Curse

Didymus C. Corax · Apr 08, 2005, 03:27 AM // 03:27

Other than the discussion about 32v64 bit processor differences, you folks sound a LOT like the conversation going on in my head (no...I don't hear voices...they type in a forum like the rest of us). The gig of RAM is a given, but I hate to get a new video adapter and then have to buy another when I switch over to a mobo that takes the PCI-Express.

I've read some about the improvements of PCI-E over PCI and it sounded like I'd get a bit more performance...The mobo I mentioned takes a much better CPU than the 3000+, but I figured that in the future (when, say, Longhorn came out) I'd upgrade the CPU. I was thinking the 64 bit 3000+ would not be that great a dip from the XP 3200+ I'm running now (truth be told, I could easily get the 64 bit 3200+ if I waited a month...*digs toe into floor*...hate waiting

)...I'm mostly looking at the future here...The mobo is the base (highly upgradable...plans for RAID 0+1), the CPU is more-or-less a lateral move, and the PCI-E card gets me a bit of a boost and I'm ready for the next baby-step...That's what I was thinking...

I know there is no huge immediate jump in performance, but $500 now, $500 later is much more do-able than $1000 or even $800 in one chunk later (I'm a teacher and cash doen't last long around here). I'm hoping to be getting a little boost now with the greatest potential for later...Oh, yeah, there IS the little matter of that mobo being easily tweaked to run SLI (warrenty-voided...I'd never go there)...I can see waiting for the next incarnation (Venice) of the 64 bit, though...It's just hard to wait...seems like every time I get $500 in my pocket, my car dies...law of the realm, I guess.

You've given me some things to think about...I'll let you know which way the wind blows as soon as I figure it out...gonna be a looooooong weekend...

Lansing Kai Don · Apr 08, 2005, 03:28 AM // 03:28

Quote:

Originally Posted by SSE4

You can easily measure the number of clock cycles it takes a processor to scan the cache using benchmarking programs, and instructions can also be measured in this way. I would suggest searching up some information on benchmarks that will show you that by increasing the cache, you increase the amount of "cache" memory the computer must search, and will naturally create latency. It's not a good idea to assume that the processor will search through the cache without latency. It's like assuming RAM doesn't have latency either, which is false. Some have a CAS Latency of 3, and others have 2.5, and some of the faster DDR400 modules (Or less of course) will have 2. I don't remember the sites. The best way for you to find out is to go looking for information yourself. I wasn't spoon-fed, and I wont give you special treatment because you don't believe me.

I give way on the Pentium M discussion, I found the source and it is there is no difference between Pentium 4 and Pentium 4-M. I wasn't sure if I was right, and it turns out I wasn't. Ok, wow, how am I going to explain this. L1 cache shouldn't have latency (why do you think they don't us a larger die half of the processors is cache, they could have alot more cheap cache if they just made the die larger, it'd be more EXPENSIVE, but there'd be a market for it?). RAM works completely different than the registers on your cache. First of all you have a completely different addressing scheme. There is no simple this set of transparent latches are 00, 01 etc.. You have an address bus which contains the address of the RAM module and the offset. A multiplexer/decoder deems which RAM module, then the offset for RAS,CAS are used to specify which row/column to start the transfer. Caches is straightforward last I checked, you just have on multiplexer and the cache is selected and taken from 32 bits at a time (if it's 32 or 64 bit if it's 64 bit cache). Now L2 cache may experience some drop in speed, and probably picoseconds of latency, but it shouldn't by no means be large.

Lansing Kai Don

P.S. L1 cache shouldn't have latency (delays yes, latency no), I read my post again and I said L2 sorry. L2 cache shouldn't have really any affectual latency. But comparing RAM to cache in terms of latency, not a good idea.

SSE4 · Apr 08, 2005, 03:34 AM // 03:34

Quote:

Originally Posted by Lansing Kai Don

P.S. L1 cache shouldn't have latency (delays yes, latency no), I read my post again and I said L2 sorry. L2 cache shouldn't have really any affectual latency. But comparing RAM to cache in terms of latency, not a good idea.

Of course, using RAM was a bad example but it gets the point across. Because the cache still has latency much in the same way RAM does. The difference will be more than picoseconds on a program that relies heavily on how fast the computer can search through a larger cache. However it isn't a big difference for the majority of other programs. And I know L1 shouldn't have latency, that's why I don't really count it in any consideration, it's different.

And remember Didymus, patience is a virtue.

Lansing Kai Don · Apr 08, 2005, 03:38 AM // 03:38

Quote:

Originally Posted by Didymus C. Corax

Other than the discussion about 32v64 bit processor differences, you folks sound a LOT like the conversation going on in my head (no...I don't hear voices...they type in a forum like the rest of us). The gig of RAM is a given, but I hate to get a new video adapter and then have to buy another when I switch over to a mobo that takes the PCI-Express.

I've read some about the improvements of PCI-E over PCI and it sounded like I'd get a bit more performance...The mobo I mentioned takes a much better CPU than the 3000+, but I figured that in the future (when, say, Longhorn came out) I'd upgrade the CPU. I was thinking the 64 bit 3000+ would not be that great a dip from the XP 3200+ I'm running now (truth be told, I could easily get the 64 bit 3200+ if I waited a month...*digs toe into floor*...hate waiting

)...I'm mostly looking at the future here...The mobo is the base (highly upgradable...plans for RAID 0+1), the CPU is more-or-less a lateral move, and the PCI-E card gets me a bit of a boost and I'm ready for the next baby-step...That's what I was thinking...

I know there is no huge immediate jump in performance, but $500 now, $500 later is much more do-able than $1000 or even $800 in one chunk later (I'm a teacher and cash doen't last long around here). I'm hoping to be getting a little boost now with the greatest potential for later...Oh, yeah, there IS the little matter of that mobo being easily tweaked to run SLI (warrenty-voided...I'd never go there)...I can see waiting for the next incarnation (Venice) of the 64 bit, though...It's just hard to wait...seems like every time I get $500 in my pocket, my car dies...law of the realm, I guess.

You've given me some things to think about...I'll let you know which way the wind blows as soon as I figure it out...gonna be a looooooong weekend...

I can only recommend, the decisions are up to you.
First of all the PCI-E architecture is naturally better than PCI. Where PCI uses a shared bus in a parallel mode. They went to individual serial buses for the PCI-E (at least I hope I got it right). The PCI-E will give you bang for the buck but there is no real improvement over AGP (i.e. except Intel pushing people toward where they want them to go, AGP would probably still be in the market for MANY years to come). But if Intel gets their way (and they usually do), then AGP will be cycled out in the next 3 years. So if your looking for a very long term commitment go with PCI-E, otherwise, it's irrevelant IMO. I don't like RAID 0+1, either go RAID 0, RAID 1, or RAID 5 (if you got 3 drives). RAID 0 if you want performance, RAID 1 if you want safety and reliability, and RAID 5 if you want both. If you have an XP 3200+, I see no reason to move your position till the Venice core comes out (and of COURSE it is going to be better, or ppl will find out right away and AMD will look bad, and AMD can't afford that). It seems your upgrading fast so I'd keep the 3200+, go with a different RAID, and buy whatever video card you want for now. Then next time it comes along, check the price of the Venice core, my personal favorite is the Toledo Dual Core (drools). Most likely over 500 dollars there. It's up to you though

Lansing Kai Don

Darkmane · Apr 08, 2005, 03:41 AM // 03:41

Quote:

Originally Posted by Lansing Kai Don

I give way on the Pentium M discussion, I found the source and it is there is no difference between Pentium 4 and Pentium 4-M. I wasn't sure if I was right, and it turns out I wasn't. Ok, wow, how am I going to explain this. L1 cache shouldn't have latency (why do you think they don't us a larger die half of the processors is cache, they could have alot more cheap cache if they just made the die larger, it'd be more EXPENSIVE, but there'd be a market for it?). RAM works completely different than the registers on your cache. First of all you have a completely different addressing scheme. There is no simple this set of transparent latches are 00, 01 etc.. You have an address bus which contains the address of the RAM module and the offset. A multiplexer/decoder deems which RAM module, then the offset for RAS,CAS are used to specify which row/column to start the transfer. Caches is straightforward last I checked, you just have on multiplexer and the cache is selected and taken from 32 bits at a time (if it's 32 or 64 bit if it's 64 bit cache). Now L2 cache may experience some drop in speed, and probably picoseconds of latency, but it shouldn't by no means be large.

Lansing Kai Don

P.S. L1 cache shouldn't have latency (delays yes, latency no), I read my post again and I said L2 sorry. L2 cache shouldn't have really any affectual latency. But comparing RAM to cache in terms of latency, not a good idea.

Not that I am not guilty also, but lets stop cluttering this poor guys' thread with our discussions about 32 and 64 bit. LOL

I still say wait on the cpu and motherboard, .... get yourself more memory- and a good AGP card. You can prolly sell the AGP on eBay when your ready for your new motherboard and PCIx.

Good Luck to you and your decision.

Lansing Kai Don · Apr 08, 2005, 03:41 AM // 03:41

Quote:

Originally Posted by SSE4

Of course, using RAM was a bad example but it gets the point across. Because the cache still has latency much in the same way RAM does. The difference will be more than picoseconds on a program that relies heavily on how fast the computer can search through a larger cache. However it isn't a big difference for the majority of other programs. And I know L1 shouldn't have latency, that's why I don't really count it in any consideration, it's different.

And remember Didymus, patience is a virtue.

Well cache is set straight, I still disagree with you referencing Pentium M better w/o sources. I don't understand why the latency would be larger than pico seconds, unless your running into a natural deadlock where your writing information you need to read before hand. But that would be a near stop on the processor, maybe preventative measures keeps the latency higher? You'd have to show me sources, I can't find any on cache latency that I'd consider reputable.

Lansing Kai Don

SSE4 · Apr 08, 2005, 03:44 AM // 03:44

Quote:

Originally Posted by Lansing Kai Don

The PCI-E will give you bang for the buck but there is no real improvement over AGP (i.e. except Intel pushing people toward where they want them to go, AGP would probably still be in the market for MANY years to come). But if Intel gets their way (and they usually do), then AGP will be cycled out in the next 3 years. So if your looking for a very long term commitment go with PCI-E, otherwise, it's irrevelant IMO.

Intel and AMD have both been pushing it of course. Partly because the PCI-E BUS has more room for growth, whereas they figure that AGP could be saturated in the next generation or two of video cards. It makes more sense to get PCI-E out now so by the time AGP is saturated, hopefully at least 80% of the market will already have a PCI-E motherboard. But it's true. He wont find any performance advantages with a PCI-E video card in comparison to an AGP one because there aren't video cards capable of saturating the AGP BUS, but it's always good to think ahead.

The Pentium M: It really depends on what you do with the processor of course.
http://www.anandtech.com/cpuchipsets...oc.aspx?i=2342
http://www.behardware.com/articles/546/page1.html
http://www.gamepc.com/labs/view_cont...ngaming&page=1
http://www.pcper.com/article.php?aid=123&type=expert
http://www.hardocp.com/article.html?art=NzA5LDE=
http://www.legitreviews.com/article.php?aid=181
http://www.anandtech.com/cpuchipsets...spx?i=2382&p=1
These are just some reviews/benchmarks on the Pentium M. The first one from anandtech takes a look at its cache as well.

Lansing Kai Don · Apr 08, 2005, 03:47 AM // 03:47

Quote:

Originally Posted by SSE4

Intel and AMD have both been pushing it of course. Partly because the PCI-E BUS has more room for growth, whereas they figure that AGP could be saturated in the next generation or two of video cards. It makes more sense to get PCI-E out now so by the time AGP is saturated, hopefully at least 80% of the market will already have a PCI-E motherboard. But it's true. He wont find any performance advantages with a PCI-E video card in comparison to an AGP one because there aren't video cards capable of saturating the AGP BUS, but it's always good to think ahead.

We need to keep this agreeing thing up, it works better on my fingers. The only reason I didn't mention that AGP hasn't reached it's full potential in bandwidth and there is no point in switching "right now" except for compatibility for the future is that I have said it like on 3 different threads

I guess I should have linked it.

Lansing Kai Don

SSE4 · Apr 08, 2005, 03:50 AM // 03:50

Agreement is good. It's pretty easy for us to agree on technology like this, since for us it's merely information. We receive it and that's it. There's no real debate.

IdNotFound · Apr 08, 2005, 03:51 AM // 03:51

Quote:

Originally Posted by SSE4

The Winchester is better than the Newcastle, but said to be a lot worse than Venice. It's built on the 90nm process, which means it uses less energy and runs slightly cooler, but it is debatable if performance beyond these two things will be noticed. Nevertheless, it's a more viable option than any Newcastle core, because Winchester is better. To be honest, I don't know how expensive the Venice cores are intended to be, so if someone else knows I would be delighted if they could tell me. I know that the Venice cores will feature SSE3 instructions and it also appears that they will run a fair bit cooler than current models. It also has been documented to be more easily overclockable than all previous models of AMD64s. This could mean that if you buy a cheaper model, you may find yourself capable of overclocking the processor an extra 100-500MHz (Depending on how extreme you might go) and save money. I myself (So no scientific or economic document backs this) feel that the Winchester (But especially the Newcastle) will likely go down in price with the introduction of the Venice cores. They wont be as good, but if they go down in price you can't really argue in consideration to value.

It's safe to assume that the Venice cores will cost a fair bit more than current AMD64 models, but I honestly don't know.

I did some researching a while back on hardware websites (basicaly Anantech, but I did take a peek on ExtremeTech and TomsHardware -- are those good/reliable?) and it really looked like there were no significant (if any) boosts in performance with the 90nm.

I agree with you the prices tend to drop with the Venice launch. Thing is, how much and how fast. :/

Could you please give any more details on what the SSE3 is? As in which areas - if not all - will it improve performance? Gaming, graphics, regular PC use, ...

I guess you can never be 100% satisfied or up to date when it comes to hardware, uh?

Thanks again.

SSE4 · Apr 08, 2005, 03:57 AM // 03:57

Quote:

Originally Posted by IdNotFound

I did some researching a while back on hardware websites (basicaly Anantech, but I did take a peek on ExtremeTech and TomsHardware -- are those good/reliable?) and it really looked like there were no significant (if any) boosts in performance with the 90nm.

I agree with you the prices tend to drop with the Venice launch. Thing is, how much and how fast. :/

Could you please give any more details on what the SSE3 is? As in which areas - if not all - will it improve performance? Gaming, graphics, regular PC use, ...

I guess you can never be 100% satisfied or up to date when it comes to hardware, uh?

Thanks again.

They are all reliable to some extent. I would suggest looking up more. As Lansing had brought up, he wants reputable sources, but it's debatable whether or not any "internet" source can be truly reputable. SSE3 is, basically, a media-based instruction set that allows the processor to perform more floating-point calculations. This will make it calculate certain things faster and can be especially useful in games or multimedia. It is likely there will be a fair increase in performance because of this. And those sites were right about the Winchester vs Newcastle, but it's rather nice to have a Winchester as opposed to the Newcastle simply because of the fact that it runs cooler and uses a bit less energy. It isn't much but for the price difference (Which seems to typically be very small) you might as well go for it.

And no, you can almost never be completely satisfied with computer hardware, since it's almost always changing. You could say the people who pushed the original $1300/$1400CDN for the FX-55 have been enjoying "top of the line" performance for quite some time now, but some of us don't have that kind of money.

Lews · Apr 08, 2005, 04:00 AM // 04:00

I have to disagree there though, I think the 939 system you mentioned would be way better for you. Then you could just buy another stick of ram in a month or whenever you can afford it, instead of waiting to have enough for the rest.

Lansing Kai Don · Apr 08, 2005, 04:03 AM // 04:03

Quote:

Originally Posted by SSE4

Agreement is good. It's pretty easy for us to agree on technology like this, since for us it's merely information. We receive it and that's it. There's no real debate.

That's not exactly true. I work with most of the products in some fashion. The processor's we've been discussing are in my companies controllers (I don't know if I can say which one and don't want to get in trouble). I also get some pretty savory confidential reading material from one of these companies that I can't brag about so it's useless. I might be able to generalize after I've read more. And I agree that internet sources aren't reputable, but if you take let's say 10 of them and see how close their numbers approximate on DIFFERENT testing equipment. Then you can get a general idea.

Lansing Kai Don

P.S. Wish I had a FX-55, better yet a Toledo core.

Lansing Kai Don · Apr 08, 2005, 04:05 AM // 04:05

Quote:

Originally Posted by Lews

I have to disagree there though, I think the 939 system you mentioned would be way better for you. Then you could just buy another stick of ram in a month or whenever you can afford it, instead of waiting to have enough for the rest.

I'd agree to some extent, but the Venice is on the doorstep. And I'm positive prices will drop when it comes. But then again, it's easy to argue that there is always going to be another processor core that comes along that could do the same and you could wait indefinitely. But if it was me, I'd wait. Then get a processor about 3-weeks to a month after the Venice (give the drop some time, they never drop immediately).

Lansing Kai Don

SSE4 · Apr 08, 2005, 04:08 AM // 04:08

Quote:

Originally Posted by Lansing Kai Don

That's not exactly true. I work with most of the products in some fashion. The processor's we've been discussing are in my companies controllers (I don't know if I can say which one and don't want to get in trouble). I also get some pretty savory confidential reading material from one of these companies that I can't brag about so it's useless. I might be able to generalize after I've read more. And I agree that internet sources aren't reputable, but if you take let's say 10 of them and see how close their numbers approximate on DIFFERENT testing equipment. Then you can get a general idea.

Lansing Kai Don

P.S. Wish I had a FX-55, better yet a Toledo core.

True. That is why I like to take as many sources as I can, and I don't rely on any one source. I don't take any one sources "word" as law, because that can be dangerously ignorant.

Ellestar · Apr 08, 2005, 06:22 AM // 06:22

Quote:

Originally Posted by Didymus C. Corax

The gig of RAM is a given, but I hate to get a new video adapter and then have to buy another when I switch over to a mobo that takes the PCI-Express.

Your NEW video card will be obsolete before your OLD processor will be obsolete. So, if you really want to save money, do it. Don't waste money on a processor you don't need at all. When you'll need to change your current processor, you'll need to change your video card anyway, be it AGP or PCI-E. So save your money now and use it later.

Quote:

Originally Posted by Didymus C. Corax

I've read some about the improvements of PCI-E over PCI and it sounded like I'd get a bit more performance...

There are some improvements of PCI-E over AGP in theory. Do you need a theories you should pay your money for or you want just to play games on your computer? Choose wisely.
Video cards that can use PCI-E advantage aren't in existence yet. So, there is absolutely no advantage.

Quote:

Originally Posted by Didymus C. Corax

The mobo I mentioned takes a much better CPU than the 3000+, but I figured that in the future (when, say, Longhorn came out) I'd upgrade the CPU. I was thinking the 64 bit 3000+ would not be that great a dip from the XP 3200+ I'm running now (truth be told, I could easily get the 64 bit 3200+ if I waited a month...*digs toe into floor*...hate waiting

)...I'm mostly looking at the future here...The mobo is the base (highly upgradable...plans for RAID 0+1), the CPU is more-or-less a lateral move, and the PCI-E card gets me a bit of a boost and I'm ready for the next baby-step...That's what I was thinking...

And you think the wrong way. You said you don't have a lot of money. So don't waste it. You don't need 64 bit, PCI-E, RAID, future etc. You need to play games so you need enough FPS and no lags. Realizing this will save you a lot of money.

Quote:

Originally Posted by Didymus C. Corax

I know there is no huge immediate jump in performance, but $500 now, $500 later is much more do-able than $1000 or even $800 in one chunk later (I'm a teacher and cash doen't last long around here).

You're wrong again. There will be a huge immediate jump in performance if you'll change your video card. So buy GeForce 6800 GT AGP + 512 Mb RAM or maybe GeForce 6600 GT AGP + 256 Mb RAM if you want to save some money - GeForce 6600 GT AGP and 768 Mb RAM will be good enough anyway.

Quote:

Originally Posted by Didymus C. Corax

I'm hoping to be getting a little boost now with the greatest potential for later...Oh, yeah, there IS the little matter of that mobo being easily tweaked to run SLI (warrenty-voided...I'd never go there)...I can see waiting for the next incarnation (Venice) of the 64 bit, though...It's just hard to wait...seems like every time I get $500 in my pocket, my car dies...law of the realm, I guess.

Sounds like you want to find ways to waste $500 now just because you have it. Fine, you can do it without any advices. Sure, if you want "potential" (i.e. nothing), you can waste $500 on it. You can also buy snow in Alaska or sand in Sahara - it will be as good for you as buying potential. Looks like marketing departments work better than i thought - they can sell a virtual potential for real $$$.

Armaio · Apr 08, 2005, 02:43 PM // 14:43

damn you are getting your evga e-force 6600 GT cheap! I had to pay $269 for mine because I'am in Canada, and this was a good price!

Didymus C. Corax · Apr 09, 2005, 01:31 AM // 01:31

Ellestar, you made me shoot milk through my nose!!! (I wasn't even drinking any at the time)...seriously, though, you made me laugh at myself

. I thought about it and after reading yours and several other's words of wisdom and checked the price on a new AGP card (the 6600GT) and came up with a fairly good price ( a tad more than the PCI-E card I was looking at, to the sum of $80 over) and the price of another 512mb stick o' RAM (cheap these days)...I end up with a couple hundred bucks left over and a comparitively decent boost in performance over my already servicible machine.

I think I spent so much time drooling over the new technology, I became dehydrated. I just bought a laptop a couple months ago (Intel P4 2.8GHz, ATI Radeon 9000 mobility Vid, 512 RAM, blah, blah, blah) and though I use it at work, I play GW on it when the wife wants to use the primary machine to shop for shoes on ebay. It does the job just fine, so why am I obsessing over bigger, better, fasster, more on my primary PC? I've rebuilt it twice and it has one original part from when I bought it (DVD drive, and a mediocre one at that....hmmmmm...a couple hundred dollars left over....hmmmmm). Truth be told, I'm going with the AGP and Memory...thank you folks for helping me see the light...as for the future....I'll wait till it gets here THEN decide what to do.

Didymus C. Corax · Apr 12, 2005, 02:27 AM // 02:27

Thanks again for the help folks...I ended up getting an MSI 6600GT (my mobo is MSI and I like it just fine thank you very much, so no snide remarks

) and I put in 512 more RAM...I haven't benchmarked it yet, but primary testing has caused me to reach this conclusion:

sweeeeeeeeeet*drools*

PhineasToke · Apr 12, 2005, 02:48 AM // 02:48

I am a firm believer in the old adage;

No software written in 64 bit, waste of money.

By the time the developers catch up (only when Intel releases their 64bit version for public consumption) it will be worthwhile. Intel is like Winblows, when you dominate the market, you call the shots.