PCIe Upgrading

[atc_traffic856]

Just alittle expert information if you will. To do any new upgrading in building your system-since I see a few new developements. Would it be better to wait alittle longer [4 to 5 months I dare say] for designing your system to a 64bit and PCIe architecture? As it stands my developement plan has all the ingredients, but as I see the PCIe MB slots, and what I have so far in the planning stage is something like 2 AMD 64bit cpu's one with a Watercooler the other with Thermalite. Space wise I am OK for the two cpu's, the question is-that the MB for the dual does not have the PCIe slot.
I am not really big into gaming, would the PCIe vid card work better with the 64bit architecture? This I am not sure on, or do not know the amount of knowledge one should know- on the relationship for PCIe and the 64bit cpu.
Also I see that eventually more and more applications going to the newer
Structures, I have so far spent some time developing, and researching, asking mega questions, going through the reviews, pricing -picking everyones expertise
on the forums here. Then now I am a bit hesitant as what I shall do with the Video card??
Thank you

[atc_traffic856]

For this post I have found most of the info I was seeking, through various manufactures etc..

[pentachris]

So, what did you find & decide?

[atc_traffic856]

Tk for your interest, and random question>
Some knowledge for the dual sys chipset>
The Intel 840 chipset is a dual processor-capable chipset designed for the
> Pentium III and Pentium III Xeon Processors. The 860 chipset is also a
> dual processor-capable chipset designed primarily for the Pentium 4. Once
> of the differences between the two chipsets is that the 840 is designed
> for the 133 MHz-system bus while the 860 is designed for the latest
> 400MHz-system bus (and supports the dual channel RDRAM). The 815 chipset
> is the single processor support chipset from Intel designed for the
> 133MHz- and 100Mhz system bus freqencies and Celeron processors in the
> Pentium III family.
But as you see this is with all; or most of the dual structures, by looking at this the closes one to the best support is VIA MB Intel plans to give more in the fall, I guess it is close for there Longhorn coming available near that time.
From what returns I have been given PCIe is not available for the dual sys mb.
Do not quote me on it, though I believe one should wait out this year there is to many fires going for this year> 64b applications, PCIe, new windows [possibly forcing cpu's to change to the higher registers] I have most of what I am looking for, but to understand it all is different.

[saphalline]

Dual-CPU architectures in general make horrible game playing systems. They're always a step behind the consumer-oriented gamers' market. They do, however, make good gamer making systems, like 3D rendering workstations.

What it really comes down to, though, is why do you need two CPU's? Are you doing that much multithreading? Going for a dual-Opteron system just for playing games is a bit overkill, and definitely unnecessary. That second Opteron won't be doing much!

[mario]

First of all I apologize if this might be a new threat, but it looks somewhat connected. I have seen recently advertising and commercial proposals of two identical top-class video cards sold in couples (for example a couple of cards with NVidia 6800 Ultra PCIexpress on board) that promise exceptional results. Have this commercial proposal something to do with dual CPU architectures? Provided one has the budget for similar things, is in this case any real improvement in gaming achievable?
Mario

[pentachris]

It sounds like you're talking about SLI, which has nothing to do with dual CPU's. Read here (http://graphics.tomshardware.com/graphic/20040628/index.html) for a brief introduction.

[saphalline]

Provided one has the budget for similar things, is in this case any real improvement in gaming achievable?Assuming you're talking about SLI, the answer is yes. Two GeForce 6600 GT's in SLI is not only cheaper than a single GeForce 6800 Ultra, but they often perform on par or better! And two GeForce 6800 Ultra's in SLI, well... Let's just say that Doom 3 and HL2 showed us that they are already CPU-limited with this much graphical power! And we're talking 1600 x 1200 with max details and 16x AA and 16x AF!! :eek: Wow!

To answer your other question, no, vid card development doesn't really have anything to do with CPU development. The two are quite different indeed when it comes to purpose and design.

[mario]

I realize that I was talking about SLI without knowing what it is. I have read the suggested article. Thanks

[atc_traffic856]

Thank you for your response Saphalline>

Your question is hard to justify an answer to. "Why dual core cpu's?"
During the last 4 to 5 months I have ponder this same topic. The real insight was to your reply and article on seepage + power consumption. The more activity, and the more applications tasking the cpu the greater the problem.
Please correct me if I am wrong, in saying that the single core cpu has now reach its limitations. What I mean as for limitations; is that throughout the years of improvement for cpu's there was a continued per-centage gain. Now we are seeing a decrease in per-centage gains for efficiency. Maybe we are leaving the Instruction Parallelism single core, and are now going Multi core Multi Thread Parallelism. Of course reading many articles within the Forums, and researching the topics that were reiterated by the many well knowledge personnel, it led me to different articles for cpu technology. The only next step left is Multi core, sure there are ways and means for patch up jobs to defend against the seepage, and high power consumption. Suggestive insulation of the registers, plus the K materials that will becoming more common. But that will only last for so long, and then more break down.
Of course Multi applications are limited to a varied amount-more reserved for the enormous data basing, or what I have is a big CAD system.
Time is also a factor > when> is a big question, but I am seeing a continued advancement already for dual core, once this starts than it will become like a siphon for newer appilcations and software advancement towards the necessity of smoother computing with dual core mbs. Will it be better to have two small cpu's, whereas one is running Instruction parallelism,and as you said one is off, but when one builds up heat then you switch over, and or to also have the capability for dual multi threading? There was no more gains with AGP so now we are looking at the dual video PCIe architecture.
I realize that this is alittle rash or overkill, but if one is to build a new system
with the time and effort you will spend on it, maybe that is the way to go.
At least you may have a easier time with the new upgrades>
Thank you again. Since I am still learning I know all the comments will be more to my advantage towards decisions.

[saphalline]

Well, it seems that both Intel and AMD are looking at doing major overhauls to get two CPU cores onto one die, so the end result will be all of us upgrading anyway. No way to avoid that. You'll need a new mobo and CPU, just like always.

It's interesting that you bring up NVidia's new SLI technology, because it's really very strange and cool. The last time SLI was around, it was 3Dfx leading the charge! On PCI, no less. AGP was designed to be a vid card only slot, so it's ironic that AGP was destined to be alone, cutting off all hope of any SLI technology. Then comes PCIe, which is for more than just vid cards! So lo and behold, we see SLI make a come-back. Strange that we had to come back to a PCI-based slot to see SLI again, huh? :p

The bulk of your question/comments centers around a very important issue for the PC as a whole. There's not enough threads!! This is a problem that has plagued the PC for years! All other proprietary architectures have moved beyond the basic sequential modeling of computing, but PC's have been slow to move on. It's a catch 22 really - the CPU companies aren't willing to create massively parallel x86 architectures because there's no massive performance gains with current software, while the software companies won't make massively multi-threaded apps because there's no parallel CPU architecture to run them! But now, you're right, the CPU companies no longer have a choice. Intel especially has hit a wall with their P4. It can't go much faster without some major help! And AMD is only 18 months behind. Sooner or later, Intel and AMD will have to lead the charge not only into 64-bit territory, but also into pressuring for more multi-threaded software. The power is there! But no one is using it. Not really. Just look at Intel's HT technology. It's the greatest CPU idea to hit silicon in a long time, and it's also the greatest tale of tragedy in non-usage. Properly used, HT gives the P4 architecture a nice 20-30% boost in performance! :eek: Without increasing the clock speed, this is enormous!! Why is no one using it!? For heaven's sake, even Excel would benefit from this, not to mention all games and multimedia apps! The software companies need to get on the ball about multi-threading, or our treasured PC's will become known as the most ridiculously under-utilised computing platform of all time!

In a related story, I recently read a most informative article on the new Cell technology. For those of you who don't know, Cell is the code name for the new CPU architecture being created jointly by IBM, Sony, and Toshiba for Sony's upcoming PS3 line of media centers/gaming consoles. The Cell architecture is basically a die-localized grid-based CPU design that will rely heavily on many small & basic execution units running on the same CPU die in a massively parallel environment. As long as the software will use the full power, the performance benefits are outstanding! There will be several versions of Cell CPU's when released, some with more processing units than others, which perhaps means that Sony won't just release a single PS3, but many with varying degrees of performance and capabilities. The PS3 aside, I see big things coming from Cell! So either Intel, AMD, and the software companies need to get together, or we may be building Cell-based computer systems two years from now!