I doubt that I am telling you anything new by pointing out that the last 12 months saw a major change of who is supplying the majority of high-end x86 processors. While Intel used to own this position for over a decade, AMD has successfully taken that place away from once almighty Chipzilla and the guys in Satan Clara (yeah, I know it's "Santa Clara") can only hope that the upcoming Pentium 4 will finally change this fact.
The major problem that keeps Intel from playing a significant role in the Giga-Hertz-and-beyond-segment is not an inability to actually release processor solutions that would be fast enough, but Intel's utter inability to provide those wonderful products in actual volume. When AMD released its Giga Hertz Athlon in March this year Intel was able to counter with the release of its Giga Hertz Pentium III two days later (please read The Giga Battle and The Giga-Battle Part 2). However, until this very day Intel continued its inability to ship this processor in volume and only recently you could finally find some Pentium III 1 GHz processors at retailer shops. In those last 5 months the vast majority of 1 GHz systems shipped with AMD's 'Athlon inside', which made many journalists use the term 'paper release' for Intel's Giga Hertz Pentium III release.
AMD started to take its newly achieved position as provider of super-fast x86-processors that seriously, that it didn't panic after Intel's recent release of the Pentium III at 1.13 GHz on July 31. Instead of launching their counterpart, the Athlon at 1.1 GHz, right away, it continued to ramp up production for this product and stick to the planned release date of August 28, 2000. This is why you can find this new AMD processor at PriceWatch already now. Different to the anyway rather flaky Pentium 1.13 GHz from Intel, you can actually either buy Athlon 1.1 GHz processors on their own or in systems right now.
The Specs of AMD's New Athlon/Thunderbird Processor
I don't want to bore you with stuff you have read and heard a hundred times before, so I'll keep myself short. AMD's new Athlon 1100 is another version of AMD's Thunderbird core. This core is based on the original Athlon architecture, but different to its older brother for SlotA it comes with its L2-cache combined with the CPU-core on the same piece of silicon. This silicon piece, or 'die' happens to consist of about 37 million transistors, its package is the now well-known 462-pin SocketA-PGA and its favorite platform is based on VIA's Apollo KT133 chipset.
The new 1.1 GHz Athlon requires the same core voltage as the Athlon/Thunderbird at lower clock speeds, which is 1.75 V. Its thermal power dissipation is 1 W higher than what we know from the Thunderbird at 1 GHz, which in return means that you won't need a significantly altered cooling solution over what is used for the first Giga-Thunderbird.
AMD's new Athlon 1100 has one invaluable advantage over Intel's recent Pentium III 1.13 GHz. Different to Intel's desperate 'beyond-giga' solution the new Athlon does not require anything different than its less powerful brothers. The Athlon 1100 runs on ANY SocketA platform, while Intel guarantees the, by me yet unseen, 'stability' of the 1.13 GHz Pentium III only on ONE specially modified VC820 motherboard. Athlon 1.1 GHz is allowed to get as hot as all its slower siblings, while Intel tries to insure some reliability of it's 1.13 GHz solution by specifying the maximal 'junction temperature' down to 62 degrees centigrade, which in return requires massive cooling solutions.
Finally there is one more reason to favor AMD's new processor - the price. I already mentioned that I found Athlon 1100 at PriceWatch. With 719 US-Dollars it's almost $250 cheaper than the best offer for a Pentium III 1 GHz at US$ 962! I guess it's needless to add that no retailer in the world is selling a Pentium III 1.13 GHz. This miraculous piece of, excuse me, silicon-trash only ships in a few OEM boxes for the gentleman who's got really everything, even the silver telephone dialer.
I decided to run this new CPU against its slower brothers, against Intel's Pentium III and against Duron and Celeron. All CPUs had to be faster than 900 MHz though, which required Duron and Celeron to be overclocked.
For all AMD processors I used a SocketA platform with VIA's Apollo KT133 chipset from Asus, the A7V. The recent BIOS update 1003 produced a surprising performance gain, particularly under Sysmark. All Intel processors were run on an i815-platform, the Asus CUSL2.
I was unable to include benchmark results of a Pentium III 1.13 GHz, as I am not in the possession of a sample that works reliably. I also would have had to run it on Intel's customized VC820 motherboard, which would have made this CPU look rather sad against its brother at 1 GHz on a i815 platform. Believe me, you don't miss anything.
A Celeron 667 was overclocked to 950 MHz on the CUSL2, which happened to be a very easy job. The latest BIOS 1002beta2 from Asus' FTP-server gives you the option to directly choose the processor clock and it automatically adjusts FSB, memory and PCI clock for you. At 950 MHz the Celeron 667 ran with 95 MHz FSB, 142 MHz memory clock and a hefty 47 MHz PCI clock, which surprisingly didn't cause any problems whatsoever. Thanks to the high memory clock the Celeron 950 performed rather impressively.
For the Duron 900 and 950 I overclocked a Duron 700 using the dipswitch overclocking feature of the Asus A7V. I am aware of the fact that overclocking Durons has become much more difficult these days and we will dedicate a new article to this topic very soon. The Athlon 1100 needed to have its L1-bridges closed in order to be overclocked to 1150 MHz. You will notice that the Athlon 1100 overclocked to 1150 MHz was never able to pass the Linux kernel compilation. This is the only test that the overclocked CPU was unable to finish. It shows that the kernel compilation is a very good test to check if your overclocked processor runs reliably. I was unable to overclock my 1 GHz Pentium III to anything more than 1035 MHz, which is hardly worth talking about, and so I omitted the results of this 3%-overclocking.
I compared the processors under Windows98 using Sysmark2000, Quake 3 Arena, Unreal Tournament, 3D Studio Max 2 and SPECviewperf 6.1.2. Additionally all CPUs had to finish a timed Linux kernel compilation.
New Price-Performance Evaluation Added
For the first time I decided to include price-performance charts. The prices were taken from PriceWatch on August 28, 2000.
|SocketA Platform for AMD Athlon and Duron Processors
||Asus A7V, BIOS 1003 final
|Socket370 platform for Intel Pentium III and Celeron processors
||Asus CUSL2, BIOS 1002beta2
||128 MB Wichmann WorkX PC133 SDRAM CL2, setting 2-2-2-5/7
|Hard Drive for Windows 98 Tests
||IBM DTLA-307030 ATA100 IDE, 30 GB, FAT32
|Hard Drive for Linux Test
||Seagate ST320430A ATA66 IDE, 19 GB, ext2
|Graphics Card for Sysmark2000, Quake 3 Arena, Unreal Tournament and 3D Studio Max 2
||NVIDIA GeForce 2 GTS Reference Card
Core Clock 200 MHz, Memory Clock 333 MHz, Driver 6.16 Detonator 3
|Graphics Card for SPECviewperf
||NVIDIA Quadro2 Reference Card
Core Clock 230 MHz, Memory Clock 400 MHz, Driver 6.16 Detonator 3
|Windows Resolution for Sysmark2000
|Windows Resolution for SPECviewperf
||SuSE Linux 6.4, Kernel 2.2.14, THG benchmarking kernel, gcc 2.95.2
|Quake 3 Arena
||Version 4.28 (patched)
|3D Studio Max
Office Application Performance
AMD's new Athlon 1100 takes the lead under Sysmark 2000. Who knows, maybe a working Pentium 1.13 GHz would be faster, but why would you care, since this processor is hardly available and seems to be a rather dangerous investment on top of that.
Let's have a look at the price-performance ratio for the first time:
It is rather obvious that you are getting by far the best bang for the buck if you should succeed to overclock a Duron processor. Besides Duron, Celeron 667 at 950 MHz is also a rather attractive option. The SocketA-Athlon 900 offers the best choice for non-overclockers who want to pay a fair price for their CPU. Its price/performance ratio is double the score of a Pentium III 933. The new Athlon 1100 offers of course the worst price/performance ratio of all AMD-processors, but it still looks better than the Pentium III at 1 GHz.
3D Game Results - Quake 3 Arena
Again Intel might be able to take the lead if its Pentium III 1.13 GHz was a reliable part. AMD's new Athlon 1100 reaches the top scores instead.
The picture won't change much in the coming tests as well. The overclocked Duron shines due to its low price and great performance, and then comes the overclocked Celeron. The worst ratio is reached by the highly overprized Intel Pentium III 1 GHz.
3D Game Results - Unreal Tournament
The differences between all the processors is rather small, which is why Unreal Tournament should not really be a reason to buy a beyond -Giga processor of either CPU maker.
Obviously the price difference between the processors is rather huge while the performance under UT turned out to be rather small. Thus we simply see an amplification of the previous price/performance results.
Linux Kernel Compilation
The kernel compilation under Linux with gcc 2.95.2 is clearly ruled by the AMD-processors, which I consider as rather surprising. The compiler runs pure integer-instructions. Neither FPU-performance, nor SSE or 3DNow! can help here. Still Pentium III loses.
Once more Pentium III is looking rather bad in this comparison, but what really surprises is that AMD's Athlon /Tbird 900 is almost reaching the highly overclocked Celeron in price/performance. An overclocked Thunderbird 900 would easily be an even better value.
3D Studio Max 2
This benchmark is already a classic here at Tom's Hardware. While it used to be ruled by Intel processors in the past, it's now dominated by Athlons. The FPU of Pentium III doesn't have the slightest chance against any of the AMD processors.
If you are into 3D-modelling you might want to have a look at this chart, although it doesn't tell the whole truth. SMP systems could score particularly well here, especially systems equipped with two inexpensive processors. Intel's processors are obviously the only x86 SMP-solution for the time being, as long as AMD is still not offering SMP-platforms for Athlon.
OpenGL Workstation Performance
OpenGL Workstation Performance, Continued
The latest SPECviewperf seems to focus on the performance of the OpenGL-card rather than the processor. The difference between the CPUs is not very impressive. Still AMD's new Athlon 1100 is looking very good.
Light-04 is of all the benchmarks in SPEC's new viewperf suite the one with the highest processor impact, although it is still not exactly a remarkable distance between the fastest and the slowest processor in this comparison. As a result, the processor price is obviously defining the scores.
There you have it. AMD has simply taken the next step and brought the Thunderbird core to the next speed grade. It is good to see that this did not require any strange tweaks to the core or to the platform. Basically, the 1.1 GHz Athlon is 'just another Athlon' really. This can only be a good thing compared to Intel's 1.13 GHz Pentium III, which turned out a rather troublesome solution. It is rather obvious that the Thunderbird core offers a lot more clock speed headroom than Intel's Coppermine-core. However, while Coppermine will soon be replaced by 'Willamette' or 'Pentium 4', Athlon will have to stick around a bit longer. AMD will probably be forced to put an even faster Thunderbird up against Intel's future Pentium 4, unless 'Mustang' should indeed be ready very soon.
The same that I said in the review of Intel's Pentium III at 1.13 GHz is obviously valid for AMD's new super-processor as well. Make sure you've got software that justifies the expense!!! I wouldn't know a lot of people that really need this processor right now. However, if you want to go for a top-notch x86-CPU, I'd rather go for AMD's offering. The reason is pretty simple. First of all it seems rather obvious that the Athlon/Thunderbird 1.1 GHz is the more reliable and more flexible solution of the two. It also offers excellent performance and it comes for a much more attractive price than Intel's counterpart.
I am still wondering if Intel won't retract the 1.13 GHz Pentium III very soon. If you are in doubt have a look at Kyle Bennett's article about the Tri-1.13 GHz testing he just published today. I consider it as a mere question of time until even Intel's not-so-infallible engineers will finally realize that there's something seriously wrong with Intel's current high-end offering. I can only repeat once more that the upcoming Pentium IV will hopefully have a better stability record. Wouldn't it be nice if I finally had a reason to write nice things about Intel again?
Please follow-up by reading Intel Admits Problems With Pentium III 1.13 GHz - Production and Shipments Halted.