3.2GHZ is out, AMD raped badly

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
If someone can afford/tolerate the depreciation of buying the latest thing every month, then good for them; the extra peformance does make a difference in things like rendering and optimising/converting large gif/flash files etc.

I tend to buy complete midrange systems (current one is an XP2100) that I can run for at least 12 months without being tempted to upgrade. Of course, after 12 months it makes more sense to buy another midrange system instead of upgrading the existing one (and use the other one for doing those 72 hour rendering jobs...)
 
You, imgod2u or Slvr_phoenix could try to counter every line this fanboy has launched. Or if you don't want to bother, I understand, it's way too FUDed! :wink:

--
If I could see the Matrix, I'd tell you I am only seeing 0s inside your head! :tongue:
 
If the P4 was an efficent cpu it wouldnt need 3ghz to beat a 2ghz athlon with a design reaching back to what, 1998?"
Click to expand...

Because the P4 has less IPC but more MHz. But one couldn't possibly matter more than the other, because the performance turns out the same, if the rest of the CPU's are the same of course.

But then again most nerds TALK about a 3.2ghz cpu, and will not, do not, pony up the dough.
Click to expand...

How I absolutely hate that word. Just because you have an interest doesn't mean you automatically are a close-minded person who doesn't go out of the door to see his/her friends.


My system: Intel Pentium 4 2.8, 800FSB / TwinMOS 1Gb DDR400 / MSI 875P Neo / Sapphire Radeon 9800 Pro / Antec True Power 550W / Western Digital Raptor / Hercules G.T XP /
Samsung DVD / Lite-On CDRW
 
The newest intel cpu I own is a pentium 90mhz. It was being used about 6 months ago, but now it's in the closet again.

My roommate has a 2.2Ghz pentium 4 OC'ed to 2.4ghz or something and my other roommate has a 1.3Mhz pentium 4 from dell. My other roommate who isn't here right now has 2 amd athlon xp's. I think one is 2200+ and the other is 1700+.

Personally I have: Athlon XP 2100+, AMD Tbird 1.0ghz, and VIA C3 600mhz.

I don't have much cash to fork out so the latest stuff doesn't really matter to me. The way I see it, sales of most hardware these days depends on how much the person is willing to spend. So my roommate got a pretty beefy system, because he had the money. Me and my other roommate with the AMD's don't have much money when we buy stuff.

So my roommate with the P4 loads warcraft3 games 5 seconds faster than me. I don't really care much, not much of a difference in gameplay. None of the games I play really use all of the power so my computer does all I need. Besides, my computer loads windows faster than my roommate's P4 2.4ghz (and most people's computers) but that's probably because I tweaked my system and I'm a computer person.

The point I want to make is, the latest processor only affects the people that are interested, have the cash, and are willing to use the cash on it. I'll probably turn into one of those people when I get a good paying job and have extra money I don't know what I should do with.

But right now, I'm the average consumer trying to maximize the "bang for the buck." In that sense the performance you can get for so cheap from amd is pretty amazing. I don't mind intel, they're a good company. If they made their products more attractive as far as price, I'd probably buy their products too (low-mid range, but not top of the line). I don't like buying the latest stuff because then a week or two later something better comes out and I get depressed. At least when I buy something in the middle I don't get depressed until they stop selling the product :).
 
You're right, it's also a matter of interest about how much you're willing to spend. And we all have different priorities. It has been quite many times when I had to sacrifice lots of other things just to be able to afford what I had in my computer at the time.

But sure, buying mid end or lower high-end is probably a better idea, (hey I know I'm crazy) 😛
My current setup is just too...vulgar with 2x WD Raptor hard disks 😛
Hey, even the next-to-best CPU at the time, usually gives quite a good price/performance ratio.
But your arguments are all valid and sane as well 😉, I wish I had the calmness inside to avoid buying such high-end stuff ;P
My economy also suffers with loans to family and such because of it. But hey, that's computer business :)

My system: Intel Pentium 4 2.8, 800FSB / TwinMOS 1Gb DDR400 / MSI 875P Neo / Sapphire Radeon 9800 Pro / Antec True Power 550W / Western Digital Raptor / Hercules G.T XP /
Samsung DVD / Lite-On CDRW
 
I "discovered" something recently that made me feel a bit vindicated in light of my previous comments. If one were to puruse older benchmarks, he would find that AMD Bartons and T-Birds hold up quite well against P4's when the P4 was "saddled" with the i850 and, especially, the i845.

I would venture to say the AMD's PR numbers are correct with regards to thier Intel counterparts. The weak link is the chipset. Hence, AMD did not get "raped", it was nVidia.

Dave

Intel giveth and Microsoft taketh away.
 
Well, Yeah - but the point is nforce2 compared to ANY other AMD chipset romps home easily in the lead.

I would say it's more the early P4 Chipsets that were raped by the early AMD ones, but the whole P4 & its chipset architecture were just that little more future-proof from the start, and now we see the evidence of that, with Intel running off into the sunset with all the performance awards, while AMD is struggling with their architecture being so far grounded in the past.

If Intel had simply improved the P3 to get their P4 (much like AMD did with athlons) then I'm sure they'd be facing the same problems as AMD, but they had the financial balls to go for a complete re-design for the new chip.

---
$hit Happens. I just wish it would happen to someone else for a change.
 
I am not sure that you all get the point here:

If you take a look at the Intel P4 3.2ghz HT CPu and a I875 motherboard(this is the most feature comparable to most Nforce2 chipsets IMHO) you pay 534 dollars just for those alone (Abit IC7+P4 3.0ghz(couldn't find a price for the 3.2ghz)). If I take an Nforce2 mobo(Abit NF7-S) and a Barton 3000+ it is 467 dollars. Now while that ~70 dollars may mean squat to a select few who have a big budget, that means 2 sticks of ram to me, or the difference between getting a GF4ti or a Rady 9700. That matters to people. The 3200+ may be slightly slower than the P4 3.2ghz, but it is priced slightly lower as well.

I do agree that people that can afford bleeding edge technologies should go Intel for now. However the mainstream user who wants a more cost effective upgrade should go AMD. Also the tech industry was hit hard during this latest recession and I think prices reflect that. People are not buying the high end CPU's because they cannot afford to pay the bills and spend 400-700 dollars on a cpu. I think that you will see a big downward turn on prices very soon and when that happens, if intel does what it should do, you will have my vote for intel as a good value. Until then it is too overpriced IMHO for the small performance gained over a cpu that runs almost a full ghz slower clock for clock, but doesn't give a full gig worth of performance boost.

Just a computer junky
 
Well, not so much their (AMD) architecture as their cache structure. But yes, I would tend to agree with the atatement that early and interim Athlons raped Intel P4s. Despite the superior chipset performance of the P4 platform (memory and ide throughput), AMD still managed a significant lead.

I wish I had a reliable source at AMD to ask why they have problems with chipset support to the level that Intel has. I know the official stance is AMD does not intend to be in the chipset business. Well now the memory controller is embeded in the Opteron/A64. So, we shall see.

Regards,

Dave

Intel giveth and Microsoft taketh away.
 
No, no! We get the point. The original post of this thread concerned performance comparison of AMD and Intel flagship CPU's-- specifically mentioning AMD's Performance Rating. No mention is made of price/performance, and is mostly irrelevant for this <i>particualr</i> discussion.

We can know that conclusions drawn from supposed direct comparisons between the Intel P4C and newer AMD Athlons have the potential for fallacies, since it is obvious that the CPU performance in applications depends on the system board as well.

I have offered emperical evidence to refute the notion that AMD's PR rating is invalid.

Regards,

Dave

Intel giveth and Microsoft taketh away.
 
I fully understand the whole Price/Performance thing - that's why I own two Athlon systems myself.

I was merely responding to dwellman's point that Athlons were more than a match for the early P4s - just saying that the P4 architecture was ultimately going to win out in the end - just as it's doing now - the Basic design of the Athlon cannot really get much faster, whereas the P4 can, and will. That's all I was saying....

If you don't have 2 grand spare to build a top-notch PC, then yes, I completely agree that AMD is the way to go.... My most recent system is an AXP1700+ based one, after all (but 2.25Ghz... :smile: ).. you gotta love those t-bred 'b' chips.

Although, I would say that the price difference between the XP3200+ and the 3.2Ghz P4 does not reflect the difference in performance. I can buy a P4C 3.0Ghz for slightly less than the XP3200+, and that would still outperform it. The price is correct if you look at the model number, but the model number is incorrect compared to the performance...


---
$hit Happens. I just wish it would happen to someone else for a change.
 
The weak link is the chipset
Click to expand...

It is not the chipset at all. nForce 2 nearly MAXES the bus out and memory bandwidth. Its efficiency is at about 90% which is far higher than any chipset out there, including Intel's. Intel's aren't THAT much more performing, they simply follow the logic of evolution of performance, period.
You see, it's not chipset that hinders here, as that has been a resolved story a while ago, it's the architecture. Not only that, it's x86. x86 asks you to go too far in clock speed and pipelines to have more parallelism. You literally needed 3 x86 decoders on the K7 to get 10-15% more performance per clock over the P3. All of that after you had 3 decoders which could've theoretically made you calculate 3 instructions per clock, which is a huge feat that rarely ever happens.
The K7 architecture is as doomed as the P4's. We will continue to max out our CPUs, even the P4s at some point, and want more until we get Itanium's like CPUs, which use an architecture that TRULY utilizes and maxes its resources. Itanium, at 1GHZ has the FPU of about 3 Athlon 1.4GHZ together. Imagine it at GHZ. It would make Pixar's latest CPUs run in "motion" away from the animation giant!

In conclusion, chipset is not hindering it, I DOUBT IT, we have maxed out almost the bus and memory bandwidth and more won't help. The Athlon doesn't fetch in huge packs the memory data. P4 does twice, therefore needs more bandwidth.
Of course guess why it needs more at the same clock? Because it needs to have twice more data in the pipeline, I am guessing.

--
If I could see the Matrix, I'd tell you I am only seeing 0s inside your head! :tongue:
 
It is not the chipset at all. nForce 2 nearly MAXES the bus out and memory bandwidth. Its <b>efficiency</b> is at about 90% which is far higher than any chipset out there, including Intel's. (emphasis added)
Click to expand...

There. Now, you've gone and done it.

Effieciency is a meaningless metric. I believe I'm talking about aggregate throughput of the memory subsystem. Using the Dual DDR topology, the Intel 865/875 acheives 25-75% (according to SiSoft and PC Mark) more throughput than and AMD XP 3200+ on an nForce2 using Dual DDR 400.

Now, do you maintain that the reason is that the P4 is <i>requesting</i> more data per clock?

Compare the benchmark results for a P4 on an i845 DDR (or even an i850) to a P4 on an i875. See a difference? An alternate explanation could be that Intel did much more to the P4 revision C than revise it to operate with a quad-pumped 200 MHz FSB. I doubt it.

Regards,

Dave

Intel giveth and Microsoft taketh away.
 
Itanium, at 1GHZ has the FPU of about 3 Athlon 1.4GHZ together.
Click to expand...
That is correct. Itanium's architecture is a very respectable one. Its utilization of resources at hand is most impressive. A 1.5Ghz <b>Itanium</b> has FP performance that is <b>2.5 times</b> the performance from a 2.2Ghz <b>XP3200+</b> from AMD. And AMD's processors have high IPC! IPC king is definitely Itanium.

Now that Intel has shown us that they can design a devastatingly powerful IPC, they might as well USE that knowledge on Prescott... 😡
 
For the last 3 reply

Sisoft is big scrap there some much better software.Nividia chipset only reach 1.5 GB/S largely under old I850E.Remember I875 on 800FSB vs 3.06 on 533 RDRAM.Only 5% distance bolt.

Instruction parralle or ILP is made at the compiler level for there native ISA.

K7 can decode 3 simple instruction per cycle but 1 complex instruction per cycle like P4 or P3.

Longer the pipeligne more stall you will got or hole in a pipeligne.

IPC form P4/K7/P3 is pretty much the same only the intruction lantency differ the longest is P4 K7 and P3 are equal even if K7 it got longer pipeligne.The rest it a result of data logistique more cache faster cache less ram lantency and more bandwith.SMT can change something on ILP.The rest dont change much.

[-peep-] french
 
One more thing:

LATENCY.

A the memeory controller on a i865/875 gives a P4C up to 39% lower latency when accessing system RAM.

Look <A HREF="http://www.anandtech.com/cpu/showdoc.html?i=1815&p=6" target="_new">here</A> and <A HREF="http://www.aceshardware.com/read.jsp?id=55000278" target="_new">here</A> (also more ammo to my bandwidth arguement).

Advantage me.

Intel giveth and Microsoft taketh away.
 
AMD's PR rating is invalid.
Click to expand...
Well, if it is indeed a comparison between AMD's current CPUs and some other CPU (like the T-bird 1400 or something) then the sad truth still remains: AMD has been technologically defeated at this point. Its top-of-the-line, max-price niche is completely overpowered, regardless of performance ratings or Mhz/Ghz. Maybe A64 will give them a breath of fresh air.

But right now, AMD's main problem is that nothing will change, because no new processor will come to anyone's aid right now. No new Barton, no new CPU from AMD for the next few months...
 
Now, do you maintain that the reason is that the P4 is requesting more data per clock?
Click to expand...
I certainly do. Intel designed the P4 to fetch 64KB lines instead of 32KB like the AthlonXP.
Imgod2u explained this months back, wish he were here to describe it more accurately than I.

Knowing the P4 NEEDS info in its long pipeline at all cost and knowing pipeline stalls cost a lot on the P4, fetching more memory at once is A MUST. It's not because it needs so much bandwidth to perform well. Otherwise explain why a K7 needs twice less really.
Pentium 4 averages the resulting bandwidth at not even 80% of the max 6.4GB/sec. K7 is closer.
I am basing THAT on why the nForce 2 chipset is already one of the best. Look at VIA, they can't beat the nForce 2! Yet, logically, the only real advantage nForce 2 has is Dual Channel, and even THAT can't be used, as Athlons have a maximum bus speed of 200MHZ. nForce 2 uses an optimized Dual Channel mode fetching, which only activates then, but in reality does not even touch Dual Channel, likely.

Compare the benchmark results for a P4 on an i845 DDR (or even an i850) to a P4 on an i875. See a difference?
Click to expand...
I don't understand what are you trying to say here.
That i845 is weaker than i875 for a reason?
Damn right! i875 allows 800MT, AND uses PAT, period! That's all there is man. Come on, dig up some real other optimizations for i875 and come prove me wrong.
This only proves Pentium 4 needs more bus bandwidth. Now look at Athlon, can BARELY use any more. You can overclock to a 450MT bus and use 225MHZ memory yet you will barely get more than 2% better performance. It is clear Athlons won't benefit from better chipsets that support higher bus and memory speeds and therefore if your argument about better chipsets for Intel was because of higher bus speeds (which in turn classifies them as "better chipsets"), then you're just wrong to link this to why Athlons lack chipset strength.

Furthermore, <A HREF="http://www6.tomshardware.com/motherboard/20030303/index.html" target="_new">this article</A> shows that chipsets for AMD have gone a long way and NOT BECAUSE of bus speeds and such, but because of optimizations. See how clear the performance varies, for the latest chipsets which all used almost the same bus? It is clear a lot has been spent to improve the Athlon platform, and that further chipsets won't help as much as this limit broken finally by the most recent K7 chipsets.

--
If I could see the Matrix, I'd tell you I am only seeing 0s inside your head! :tongue:
 
Pentium 4 averages the resulting bandwidth at not even 80% of the max 6.4GB/sec. K7 is closer.
Click to expand...
The SiSoft and STREAM results I've seen show Athlon around half 6.4GB/s.

You may be on to something where the AMD parts simply cannot benefit from increased memory bandwidth. However, you have shown that merely raising the clock of the memory has a barely significant increase (statistically, 2% is significant). You assume the memory controller on the nForce2 can provide more memory bandwidth, and I assume that it cannot-- <b>at a given clock speed</b>. I must admit that I am still not wholly convinced whether the memory bandwidth discrepancy is the fault of the processor or the memory controller. Hell, it could even be the benchmark code!

I'd be interested if KT600 results shed any light on this.

Dave


Intel giveth and Microsoft taketh away.
 
However, you have shown that merely raising the clock of the memory has a barely significant increase (statistically, 2% is significant).
Click to expand...

No, I was talking about the sum of both, bus AND memory upping. This is the average result moving to 333MT! It was really not that significant and very much unnoticeable. Such is what we'd get from now on if we upped the bus and memory throughput. It simply isn't anymore helpful. The Athlon doesn't need to fetch more data if its pipeline has few stalls and twice less stages.
Not only this but its clock speed doesn't even need (for its current 32KB fetch mode) more bandwidth. Perhaps at 3GHZ.

Yes a KT600 benchmarking is in order.

--
If I could see the Matrix, I'd tell you I am only seeing 0s inside your head! :tongue:
 
Now that we finally agree on something (KT600), it seems dubious to further belabor this . . .

BUT

If memory badwidth improves via the KT600, then do I get a cookie?

Dave

PS If it doesn't, its because VIA makes crappy memory controllers. Yeah, you heard me! I'm never wrong. Except that one time. . . Damn you, Ross Perot!

Intel giveth and Microsoft taketh away.
 
Ahhh Ross Perot, a people's man.... lol

I dunno about a cookie, I could give you a free pass for a day with Wingding if ya want.

--
I am my own competition. -VJK
 
For clarification, the cacheline size on the P4 is 128KB (2 stripes of 64KB), the cacheline size on the Athlon is 64KB.
As far as memory subsystem, it would depend on what you're doing. Programs with high data locality will benefit from the P4's superior memory bandwidth and cacheline size. Programs with low data locality will just waste memory bandwidth with no benefit on the P4.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom