Introduction
I’ve promised this 2nd addendum to my initial article ‘New Hope for Athlon — VIA’s Apollo KX133 Chipset‘ last Saturday and I had the results ready for publication by Sunday. The results are far from exciting, which is why I decided to also include information about further findings of the AGP-issue within the GeForce drivers. The testing of the AGP-stuff is not finished yet though, which would have delayed the posting of this article by a bit.
Is Super Bypass the Holy Grail?
I finally decided against waiting until the AGP-testing is finalized and will now simply post the added benchmark data of a Gigabyte GA-7IX-motherboard with enabled ‘super bypass‘. The reason behind this decision is the high interest into those ‘super bypass results’ and I appreciate that. I would also appreciate however, if people would understand that ‘super bypass’ is only a little feature of Irongate and not the Holy Grail. Irongate is logically inferior to KX133 and the facts should really speak for themselves. Irongate lags behind KX133 by 25% in terms of memory bandwidth and pretty much 75% in terms of AGP bandwidth. My AGP-testing has pretty much shown that all graphics cards I’ve tested only run at AGP1x on Irongate platforms, which seems to be due to instability issues of AMD750 running at AGP2x. This should neither upset nor surprise anybody, because Irongate has never been a particularly mature and highly sophisticated chipset in the first place. Why do you think did it take several months until AMD could get a grip on the ‘super bypass’-bug in Irongate-revisions 1-4? ‘Super Bypass’ was meant to be implemented into Irongate from the very beginning. Why do you think that AMD hoped that KX133 would be ready at Athlon-launch? Because it was known that Irongate was not exactly the most sophisticated chipset for Athlon. Until now it was the only one though.
Now KX133 itself is certainly not the Holy Grail either, as I already pointed out in the initial article. However, most benchmark-data still speaks for KX133 vs. Irongate, as you will see in the following results.
Benchmark Setup
| Hardware Information | |
| CPU | AMD Athlon 800 |
| KX133 Motherboard | VIA VT5249B1 Reference Board, BIOS date Jan 4, 2000 |
| Irongate Motherboard | Gigabyte GA-7IX, BIOS Dec. 1999 |
| Memory | 128 MB Micron/Crucial Technologies PC133 CAS2 |
| Network | Netgear FA310TX |
| Graphics Card | NVIDIA GeForce256 DDR Reference Board |
| Driver Information | |
| GeForce256 | NVIDIA Reference Driver rev. 3.68 NVIDIA’s Special KX133 Driver, rev. 3.76 |
| KX133 Chipset Drivers | VIA 4in1 4.19 |
| AMD AGP Miniport | 4.45 |
| AMD Bus Master IDE Driver | 1.24 |
| Environment Settings | |
| OS Version | Windows 98 SE 4.10.2222 A / Windows NT SP6a |
| DirectX Version | 7.0 |
| Screen Resolution under Win98 | 1024×768 |
| Color Depth | 16-bit |
| Screen Resolution under NT | 1280×1024 |
| Color Depth | 32-bit, true color |
| Refresh Rate | 85 Hz |
| DirectX Version | 7.0 |
| Quake 3 Arena | Retail version command line = +set cd_nocd 1 +set s_initsound 0 |
| Expendable | Downloadable Demo Version Command Line Setting ‘-timedemo’ |
| Descent III | Retail version Settings = -nosound -nomusic -nonetwork -timetest |
| Quake 2 | Retail Version, 3.20, plus AMD 3DNow!-Patch |
Office Application Benchmarks
With enabled ‘super bypass’ Irongate scores identical to KX133. AGP-performance and memory bandwidth have little to no impact on office applications, as already mentioned in the initial KX133-article.
Those results were run with the new NT-driver for GeForce, described in the first addendum to the KX133-article. As you can see KX133 has still got a slight edge over Irongate with enabled super bypass, but S/B does definitely make a difference here.
3D Gaming Benchmarks
In Quake3 Irongate can logically not reach the performance of KX133. Irongate is damned to run AGP1x.
Quake 2 doesn’t seem to be particularly impressed by ‘super bypass’, so the lead of KX133 still remains.
Super bypass is able to give Irongate a boost in Descent3, but it’s not enough to reach KX133.
The same story is valid for Expendable. Super bypass makes Irongate faster, but KX133 can retain its lead.
OpenGL Benchmarks
SPECviewperf running under Windows98 shows the AGP1x-problem of Irongate quite clearly. It’s also interesting to see that super bypass doesn’t improve the viewperf-scores or Irongate under this operating system.
Unter NT super bypass is able to help at least a little, but the lead of KX133 with the new AGP4x-enabled GeForce-drivers is far beyond reach of Irongate.
Conclusion
I hope all the upset souls have finally calmed down. ‘Super bypass’ is certainly an important feature for Irongate, but it’s not making a Porsche out of a Chevy. Anyone who is in the process of buying an Athlon-system or platform should make sure that it carries the KX133-chipset. It’s a matter of fact that Irongate has just become obsolete. This may be a bitter pill to swallow for all Irongate-owners, but those should please appreciate that in the PC-world there’s always something better and faster around the corner. Changing from Irongate to KX133 is a rather questionable move though. Unless you are using OpenGL development software you might want to wait until the next Athlon-chipset that supports DDR-SDRAM.
