin.gif">
Posted by reaper47 on Thu Mar 22nd at 12:45pm 2007
? quote:
You might have an idea along the lines of, "well what if we just put a whole bunch of processors together and chain them up like a hypercube with n > 3 dimensions".
I doubt many of us had this idea.
© Snarkpit.net 2001 - 2023, about us, donate, contact
Snarkpit v6.1.0 created this page in 0.1517 seconds.
Snarkpit v6.1.0 created this page in 0.1517 seconds.