800 Mhz P4's = no big deal?

[SupermanCK]

wait.........i still can't figure out how u ppl come up with the "50% increase in fsb"??? 😕

 

[Tom]

It's a completely subjective opinion to define something as a big deal or not a big deal, so the question is unanswerable in any meaningful way.

The reasons why the new processors are desirable have been addressed, what's the point beyond that ?

1.they are faster than 533FSB chips at the same speed
2.the newest motherboards are designed for them and have other new features besides the faster bus speed.
3.there is apparently a new stepping,D1, which may run at faster speeds than previous steppings
4.there will be lower priced processors with hyper-threading
5.the new motherboards support dual DDR400 officially if used with 800fsb processors

 

[SSXeon5]

Originally posted by: Jeff7181

Originally posted by: Jombo
hey jeff7181, read this <a class=ftalternatingbarlinklarge href="http://www.anandtech.com/cpu/showdoc.html?i=1810" target=blank>article</A> and understand it before posing the same question over and over. prolly will save your time, along w/ everyone elses... btw, thanks for the article old man, i know a bit more about the fsbs now. 😀

getting a faster fsb is like having a 10 lane highway into the mall, but w/ the memory speeds being way not being able to match the fsb, 2 lane road leaving the mall, due to the bottleneck, there won't be much, if any improvements in speeds. that was the whole issue w/ cpu speeds being too fast. other components can't catch up, and deliver the instrucitons, info needed by the cpu, thus the cpu just sits there, wasting clock cycles. (so intel includes HT into their new chips to make it more efficient) and dun forget yer good ole HDs, at ATA 100/133 speeds, gains from faster fsb becomes even more marginal...

The wonderous article that you guys keep mentioning would lead one to assume, once you ad so many doors to the mall, your cashiers won't be able to keep up with the flow of customers... which when related to the P4, the CPU core would be saturated, and increasing the bus's bandwidth or speed wouldn't do a damn thing. Which brings me back to my question... is the P4 at 3 Ghz actually benefitting from an 800 Mhz bus vs a 533 Mhz bus? A 50% increase in speed, and a 12% MAXIMUM increase in performance leads me to believe no, it is not fully utilizing the 800 Mhz bus.

I cant believe I read this WHOLE THREAD i want to gouge my eyes out with a spork after reading Jeff7181's replies. I read through this thinking "o another freaken AMD fanboy talken crap" well that is what is going on. So what is your point, to bash the new fsb because YOU dont think its good enough for a 12% increase in performance. DT DT DT DT DT DT DT DT DT

SSXeon

 

[AlexKN1]

No this 800fsb is not big deal, however prescott will be a big deal.

 

[SinfulWeeper]

Well Jeff7181. Based on your system rigs. And your compaints that FSB is not a big deal. You will never have to get a new computer for say... 6 years. In fact, why not just sell your systems now and get back to an older AMD K6 266MHz. You'll make some money at it and will be happy you got rid of a computer that just scooped you out of the money for more bandwidth.
...
...
...
But wait...
that 266 won't have enough bandwidth in every place and corner.
But hey...
Those 66fsb computers paved the way for 100fsb cpu's...
which in turn made the way for the 133fsb... then the 166... and soon a 200fsb...

Even if it does not have a big gain. The 3GHz chip is clocked slower than the 3.06GHz chip but still scores higher for the most part. Now intel has a road map to get more into the future. AMD see's a future of faster CPU's... but it's kinda foggy at the moment. Not to mention AMD clouded their flollowers minds.

Lets take a look here of price/performance down to the bottom line. Price via Newegg for retail CPU's.

Intel 3.06GHz $381.50 which equaltes to $0.12/MHz
AMD Barton 3000+ $323.00 which equates to $0.15/MHz

Intel 2.8GHz $314.00 which equaltes to $0.11/MHz
AMD Barton 2800+ $230.00 which equates to $0.11/MHz

Intel 2.66GHz $217.00 which equaltes to $0.08/MHz
AMD XP2600+ $157.00 which equates to $0.08/MHz

Intel 2.53GHz $189.00 which equates to $0.07/MHz
AMD Barton 2500+ $131.00 which equates to 0.07/MHz

Intel 2.4GHz $161.00 which equates to $0.07/MHz
AMD XP2400+ $105.00 which equates to $0.05/MHz

That concludes this because as to my knowledge Intel no longer produces anything slower than 2.4GHz. Whats out there now is whats left on the market.

Now AMD's design is uncertain whats in stock for the future. Intels 800FSB motherboards are backwards compatable with their older 533 CPU's. Will AMD be the same? I don't know. I want to build an AMD system.
But overall Intel is the better value. I do very little office apps. Mostly play games, surf the net and do occasional MP3 burning and soon DivX encoding. Intel is the best bet, and on top of that, their top of the line CPU's which stomp all over AMD's processors for what I do. Cost less, or the same amount than AMD's. Not only that, I 'see' absolutely no difference in speed between my 2.53GHz computer running the occasional M$ Word or Excel as I do with my PIII 650MHz system.

So AMD smarter for business? Give me break. Maybe they are in benchmarks, but not in the real world. They made themselves the underdog by pricing their CPU's so high /MHz. If they want more market share, they need to lower their price/MHz.

But I already know the answer coming. AMD followers praise it like the bible, MHz is a myth.

Well perhaps so. AMD users are used of paying big dollars for their cycles. I could not afford it myself. Why go out and Lincoln Navigator for $50,000 when the same vehicle... the Ford Expidition costs much less?
AMD just fooled all their followers to pay prime dollar for their cycles when anyway you slice it or dice it. 1MHz is 1MHz It's just the way that you get there that is different.

*I'm am very interested in AMD answers that will be coming up. Although I expect Intel fans to ridicule me for this one alse

But It's well deserved 😀

 

[motojeff]

Jeff makes some valid points.

1) there is a limit at which higher fsb at a given processor frequency will not
help. that is a fact. but the limit is the fsb running at the same frequency as the processor. hence, not
until you hit 3ghz fsb will you see 0% increase in performance for a processor running 3ghz. (ok, the really
technical people might realize that the fsb is usually not designed these days to run at same frequency)

2) yes it is only 12%, but that is because
you are optimizing the path that is not supposed to happen. the goal is for the cache to hit! it's only on
cache misses that you will see any advantage to a higher fsb. that is why you don't get a 1-to-1 percent
increase in fsb and percent increase in performance.

3) a bigger cache (like prescott) will be a much bigger deal than a higher fsb. but, if the program already
fits in the cache, guess what, you wont see any gain in performance.

in the war of intel vs. amd, if you want to say who won, look at the stock prices. thank God for amd
though, cause they ended up getting intel to build better cheaper processors...

 

[Jeff7181]

You "guys" are funny... you call me an AMD fanboy, then you proceed to try to validate the 800 Mhz bus by slamming AMD. I made reference to AMD twice before I had to set people straight that I wasn't arguing against Intel or for AMD. Look how all the "Intel fanboys" get their shorts in a knot when someone questions the importance of an 800 Mhz bus with today's processors. Get real people... the only reason you felt I was bashing Intel is because you have some sort of emotional attatchment to Intel and got offended that someone would question what they did.

When the Barton was released... yeah... lots of hype about nothing spectacular.
When the Opteron was released... yeah... lots of hype, but it wasn't about nothing. Look at the facts. It's the most significant architechtural changes AMD has made since move from the K6 to the Athlon.

Anyway... MOST of my questions have been answered by people who present facts/information in an intelligent way, so I'm done with this thread now. The rest of you "Intel fanboys" can keep a lookout for more threads that question the performance of the P4.

 

[SSXeon5]

Originally posted by: Jeff7181
You "guys" are funny... you call me an AMD fanboy, then you proceed to try to validate the 800 Mhz bus by slamming AMD. I made reference to AMD twice before I had to set people straight that I wasn't arguing against Intel or for AMD. Look how all the "Intel fanboys" get their shorts in a knot when someone questions the importance of an 800 Mhz bus with today's processors. Get real people... the only reason you felt I was bashing Intel is because you have some sort of emotional attatchment to Intel and got offended that someone would question what they did.

When the Barton was released... yeah... lots of hype about nothing spectacular.
When the Opteron was released... yeah... lots of hype, but it wasn't about nothing. Look at the facts. It's the most significant architechtural changes AMD has made since move from the K6 to the Athlon.

Anyway... MOST of my questions have been answered by people who present facts/information in an intelligent way, so I'm done with this thread now. The rest of you "Intel fanboys" can keep a lookout for more threads that question the performance of the P4.

Sh!t ill admit im a INTEL FANBOY but all you have said here is that 800Mhz fsb is "no big deal" well suming it up in 2 words "it is" a big deal. Because you use a AMD its different to see a performance boost with a higher FSB compaired to a P4. IMHO the new FSB is great, giving up to a 12% increase over the Granite bay is very nice indeed. Remember the granite bay reviews when it was beta, after the new drivers and bios updates it got a bit faster, Same will go for the Canterwood.

Also remember that Corsiar PC2700LL (2-2-2-5) is faster then what they were using (most reviewers were using Corsair XMS Cas2 PC3200, which is (2-3-3-6), I havent seen one yet using the newer 3200LL (2-2-2-6) that has hit the market) So dont judge a book by its cover, DC DDR400 and 800Mhz fsb is brand new, give it some time and you might see a bigger jump in performance. Sorry If I came off harshly or rough, but thats the way things are, later.

EDIT : HERE is a super review Digit-life did on the performance jumps from all intels platforms, i845PE to the new i875P. Up to a 37% increase from the i845PE!!! thats pretty amazing IMHO

SSXeon

 

[chsh1ca]

Wow, the misinformed are coming out of the woodwork. 🙂

Originally posted by: SupermanCK
wait.........i still can't figure out how u ppl come up with the "50% increase in fsb"??? 😕

Uhh, do the math. 800 - 533 = 267 (which is 50% of 533).

Originally posted by: SinfulWeeper
So AMD smarter for business? Give me break. Maybe they are in benchmarks, but not in the real world. They made themselves the underdog by pricing their CPU's so high /MHz. If they want more market share, they need to lower their price/MHz.

But I already know the answer coming. AMD followers praise it like the bible, MHz is a myth.

Actually, I've never seen someone suggest that a MHz is anything more than what it truly is: A unit of measurement of the electrical speed of a circuit.

Well perhaps so. AMD users are used of paying big dollars for their cycles. I could not afford it myself. Why go out and Lincoln Navigator for $50,000 when the same vehicle... the Ford Expidition costs much less?
AMD just fooled all their followers to pay prime dollar for their cycles when anyway you slice it or dice it. 1MHz is 1MHz It's just the way that you get there that is different.

Yes, and for each 1 MHz on an Athlon XP processor, more work is being done than on the Pentium 4. That's a simple fact. Nobody is disputing that 1 MHz is 1MHz, what's disputed commonly is whether or not that really means anything. AMD pushed the 'we have a higher MHz rating' when the PIII and the Athlon were head to head because in fact, it did mean quite a bit, the pipelines were both the same length, meaning both processors were doing the same amount of work per clock cycle. When the P4 was released, the pipeline was lengthened quite a bit, meaning less actual work is being done per clock, but you can scale it to many more clock cycles. What you are suggesting is that nobody should buy any RISC processors, since they have short pipelines and lower electrical speeds. The fault in that thinking is that RISC processors do a hell of a lot more work than most CISC processors (which include the P4, and the Athlon) in that time frame, which is why they are much better when it comes to raw calculations.

You may consider it to be meaningless, but you're wrong, and that's a provable fact, and is proven quite regularly by the benchmarks released around the 'net. Price/MHz don't matter anywhere near as much as Price/Performance.

The bottom line is 800MHz P4s are a big deal if you spend a significant portion (I'd say more than 30%) of your time doing media encoding of one form or another, and other high-memory access situations, but otherwise, you won't see much benefit.

 

[acidpad]

wow .... its been a while since i have seen a group of people being so biased one way or another.

There are many advantages with going to a 800mhz bus, one , as was already described, is the fact that the p4 thrives on it ... great, but perhaps the better way to look at this, as an overclocker, is that the multiplier is going to be set lower. when it comes to overclocking, its near impossible to change the multiplier on a p4 (less ES chips) and thus you have to raise the fsb..... the lower the mult is locked, the higher you can crank that fsb. this will open up a nice area for overclocking assuming that the mfr process can take the higher frequencies. even if it cant right now, wait until the prescott hits with .09 mfr process ... overclockers wet dream.

now, as far as you were stating about diminishing returns ... YES there are diminishing returns. there will be a point that the p4 has all the fsb bandwith it needs and it becomes bottlenecked somewhere else, this WILL happen , i dont care how much the p4 is reliant on the fsb, however the point of diminishing returns is not a point that you are going to be able to reach, so its not something you need to worry about.

The fact is , there is no blatent answer to the original questioned asked, threre is a limit, just not a feasible one, and no, you would not see a difference between a current architechture 1500-fsb x 2- mul and a 3000fsb x 1 - mul. i hope this post helps settle this and not make it worse, and more over, i hope it answeres the original question more accurately.