And They All Unwind
By Van Smith
Date: August 26, 2002
As hardware sites fall under scrutiny for possible complicity in using BAPCo SysMark, some webmasters are becoming unraveled.
I want to apologize for the information disclosed here. Tom has hurt a lot of people, and he will continue to do so until someone pushes back.
I suppose I should have expected Tom Pabst, a proponent of BAPCo and its products (Tom wrote a story "vindicating" BAPCo after a heart-to-heart with John Peterson), to feel the sting of his commitments. Knowing Tom, he also craves the attention and was motivated by the knowledge that we, at times, get as much traffic as his site does -- and we are purely a volunteer site, and we don't use tricks to inflate hit counts.
Tom is a tragic figure, really, and I worked with him for as long as I did because I, for some reason, thought he would reform. Tom treats everyone around him with contempt. Tom is narcissistic, paranoid and abusive. Few people who really know Tom like him. I have seen him shout at totally innocent people in public, humiliating them for no apparent reason -- he has done this even to those closest and most loyal to him.
Except me. Tom usually left me alone because I had helped make his site credible and snatched it out of a spiraling decline. My writing brought both hits and respect to THG.
And Tom is simply not very "technical" -- which makes him all the more insecure -- and he gave me wide berth in my analyses.
Tom is also a racist. I am half-American Indian and I have relatives in Mexico. One of Tom's former employees was Mexican-American. On several occasions Tom mentioned to me how Mexicans were "lazy" and had "fiery tempers." According to Tom that's why he had to be so harsh with this employee. In fact, Tom treated this employee like a dog.
And please understand that I cannot begin to accurately describe the level of abusiveness that Tom could unleash.
Obviously Tom's racist remarks made me very uncomfortable, especially considering my heritage. Even though I told Tom how I thought what he said was wrong, he expressed these sentiments on more than one occasion.
And I have made these same statements in a legal deposition solicited by a third party.
Yes, I was asked to provide a deposition regarding Tom's and his organization's behavior. A number of former American employees banded together to leverage legal action against Tom for breach of contracts. I have been told that Tom resoundingly lost the mediation and was ordered to pay up. But instead, I have been told that he has sufficiently pulled up stakes in the U.S. so that his money can't be reached.
Except by a few...
I have also been told that he was "forced" to leave his abode in the Caiman Islands, a getaway that Tom set up so that he could "avoid paying ridiculously high German taxes" as he told me once in Munich.
So much for Tom's patriotism.
It is very humorous to read Tom write that he had to edit my articles heavily. Tom writes English poorly. I still have all of my original work and, most often, Tom never edited anything at all -- which was sometimes embarrassing since I had to work essentially without a net.
And as far as "burying" any of my articles, I only know of one, which we later published here.
As to why I left Tom's Hardware, Intel could play Tom like a fiddle. Remember the Pentium III recall that Tom took credit for? Intel was ready to write Tom off when I, serving as liaison at IDF, rebuffed Intel's ridiculous excuse that the new Pentium III was "too fast" for the Linux compilers during the failed kernel compiles. Further, I told Intel that we were ready (at least I was) to arrange a broad, collaborated, massive coalition of sites to apply pressure to the chipmaker to recall the chip.
Tom might swing away from Intel for a little while, but inevitably he'd swing back and become an Intel apologist. That happened about when the P4 launched. I tried to help him with the review, but he didn't want to listen. Of course the original review was an embarrassment to the organization and he waffled several times in his infamous "recount" series.
He also started dipping into one of his more extreme paranoid states -- probably triggered by guilt and shame. He unloaded on me one day in a truly "crazy man" way. The issue occurred because I tried to mediate an argument between him and Chris Tom (I believe Chris received these emails as well, so he can stand witness, although I still have ALL of Tom's correspondence over the time I worked with him).
Chris and his site had created an Athlon optimized version of Flask as a counterweight to the P4 optimized Flask centerpieced in one of Tom's flip-floppy "Recount" articles. Tom accused Chris of "stealing MY article." While at THG, I tried to nurture connections with other sites -- my articles were usually replete with links to other hardware sites which is atypical for THG where Tom jealously guards his audience.
I diplomatically told Tom that maybe we could work with Chris since Chris's people had already developed the new Flask program and demonstrated their competency. At that point Tom went just a little crazier and started saying that Chris was his enemy (Tom has lots of people he hates and whom he labels as "enemies": Kyle Bennett, Alex Ross, Larry Barber and, of course, yours truly are but a few examples) and that I was siding with the enemy.
In light of the fact that objective reporting was often not possible working for Tom, in light of the fact that Tom never finalized my contract or any of his promises (which got him in the trouble with other Americans he employed as mentioned above), in light of the fact that Tom, though rarely abusive to me, wrote truly hateful things to my wife who also worked for Tom, in light of the fact that I had seen Tom lie time and time again...
I quit.
The Emperor Lucius Aurelius Commodus reincarnate, Tom was shocked.
He had just offered to move the American headquarters of Tom's Hardware to Springdale, Arkansas where I was residing at the time.
But I quit anyway because Tom was that bad of a person to deal with. From pimps to prostitutes, drug dealers and killers that I have had the misfortune to meet in my life, Tom is easily the worst person I have ever crossed paths with.
Tom is the tragic modern man. Possessed by his possessions. Lost in his ego. Mad from greed and narcissism. He is alone in his self-made prison of abusiveness and paranoia.
I have been thinking about writing a refresh of Dante's Inferno. I already have a ring in Hell reserved for Herr Thomas Pabst.
I still have all of his correspondence. I have the "temporary" contract that never was signed. I still have the bills I had to carry for THG that sometimes went well above $10,000 and which I had to wait months and months for reimbursements.
But there were at times humorous moments. Although there was certainly an element of terror, I'll never forget riding in Tom's S-Class Mercedes and watching him drive like a maniac, frightening pedestrians and running over every curb in sight.
Oh, and Tom in his silk pajamas smoking French cigarettes telling sex stories at 3 AM in the Grand Hyatt-Taipei... Was that a pass, Tom? You really creeped me out that night, which is why I ran out of your room.
Perhaps serving as a character defining moment in his career, Thomas Pabst was caught plagiarizing my articles recently, even going so far as to replace my name in the byline with his. After intense pressure and ridicule, Tom removed all of my 40+ articles from his site, several of which were some of the most important pieces THG has ever carried.
The many fans of Mike Magee might also find it interesting to know that Tom has actively worked to subvert the inquirer. And this goes well beyond his recent and ironic "Scooby attacks" where he accused Mike and his site of trying to "talk up nVidia's stock price."
===================================
The Important Issues
But notice how this entire discussion serves as a diversion from the really important issues at hand.
AMD produced a presentation highly damning of SysMark. This presentation was explosively newsworthy. The facts it contains are simple and compelling. For instance:
Eight of thirteen Photoshop filters favored the Athlon XP in SysMark 2001. All of the eight filters were removed in SysMark 2002 to be replaced with repeated filters that favored the Pentium 4. The deliberate exclusion of only these filters suggests specific intent by BAPCo/Intel to skew the workload towards the P4.
Greater than 90% of the Excel portion of SysMark 2002 is devoted to sorting, when sorting is an atypical task in Excel. However, the P4 is strong sorting data due to its bandwidth. At best, grossly skewing Excel towards an atypical task suggests sheer incompetence by the benchmark developers. But much more likely, given the other evidence listed here, given the Athlon XP's powerful FPU and unsurpassed performance in random memory accesses, it demonstrates the rubber-man contortions that BAPCo/Intel were willing to go to in order to bias SysMark towards the P4.
Microsoft Access is a wildly popular database program. Although it does not carry the multi-user robustness of Oracle, MySQL or Microsoft SQL Server, it is a very fast database useful in handling even large amounts of data. Access is one of the most clear-cut candidates for performance testing since queries and reports over many megabytes of data can take a lot of time. However, the Intel Pentium 4 is very, very bad in Access which was demonstrated in SysMark 2001 where it was completely dominated by the Athlon XP. So for 2002, the Microsoft Access component of SysMark has been almost completely removed. Now, according to AMD, Access contributes less than two seconds to the overall time of the suite. For comparison, Photoshop, a test that now favors the P4 as mentioned above, contributes ten-times more to the SysMark2002 final score.
The Flash component of SysMark 2001 was gutted for 2002 when over 88% of the test was removed. This 88% happened to favor the Athlon XP. Again, this demonstrates that a very specific type of scalpel was used for a very specific purpose in the development of SysMark.
And don't forget SysMark2001 which was introduced before AMD's SSE-enabled Athlon XP. The Pentium 4 came out on top of the Athlon XP in the Internet Content Creation portion of the test -- until AMD and Microsoft distributed a patch to enable recognition of SSE in the version of Windows Media Encoder (WME) distributed with SysMark 2001.
An unusual act in itself that has not been duplicated before or since, BAPCo provided a specific version of a program -- WME -- that had to be install before SysMark 2001. Inexplicably, this version of WME comprised a whopping 31% of the overall ICC score to SysMark 2001. (The percentages are taken directly from a BAPCo white paper).
After receiving a great deal of criticism for their choice regarding WME, BAPCo distributed the corrected version in 2002. However, the scoring emphasis for WME magically dropped from 31% to only 12% now that it became clear that the Athlon XP had robust SSE performance.
Again, the fact that AMD constructed and distributed this explosive presentation is highly newsworthy. It is also unlikely that the chipmaker would risk its reputation by providing detailed data that might be possible to disprove.
===================================
How-to Really Unlock SysMark
As I wrote above, Tom is not very "technical" and the instructions he gave to unlock SysMark, are predictably incorrect.
Specifically, the commandline instructions necessary to run SysMark 2002 contain the same typographic error sent to all reviewers (so much for Tom's editing).
Here are the two corrected commandlines:
For Internet Content Creation, from the "Run" dialog enter:
sysmgr -NOREBOOT -ICCSUITE=1 PROJNAME=icchacked
For Office Productivity, from the "Run" dialog enter:
sysmgr -NOREBOOT -OPSUITE=1 PROJNAME=ophacked
The actual project names are not important.
The key to "unlocking" SysMark 2001/2002 is to cause it to halt immediately prior to completion. SysMark logs the results from the individual tests in the temporary file "c:\Program Files\BAPCo\SysMark2002\Reports\_measurement.txt".
By adding the bogus command "die" to the end of the execution scripts "SEQUENCE.txt" (one for each suite) you force SysMark to generate an error and halt, exposing "_measurement.txt" for examination.
We have included one such file that we derived and renamed "icc_hacked.txt" here.
In all honesty, Thomas Pabst does not have the competency to judge the merits of these benchmarks.
And he does not have the integrity to be trusted even if he did.
Examine the facts for yourselves. The data AMD has derived is compelling.
vanshardware.com |
