Winny - Re:"If you want to get some work done faster, get an AMD CPU."
Not according to this, Pooh.
Remember this dude who trashed the Pentium 4 when it first came out? Mr. Emulator was a former severe critic of the P4 and a favorite of the 'Droids.
This article was based on a test of a 2.53 P4 system. Imagine how the newly introduced 2.8 P4 will thrash the AthWiper.
How do you say "turn out the lights, the parties over" in Brit?
See original for long article with all the details. It's a good read.
vivisimo.com
Pentium 4: Round 4 - Pentium 4 hammers the Athlon XP
Copyright (C) 2002 by Darek Mihocka
President and Founder, Emulators Inc.
Updated May 6 2002
--------------------------------------------------------------------------------
Older postings: Round 1 Round 2 Round 3
Are you surprised?
The 64-bit hype and why you shouldn't hold your breath
How the Pentium 4 took the lead
Dual processor shootout - is a second processor right for you?
Single processor benchmarks
--------------------------------------------------------------------------------
Are you surprised?? Athlon stopped improving and got blown away!
Just before April 1st, I was rolling on the floor laughing when the irony of this whole situation dawned on me. About 6 months earlier I posted the results of the Round 3 benchmarks and declared the Athlon XP the faster chip over the Pentium 4. AMD freaks cheered and went back to playing their video games.
Those with the short attention spans apparently did not bother to read the fine print in my writing and see some ominous warning signs that I was pointing at last year. i.e., that it was within Intel's grasp to catch up to and overtake AMD in the CPU battle if they so chose, and that AMD was doing dumb things to hurt itself. Dumb advertising. Not making some improvements. Not delivering product.
How poetic then that Intel should reclaim the lead from AMD just around April Fool's Day. When we and other web sites broke the news, AMD users couldn't believe it. I was accused of posting an April Fool's Joke! Even received threats from some idiots if I did not remove the "joke". But it's no joke. The AMD Athlon (whether you call it the MP, XP, 4, Palamino, Thunderbird, etc) has lost its lead. Big time.
Not only does the 2.4 GHz Intel Pentium 4 take the lead, but the 2.53 GHz Pentium 4 running an even faster 533 MHz front side memory bus is already now out and extends the speed lead even further to the point where AMD chips are just left eating dust.
And here is the news that even surprised me when I tested the system. While Intel has been pouring all of its marketing money in the Pentium 4, it HAS done what I predicted in October they would do, which is to release faster versions of the Pentium III. In January, nobody noticed (even I missed it) when Intel quietly shipped the 1.4 GHz Pentium III-S with 512K of L2 cache. I've got a couple of these chips now running with a dual processor board, and even at 1.4 GHz, the chip in some cases beats the Athlon XP 1800.
What happened? In a one line summary it's quite simple: Intel caught up as expected, and AMD fell asleep at the wheel and blew the lead. Plain and simple, and it wasn't exactly unexpected. In April, I was certainly not alone in reaching the conclusion that Pentium 4 is now the fastest processor. Immediately after the launch of the 2.4 GHz Pentium 4 in early April 2002, the well known sites Tom's Hardware and AnandTech similarly declared the new Pentium 4 to be fastest chip around. They used the more standard suite of benchmarks to reach their conclusion. Even TweakTown is not holding out much hope for AMD.
Anyone can run a benchmark. I like to know WHY the results are what they are, and as always I've used a combination of real world benchmarks and low level analysis of the processors to reach my own conclusions.
If you're having trouble understanding how it is that I turned from calling for an all out boycott of the Pentium 4 back in December 2000 to telling you to buy one now in 2002, you've obviously not been paying attention. If you read through to the end of my Round 3 report, you will see that back in October 2001 I wrote this about the Pentium 4:
"Will be a far more attractive processor coupled with DDR SDRAM and .13 micron based speeds beyond 2.0 GHz... Given that next year's Pentium 4 will likely use a DDR SDRAM chipset, any investment in Pentium 4 today is money wasted."
So in other words, while last October I told you to buy an Athlon XP and not waste your money on Pentium 4 systems available at THAT time, I also hinted that all bets would be off come this year, 2002. In fact in January Intel delivered on all three things I hoped for: shipping a .13 micron based Pentium 4 processor, at speeds above 2.0 GHz, and supporting DDR SDRAM. That is why in January I did give the Pentium 4 my thumbs up. Apparently the AMD fanatics who read my postings failed to notice this too, as nobody bothered to send me a death threat in January, only in April when I made the more explicit statement that the Pentium 4 was now faster than the Athlon. Stupid, stupid, stupid AMD users. Dumb as nails, and all this time I thought they were buying AMD based on technical data. But it turns out AMD users are mostly a bunch of Intel-hating technology novices. So please, spare yourself the effort and do not send me yet another "OH MY GOD YOU ARE SO EVIL HOW COULD YOU TURN ON AMD" email. The writing has been on the wall for a long time.
What I find ironic is that during the year and half battle between the Pentium 4 and Athlon that AMD and Intel have practically swapped places. Intel is in first place again, and AMD is the #2 bitch. How different this is from December 2000!
If you go back to my Round 1 posting from December 2000 and study the issues you'll see how this has slowly happened. Let's review the facts:
The Pentium 4 processor was launched at clock speeds of 1.4 and 1.5 GHz in November 2000. While running at a significantly higher clock speeds than 1000 MHz (1.0 GHz) Pentium III and Athlon processors of the day, it failed to deliver proportionally faster speed. In fact in some cases the Pentium 4 was actually slower than the older 1.0 GHz processors, not 40% to 50% faster as clock speed would predict.
After a month of studying the new chip and running all sorts of software tests on it I reached the conclusion that 8 major factors contributed to this paradox that allowed the "slower" Athlon to beat the "faster" Pentium 4. Among the reasons - the removal of the barrel shifter which is critical to many address calculations and integer operations, and a smaller cache than the Athlon.
By Intel's own specifications, the Pentium 4 running at 1.5 GHz was supposed to deliver about 20% faster integer performance than a 1.0 GHz Pentium III. In other words (if you do the math), adjusted for clock speed the Pentium 4 was to be 80% as efficient as the Pentium III. Thus Intel had hoped for the 1.5 GHz Pentium 4 to be the natural successor to the Pentium III by delivering on the order of 1.2 GHz Pentium III performance. However, in actuality the performance was more that of a 1.0 GHz Pentium III, or only about 65% to 70% as efficient.
Due to last minute cuts in the design of the processor, the Pentium 4 fell about 20% short of its designed speed goal. To further this problem, Intel bet the farm on a new memory technology called Rambus, choosing not to go with the more popular PC133 SDRAM memory used by both Pentium III and Athlon processors. Rambus memory cost about 4 times the price of PC133, and added to the ridiculously high introductory price of the Pentium 4 chip itself, a typical Pentium 4 system cost about $4000 around Christmas 2000 - double or even triple a comparable Pentium III based PC at the time. At around the same time, the AMD Athlon did start shipping in 1.1 and 1.2 GHz speeds, thus completely overtaking the Pentium 4.
If history was any indication, I predicted it would be a full year or two before Intel would fix the problem (which would entail a brand new processor and likely a new motherboard) and before compiler technology would be supporting the Pentium 4 which in turn would allow software developers to ship Pentium 4 optimized applications. The reasons for this delay have to do with the fact that silicon is not like software - you can't just post an overnight patch. And even the thousands of software developers out there can't just rewrite and recompile their software overnight.
These are the reasons why in December 2000 I called for people to avoid (boycott in fact) the Pentium 4. It was neither faster nor cheaper than existing products and was doomed to be that way for the next year or two. Only 3 months later in March 2001 when I posted the Round 2 results that not much had changed. The Pentium 4 still sucked as expected. Duh.
It was in Round 3 which I posted in October 2001 (almost a year after the Pentium 4's launch) that I had listed a number of warning signs that popped up between June 2001 and October 2001 that ultimately would lead to the day when the Pentium 4 would take over.
June 12 2001 - The Day That AMD Faltered
In Round 3 I mentioned my very disappointing day on June 12 2001 when I attended the AMD Road Show here in Bellevue. It came only a few days after AMD announced its new Athlon core (the Palamino) which was supposed to push the Athlon into the realm of higher-speed lower-power multi-processor computing. We'd have Athlons in notebooks, Athlons in dual processor workstations, etc. etc. That was the hype. Yet AMD came to this Road Show without the new product! Not only did they not have any Athlon Palamino powered notebooks to show but they didn't even have the dual processor Palamino machine to show. Not even a sample of the Palamino processor which were supposedly shipping already a week earlier. To add insult to injury, the company rep for Tyan, which manufactured the one and only motherboard for the Palamino at the time, did not show up.
Hundreds of computer resellers and system builders, including myself, from all over the Seattle area showed up to this all-day conference to basically see... nothing! Without anything new to show, the AMD reps merely rehashed some PowerPoint slides which were already posted for the general public to read on AMD's web site. I know that many people left the grueling 5 hour ordeal very disappointed.
When one of the resellers questioned the rep as to why the dual processor Palamino chip (the Athlon MP chip) was being released at 1.0 and 1.2 GHz clock speeds and not at the 1.4 GHz speed of the older Athlon, he gave a cryptic response along the lines of that they want to make sure it works at the slower speeds first. Bullshit folks! In hindsight, big red warning flags should have been going off.
Two other claims that the AMD rep made - that other motherboard manufacturers would be releasing dual processor boards by August based on a cheaper 760 MPX chipset, and that the new "Hammer" chip would be out by the end of they year (2001) or early this year (2002) - both failed to come true.
In fact, the widespread release of the MPX chipset boards for the Athlon didn't even happen until January 2002. And by latest news reports, the Hammer is now delayed an additional year into 2003. Could this be among the reasons that Dell chose not to sell AMD based products and why Gateway decided to drop AMD last summer? Hmmmm.
Failed to Meet Expectations
The single processor Palamino chips, the Athlon XP, did finally ship in October, and as I wrote in Round 3 they were faster. But, the reasons that the Athlon XP won the tests had less to do with any kind of major redesign of the chip than simply that the Athlon XP was clocked about 10% faster than earlier Athlons. In fact, I pointed out in October that some of the supposed speed increases just plain weren't there to be seen. The supposed hardware prefetch - I couldn't measure it. This "Quantispeed" crap? What was that? AMD had simply cranked up the clock speed, which I later confirmed by doing low level analysis of the instruction execution and finding that between the Athlon Thunderbird and Athlon XP, instruction timings remained identical. It was the same chip, built on the same .18 micron process, but with a fancy new name. Yay.
Other dark clouds were also looming on AMD's horizon last October. The Pentium III Tualatin for one thing. It was Intel's first .13 micron based processor, and as my tests showed, the 1.2 GHz part easily kept up with a similar 1.2 GHz Athlon. And then on top of that it could easily be over-clocked by 20% without overheating or crashing after days of continuous use. I have been running Windows XP now with Prime95 going non-stop for 6 solid months 24 hours a day without a single crash of lockup on the Tualatin.
With results like that, of course Intel was going to switch the Pentium 4 over to .13 microns as soon as possible and give the Pentium 4 the same kind of huge speed boost it gave the Pentium III.
Near The End Of Its Life Cycle? Where is the .13 micron Athlon?
Tests that I have been conducting on the various Athlon chips (from the 600 MHz model to the Athlon XP) indicate that almost every single machine language instruction takes the exact same number of clock cycles, regardless of the model of Athlon processor. Almost no deviation from that for over 2 years.
The Athlon design is now approaching almost 4x its original launch speed, which means that the architecture is only going to scale so much further (especially at .18 microns) before hitting the same kind of brick wall that the Pentium III hit 2 years ago.
Is AMD doomed to repeat Intel's past mistake? Did they so heavily bet on the Hammer being ready that they didn't plan to pushing the Athlon past 2 GHz. A similar mistake as Intel made with the Pentium III hitting a brick wall before the Pentium 4 was ready.
It took Intel an entire year to scale down the Pentium III to .13 micron and re-release it in the summer of 2001 at 1.2 GHz clock speeds. Could this explain why in the past year, AMD has only managed to push the clock speed of the Athlon from 1.2 GHz to 1.733 GHz? A 45% improvement is decent, but far short of the 60% needed to keep up with Moore's Law. While the Athlon has gained 23% clock speed (from 1.4 GHz last summer to 1.73 GHz today) the Pentium 4 has delivered about a 40% speed increase in the first 5 months of 2002 alone thanks in part to the .13 micron Pentium 4 Northwood that shipped in January and the huge increases in clock speed from 2.0 GHz last fall to 2.53 GHz now.
AMD has countered this attack from Intel NOT with product, but with just more press releases telling us how wonderful their products will be in 2003. Again, how ironic, as this is sort of what Intel has had to do with the Pentium 4 for a year.
Yes, in April 2002 AMD did ship its first .13 micron Athlon part, a notebook processor running at something like 1.3 GHz. Big deal, Pentium III notebook processors were there last year, and they run faster. Where are the chips for the desktop? Where is a nice fast 1.8 or 2.0 GHz Athlon chip?????
Stability issues
My last 4 Athlon systems have now all either died or failed to work properly during the past 6 months. I fail to imagine how this can be a coincidence, as during this same 6 month period I built or purcahsed a number of Pentium III systems (dual Pentium III/1000, a Pentium III-M/1200, and a dual Pentium III-S/1400) and more than half a dozen Pentium 4 systems ranging from 1.4 GHz systems to a dual Pentium 4 Xeon to a 2.2 GHz Pentium 4. All of the Intel systems worked first time, and continue to work today.
This is the part where I should have given more thought to the strange answers that the AMD reps were giving at the AMD Road Show. Especially the part about them wanting to make sure the Palamino works at lower clock speeds first. Is it no wonder then that I've had stability problems with all 3 of the Palamino systems that I've built and the one Athlon Thunderbird system I purchased, but comparatively zero problems with any of the 7 Pentium 4 systems and many other Pentium III systems that I now own.
Even today, AMD sells a 1.733 GHz Athlon XP/2100 chip for single processor use, but only a 1.6 GHz Athlon MP/1900 chip for dual processor use. WHY? The XP and the MP are practically the same chip, just as the Pentium 4 and the Pentium 4 Xeon are. Yet Intel last month shipped both a 2.4 GHz desktop Pentium 4 chip, and a 2.4 GHz dual capable Xeon chip. The Pentium III is sold at 1.4 GHz speeds, and works just fine in a dual processor configuration. I know, I just built such a system! Why then can AMD chips not run as fast in dual processor mode as in single processor mode?
People have said that the Athlon is not as stable as the Pentium III. Up until 6 months ago, I would have (and did!) strongly argue against that statement, as my past experience with the Athlon chips (at clock speeds of 1.2 GHz and below) indicated no problem.
But starting with the Athlon XP 1800 (that I won from the back of that AMD truck on October 9 2001) and then with the Athlon XP 1900 chip a month later, and then the dual Athlon MP 1800 that I more recently built, and finally the Athlon Thunderbird 1200, I've had strange problems.
The Athlon XP 1800, having come with a nice MSI motherboard as part of the package, I can say is almost fully stable. Not quite. About once a week, the machine just randomly reboots. No blue screen in Windows XP. Just boom, black screen and reboot as you're using it. Otherwise it's a speed demon.
The Athlon XP 1900 problem I've documented on the Secrets page a few months ago, and that problem is that I am unable to get the chip running faster than 1200 MHz. It's a 1600 MHz part, yet when I try to boot it at 1600 MHz on the ASUS A7A266 board, it just plain won't boot, even though I can pop in an Athlon Thunderbird part and run it faster than 1200 MHz on the same board. I've had two other readers report similar problems to me of not being able to get the AMD chip to run at full clock speed. I only got the Athlon XP 1900 chip to work by putting it in the MSI motherboard that the XP 1800 was in. Obviously ASUS made a board that wasn't Athlon XP compatible, yet claimed was!
So when I read that dual processor Athlon MP boards were finally available in January, I went to AMD's web site and looked at their recommended boards. They specifically recommend the ASUS A7M266-D board for dual processor Athlon MP system. So I bought that board, and a pair of Athlon MP 1800 processors, and 2 gigabytes of real Kingston brand ECC registered DDR2100 SDRAM.
I put the system together. Windows Me installs and runs in single processor mode. I get speeds similar to the Athlon XP 1800 as expected. Windows 2000 installs and runs. Both processors work fine. So far so good. I go to install Windows XP. Boom. It blue screens during setup. I've tried and tried and tried. I just couldn't get Windows XP installed on this puppy. Only after tweaking the BIOS settings and basically crippling some of the options did I get XP to install.
And now most recently, on April 24 2002, as I was in the process of re-running all the benchmarks that I had gathered over the past few months in order to check and double check benchmark results, the year-old 1200 MHz Athlon Thunderbird system was dead. No boot, nothing. 4 Athlon systems, 4 problems.
It All Adds Up - Pentium 4 is now the better choice over the AMD Athlon
contiued on the web |
