|
|
July 21, 1998:
The W98 setup application ran on automatic pilot for the most of the install.
The installed options were limited to the "bare bones." The firstboot
into W98 failed. The system "spontaneously" rebooted prior toshowing
the desktop. Scandisk complained that the previous boot had failed.
Indeed , it had. Scandisk reported no disk errors andthe boot continued.
Then there was the choice of booting into Safe Mode. I punted and
did the manly thing -- insisted on a Normal boot. Not a great idea.
The desktop never appeared. Again, the system"spontaneously" rebooted.
The sequence continued. Once again, Scandisk did not find errors. This time, Safe Mode was the ticket. To my relief, the system booted into Safe Mode. Now what? I had not loaded any drivers. There was nothing to unload. There were no applications to remove. I could have looked at the W98 boot log, but that seemed like too much trouble at this time. Earlier, I read a UseNet post directing readers to the Via WEB sitefor information on installing W98. Via intsructs you to enable USBsupport before loading W98. I did not think much of this. I planned get the system running and then straighten out the miniport IRQand the AGP virtual device. Knowing that many users might never install these drivers, I thought that "before and after" benchmarking would beuseful information. Unfortunately, the system would not boot W98. This precluded benchmarking. For lack of a better plan, I entered the BIOS, enabled the USB function, and tried booting W98 for the 4th time. This time, the system booted in Normal Mode. Meanwhile, the FX83 was leisurely downloading the VIA minport and VxD for W98. By the time I finished cancelling W98 requests for update disks on the EP-51, the VIA updates were downloaded onto the FX83 HD. What happened next impressed me about W98. I have a network installed at home. I entered the workgroup name during the W98 setup but I did not load network protocols. I double clicked on Network Neighborhood (a stupid and patronizing icon if there ever was). Whammo, the FX83 and the HOT631 popped right up in the Neighborhood. I am glad they were there, but based on my experience with W95, this should not have worked. All systems on my home network are running W98. I took the EP51 towork and put it on a network having W95 systems. The neighborhood was dead. I loaded NetBEUI. Presto, the entire neighborhood was partying. Well, enough of that. I will decipher it later. I mapped the FX83 download directory and loaded the VIA "drivers" over the network. Networking is great for setting up new systems. The EP51 was heading for the 7th boot. Everything went smoothly. There were no OS requests for update disks. The Device Manager did not show any of those silly exclamation marks. Next, onto installing the video drivers. I was fighting off a brain seizure due to the60Hz refresh rate. The Nine 147P drivers are tricky to setup in W98. There is preliminary information one has to accept. Of course, I accept nothing without a fight. I loaded Hawkeye and the Optimized driver both of which are reproted to be incompatible with W98. I knew this to be true from installing the video drivers on the HOT631, the HX83, and the FX83. Nine's utilities and their Optimized drivers do not work in W98. Still, I had to try it. Why? Remember that I have something very important to say about the Nine AGP R3D and overclocking on the EP51. Heck, this is a diary. It can be tangential. My own experience and the experience of others indicate that the Nine AGP cards cannot run on an overclocked bus. Some folks go so far as to say the AGP graphics cards simply don't OC. I knew this was not true because I overclocked the STB Velocity 128 AGP without a problem. I have three Nine AGP R3Ds. They were all on the shelf because I could not OC them and the PCI version is faster. I was saving them for some boring "mission critical" system that would be relegated to running at specification. Let me just say, the NineAGP R3D will not not work at 75/2 on LX, FX, and HX chipsets. On the EP-51, the AGP R3D runs at 75/2 and at 83/2. This is with cheap"auction" memory. I always keep some cheap memory on hand just to convince myself that the premium for Crucial or AMM memory is worth it. I thought that perhaps the AGP port was not affected by OCing EP51. I doubt this because the percentage increase in performance for the overclocedAGP card is similar to the the corresponding increase for the PCI R3D in the HX83. Did you ever read Nine's description of installing their certified driverin W98? You might as well read War and Peace. It is worse than this diary. I gave up on reading and started double clicking. A key bit of information is that you have to modifiy the driver configuration file to accept ver 4.10 of Win. The setup program will not run unless this is done. Of course, the setup program installs Hawkeye. Trust me, Hawkeye is not compatible with W98. As soon as you load the Nine driver using setup (using setup is not recommended by Nine), delete Hawkeye from the startup directory. Also delete it using the control panel. Chances are that your next boot into W98 will be a problem. The o'l "driver not setup correctly" message is going to appear. Go back to the Standard VGA driver. Reboot. List the drivers. Now, the Nine Optimized and the WHQL (Standard) drivers should be choices. Chose the Standard driver. W98 then asks for the driver location.. Navigate to the driver download directory and let W98 do it's thing. This should get the Nine driver up and running. This system is stable, fast, and easy. So far, it is not appreciably
faster than an HX83 but, a tuned HX83 is pretty darn fast. The EP51is
more tolerant of memory than is the HX83. I tried 4 different brands
of memory all of which booted into W98 at 3.5 X 83 (Ip233MMX). The EP51
completed "all winmarks" of WB98 without incidence. It will not run
WS98. WS98 bombs at the task switching stage. To my knowledge,
no computer runing W98 completes WS98. The EP-51 is much easier to configure
at 83 than is the HX83 (unless you start out with magic memory).
|
|
|
July 24, 1998. For the past two days I benchmarked the EP51/233MMX configuration. Hopefully, I will start a benchmark results page in the next couple ofdays. There are a few notable observations. Most of these observation were made at 3.5X75 (266MHz).
|
|
|
July 25, 1998. Installing the K6-2/300 required upgrading the BIOS to the 06/29/98 version. Prior to upgrading, the system rebooted at the end of the memory check. BIOS setup was unavailable. A quick search of DejaNews identified others with the same problem. The fix involved a BIOS upgrade. I reinstalled the 233MMX in order to boot the system. The new BIOS corrected the rebooting loop. I learned later that, had I disabled USB, the K6-2 might have booted. It might not have been necessary to install a different CPU. I notice at Anand's site that he tested with an interleave of 4. My system does not boot unless the interleave is at 2. I suspect this is a BIOS issue. The praise Anand directed toward this boardis well deserved. There is plenty to say about the EP51E/K6-2/300 configuration. It cooks
and it is stable.
July 27, 1998. The system ran benchmark tests 24/7 for the past three days. Therewere no locks despite aggressive tweaking of the hardware and the BIOS. Like the HOT631, if the EP51MVP3E POSTs, it does not fail. An interleave of 4 works if Page Mode is disabled. However, thereis no performance advantage for configuring an interleave of 4. Infact, an interleave of 4 (with page mode disabled) appears to decrement disk performance. I am dubious of the compatibility of non-Intel platforms. TheEP51MPV3/K6-2 is eroding my doubts. I write, test, and use "in-house"software. Custom applications accessing port I/O work perfectly well on the EP51/K6-2. Of course, they are suppose to. Soon to be tested is a driver involving hardware interrupts, interupt service routines,and assembly language. July 29, 1998. Benchmarks were completed testing every BIOS setting in the "Chipset Setup." Not once in the many hours or testing did the EP51/K2-300 fail. The BIOS setting does not exist (for the current hardware) that will causea system fail. The stability of this system continues to amaze. Benchmark comparisons between the this system and a factory configured Dell PII300 Dimension attest to the viability of the K2. UseNet discussion suggest that the K6-2/300 lags behind the PII300. This system beats the Dell in everything (including CPU32) with the exception of floatingpoint. |
| Top of Page | July 31, 1998. Let
the overclocking begin. I received a second copy of the K6300. The
current CPU does not run at 3.5X100. The video simply goes dead a
few seconds into the post. This CPU runs at 4X83 (332) without a
problem. What's another 18MHz? In the o'l days, 18MHz was a
different world, but not today.
I tested the second K6300 at 3.5X100. It works -- almost! I violated my fundamental rule of first time testing of an aggressive configuration. I attempted to boot into W98 without first booting a bootdisk and testing Scandisk. Part way through the W98 boot, the hard drive jammed. It was hopelessly crippled. I managed to transfer the benchmarking data before reformating. There were two damaged clusters. I could either remap them or do a low level format. I hate loose ends. The low level format put everythink back in order. I reloaded W98, slowed the memory timing, disabled read around write, and jumpered for 3.5X100. The system booted to a floppy and ran Scandisk without a hitch. With my fingers crossed, reset pressed, and the floppy removed,the system booted to the hard disk. Everything works fine. I changed two BIOS settings relative to the failed attemp -- slowed theSDRAM cycle length to 3, and disabled Read Around Write. Previous benchmarks indicate that neither of these setting have an appreciable effecton performance. In other configurations, disabling read around writehas a permissive effect. This configuraton leaves little doubt that K6300 can be configured tooutperform marketed PII300 (and even PII333) systems. First run ofWB97:
|
|
|
Aug 01, 1998. This
morning's round of benchmarking caused me to go from dismay to a chuckle.
My most revealing benchmarks for the K6-2/300 are now part the the EP51benchmark
list. I refer to the 3.5X100
data showing that the overclocked K6-2 holds the PII333 at bay. These
data predict a great future for the upcoming AMD CPUs and the Epox boards.
Now that there is closure on K6-2 performance, my focus is back to the Ip233MMX. The intent is to publish the EP51/233 versus HX83/233benchmark data. I searched the database for the 3.5X83 data on theEP51. It is gone! I don't know what happened to it. Most likely, I inadvertantly backed up an old file over a new one and erased the data. I bit the bullet and swapped the 233MMX in place of theK6-2 in preparation for generating new 233MMX benchmarks. I set the jumpers to 3.5X75 and flipped the switch (pressed a key onthe keyboard). The system started to boot. The Nokia 445Xi indicated a video signal for about two seconds and died. "What in the Wide Wide World of Sports was a goin on?" I looked at my ownreview page on the EP51. Not only had I commented about 3.5X83, I said that the EP51/233MMX ran flawlessly at 3.5X83. Was I dreaming? Right now, it would not boot at 3.5X75. I removed the PC100 memory and installed Crucial 50nS SIMMS. Same result. No boot at 3.5X75. I tried specification (3.5X66). The system booted and everything was fine. What happened? Did the BIOS upgrade effect overclocking the 233MMX? Next, I tried 3X83. This worked fine. The233MMX ran at 250 with an 83MHz bus, but I swear that it worked before at 3.5X83. This same CPU runs 3.5X83 in the HX83. Have you guessed? I forgot to change the CPU voltage! For
anyone interested, the EP51 runs the 233MMX at 3.0X83 with the core voltageset
to 2.2! Once I caught my error, the voltage was set to 2.8 and the
system is stable at 3.5X83.
|
