I probably met Jim Navas on the Net before but I guess I forgot him. Anyway, he first contacted me directly, right after I posted the XStore Pro Comparison on the 27th of January, 1999. He e-mailed me a long comment about the XStore Pro which I answered. Soon, his comments were taking up a whole lot of time. Read the thread below if you don't believe me :-

Adrian:

Your review is simply ludicrous. You also obviously do not
1) understand how the HighPoint XStore Pro driver works; nor
2) do you have your "test system" set up properly.

Why you even have a "site" is beyond me. I am attaching my Winbench 99 results using the HighPoint XStore Pro IDE driver properly set up. As you can see, it makes your scores look ridiculous.

To paraphrase an old cliché, "If you don't know something, its better to keep your mouth shut than to open it and show everyone that you are ignorant".

Cheers;

Jim Navas
NCE Enterprises

Hello Jim,

I have prepared a table below of your scores and the scores I received from my system. Note that both my scores are with no limits to the size of the virtual cache.

It is pretty much evident that your hard disk is faster than the IBM Deskstar 14GXP I'm using. But the scores are not exactly conclusive evidence that XStore Pro works. I believe your Maxtor Plus 2500 hard disk has almost the same specifications as the 14GXP but has a faster access time of 9.0ms compared to the 14GXP's access time of 9.5ms. So, the Maxtor Plus 2500 could inherently be faster than 14GXP from a hardware point of view.

In addition, the amount of free disk space during the tests affects the results of the benchmark. An almost empty hard disk will receive a higher score in Disk WinBench than a half full or almost full hard disk. That's why I ran all the tests in one go. Thus, your results cannot be compared directly with the results from my system or any other system.

Finally, you are using a Super Socket 7 motherboard. Needless to say, there are differences between south bridges of the MVP3 and the 440BX chipsets. Again, that's another reason why your results cannot be compared with the results from my tests.

It would be really helpful if you could provide us with comparative benchmarks (averaged from 3 test runs) of the XStore Pro vs. Win98's bus mastering driver, rather than just a single benchmark run of XStore Pro. The results from Winstone 99 would also be very informative. In fact, Winstone 99 is very useful in determining the actual effect of XStore Pro on real world performance.

Regarding your remarks, since you feel I do not understand how XStore Pro works or how to setup my test system properly, I would really appreciate your frank thoughts on exactly which part of the review is wrong and how a proper test system should be set up. HighPoint doesn't seem to want to divulge how XStore Pro really works so if you know something we don't, it would be really nice if you could share it with us.

Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/

Adrian:

As I stated, you have *no* understanding of how the HighPoint IDE driver works. While I *won't* go into detail, the HDD IDE driver sets up a DRAM cache for "real world" files. It *doesn't* work with "synthetic" benchmarks, period, such as HD Tach 252. Further, SysSoft Sandra, Norton, WinTune, etc. are all junk, anyway, period.

The CDXpress sets up a HDD cache, not a DRAM cache, for the CD, BTW.

I have also done real world, proper, tests. Copying a single folder that contains MANY files (I set up a special version of Windows\System in a separate folder which contained 72+ MBs of data), the Winbench98 Business Mark score for the HighPoint IDE driver was essentially replicated in my real world test when I timed the copying of the whole folder from one partition to another. That verified that the HighPoint IDE driver DOES work as advertised. I timed the total transfer with a digital stopwatch that I am accurate with to within 4 hundredths of a second.

BTW, the Winbench98 scores are a significantly greater improvement over the Windows default driver than the Winbench99 scores. With Winbench98, due to the different files, etc., used for that test, the HighPoint IDE driver surpassed the default Windows "Standard DUAL IDE Controller" by 100% in the Winbench98 Business Disk Mark, for example. But even with Winbench99, there is at least a 13+% improvement in Business Disk Mark and a 19+% improvement in High End Disk Mark versus the default Windows IDE Controller in Win95B. The improvement over Win98's default, depending upon which chipset the MB uses, may be slightly greater or lesser. But both *are* significant improvements.

The HighPoint IDE driver is "made" by a subsidiary of Triones, so anyone who had a previous version of the Triones driver installed will get poor results due to the "conflicts" between the previously existing Triones and the new HighPoint IDE driver files. This is exactly the same as when someone installs Win98 with a VIA chipset (which loads the old VIA BM drivers included in Win98) and then tries to casually load a late version.

Also, for maximum HDD performance, Win9X should *never* be "messed" with, in terms of vcache or anything. Extensive real world testing has proven this, just as Microsoft says!

As the HighPoint IDE drivers sets up a *unique* cache in DRAM for the HDD, they do *not* work, or work well, will small amounts of DRAM. 64 MBs is essentially the minimum for proper performance, especially with Winbench99. 128 MBs will produce significantly better results. My tests were run with 64 MBs.

You also *obviously* do not understand either DMA nor UDMA. Neither one affects real world HDD DTR. DMA only "unloads" the CPU from being as involved with disk accesses. It does *not* meaningfully affect HDD DTR, period. UDMA is *pointless* as it is only good for the HDD "buffer", *not* for any transfers from the media. Even the FASTEST IDE HDDs do *not* approach the PIO mode 4 speed limit of 16.7 MB/sec for data that comes from the disk media. The current fastest IDE HDD transfers data at a maximum of 14.7 MB/sec from the media. Only from the buffer, which is limited to 512 KB for the best IDE HDDs and is 256 or even 128 KB for older HDDs, is the DTR affected. And if you understood *anything* about benchmarks *or* real world performance, even a 512 KB transfer at, say, 24.6 MB/sec (and THAT'S *if* the data required is IN the buffer!) with UDMA "Enabled" wouldn't make any real difference in the transfer rate of, say 15 MBs of data. Remember, the buffer is only "filled" from the media, so *only* the buffers initial content would result in ANY improvement! If you *understood* this, your replies to comments would obviously be different! But you obviously do not understand *anything* about HDD DTR, DMA nor UDMA.

As above, similarly, if there is not enough free space on the partition of the HDD that is used when using CDXpress, it won't work, either. It takes at least 200 MBs, minimum, to work properly.

Lastly, I checked your total "Comments" section. I did *not* see my "Comment" there, in any of the three parts of the "Comments". But I *DID* see a LOT of ignorant comments and your *equally* ignorant replies. Benchmarks are only worthwhile *when* the user understands the structure of the benchmark vs. what is being tested *and* when a *decent* benchmark is used. Also, *improper* installation of the HighPoint IDE driver, as *you* and *many* other commenters have done, totally negates their benefit, sometimes producing even worse results.

To paraphrase another cliche, "A little knowledge is a dangerous thing. Drink deeply from the wellspring of knowledge or do not drink at all". You obviously only take sips!

Your lack of understanding, and then having the audacity to post it, only makes you appear really ignorant to those that understand the subject.

You are ALSO doing *your readers*, as well as many companies, a REAL DISSERVICE with posting your ignorance.

I don't think that you will have the honesty to post this reply, just as you haven't with my prior comment.

I will also not deign to make any further comments. But anyone who reads and follows your reviews is being SIGNIFICANTLY mislead, period.

Jim Navas
NCE Enterprises

Jim,

I'll answer your e-mail in a little while but in the meantime, I think you went to the wrong page (or I posted it in the wrong page). There's a new XStore Pro comparison, if you have missed it. Your comment has been posted there as you have been previously informed.

The correct URL is here.

Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/

Jim, you are wrong. Win95/98's virtual cache sets up a DRAM cache for hard disk and CD-ROM data. The IDE driver just offloads the task of transferring data from the CPU to the motherboard chipset. And the cache will work with synthetic benchmarks. HD Tach 2.52's data transfer rate test is a hardware test that tests the hard disk's actual throughput from the platter to the IDE controller and therefore is not affected by XStore Pro. But other synthetic benchmarks that are affected by such a caching software.

The copy test is not a proper real world test. Users do not normally spend their time copying 72+MB of data from one directory to another. It is still a useful test but it cannot be used as irrefutable proof that XStore Pro works. Benchmarks aren't perfect and they can be fooled so scores from one benchmark must be backed by scores from several different benchmarks.

Taking a look at my scores, XStore Pro 1.20 with unlimited vcache scores 22.6% higher in Business Disk WinMark 99 and 2.5% higher in Highend Disk WinMark 99 than Win98's bus mastering driver. The improvement in Business Disk WinMark 99 is significant but the increase of 2.5% in Highend Disk WinMark 99 isn't because it's within the 3% error range stated by ZDBOp. In any case, you should check the 7th page of this comparison. From my tests, I discovered that with vcache left alone, XStore Pro used up an incredible 96.5MB of RAM for the virtual cache while Win98 uses only 60.4MB. So, the increase of 22.6% could well be attributed to the much larger virtual cache rather than XStore Pro's read ahead caching algorithim. In fact, I found that with the virtual cache's size fixed (even at 32MB), XStore Pro consistently performed poorer than Win98's bus mastering driver.

Curiously, I had the impression that HighPoint and Triones are one and the same. Anyway, that's irrelevant to the issue. Triones has been out of the bus mastering driver business for some time now. I seriously doubt that anyone would be using their old bus mastering driver anymore. Besides, XStore Pro 1.20 isn't a bus mastering driver per se. It makes use of the bus mastering driver that comes with Win98 or installed manually by the user.

You are quite right - for maximum hard disk performance, Win98 should be allowed to dynamically size the virtual cache. But the jury is still out on whether that's the right move for maximum system performance under Win98. I will be running some tests in the future to determine just that. BTW, I have not seen any extensive tests done on vcache size.

The virtual cache that XStore Pro "sets up" is hardly unique. It's Win98's vcache but XStore Pro adds a read-ahead caching algorithm and probably changed vcache a little. All caching software benefit from more RAM. That's natural because more RAM means a larger cache which gives better performance. All my tests were done with 128MB of RAM.

Oh.. yes, I do understand that DMA merely offloads the IDE transaction task from the CPU to the motherboard chipset. If you read the first page of the XStore Pro review, you would have noticed that. But the DMA checkbox in Win95/98's hard disk (and CD-ROM) Properties screen does more than enable DMA mode. Take a look at the two following graphs obtained by Chris Swan.

9.1GB Seagate Medalist Pro (DMA Enabled)

9.1GB Seagate Medalist Pro (DMA Disabled)

As you can see, leaving the DMA checkbox unchecked not only causes the CPU to retain the task of handling IDE transactions, it also effectively limited the total throughput of the hard disk from 14MB/s to ~7MB/s. I noticed the same thing when I ran both the Medalist Pro and the IBM 14GXP hard disks. That's why I keep asking those who wrote to me if they have enabled DMA for the hard disk or not. As you can see, not enabling DMA is detrimental to performance.

Jim, when I sent you that "your comment has been posted" e-mail, I have already posted the reply to your e-mail above. I think you have been looking at the Comments pages of the XStore Pro review. Anyway, I disagree that the comments from the readers were ignorant. Those were just their observations. Just because their results differ from yours doesn't mean they are ignorant. Also, there's no such thing as an improper installation. XStore Pro 1.20 is easy to install and there are no features that need to be set manually. I can't see how XStore Pro 1.20 can be improperly installed.

BTW, even Marco of Chaintech admits, "XStore Pro has received mixed results on different types of systems. We offer XStore Pro for free to our customers who may try the driver for themselves." ABIT is still looking into XStore Pro's performance issues.

Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/

Adrian:

I went back later and found it. I also found that you had "changed your tune" with regard to DMA and UDMA in your reply to me, which I think is very disingenuous of you. I firmly believe that this was after reading my second E-Mail and then posting my first E-mail with your "revised position" reply after learning at least something about DMA and UDMA from my second E-Mail (which was sent before you "changed" the review and added my E-Mail). You had previously made a BIG point about "DMA" affecting HDD performance in your replies to other people that you had posted before you got my first E-Mail.

You should know that people do catch you at this, just as I did. It does *nothing* for your "reputation", such as it may even be. I sent the second E-Mail in response to a private E-Mail that you sent to me. And as stated, just before I typed and sent it I check your site and my first E-Mail had *NOT* appeared. Nor was it there just AFTER I sent the second E-Mail, as I also checked just after I sent the second E-Mail.

When you "revised" your site, you "dropped" the prior comments as well as your previous incorrect replies.

I have also previously have sent you a third E-Mail in response to your "revised reply". This is the fourth.

You should also realize that you lack of knowledge and understanding of HDDs only becomes more evident with each "statement" that you make, even when you correctly revise part of your "posted reply" due to the knowledge imparted by others, in this case myself. The average seek time does *not* help the High End Disk Mark nearly as much as the Business Disk Mark. If you knew anything, you would understand this and thus not make such blatantly incorrect statements. Thus my much more superior High End Disk Mark, as opposed to my less superior Business Disk Mark, had *nothing* to do with any possibly faster average seek times, period.

I "paid out enough rope" and you "hung yourself".

You should pay someone to educate you about what you are reviewing before you go sticking your whole leg down your throat, not just putting your foot into your mouth.

As I started off, you clearly do *NOT* understand how the HighPoint IDE driver works, you do *NOT* have your system properly setup, and you also do *NOT* understand anything about HDDs, DTRs nor HDD benchmarks. Period.

I dare you to publish this E-Mail. But if you do, it *MUST* be published in its entirety, as also applies to the second and third E-Mails (except for the non relevant parts of the Winbench results only).

If you do *not* publish this E-Mail on your site, then I may take other action.

Jim Navas
NCE Enterprises

Jim,

I've not changed my stance on DMA and UltraDMA and I do not know why you still insist I did. About DMA affecting hard disk performance, please check the earlier reply.

Actually, that was not a private reply. The reply was posted on this page and you were looking in the wrong page. That reply was to point you to the correct page. If you have received my "your comment has been posted" e-mail, your comment has most definately been posted. I post all comments before sending those e-mails out. If you do not see your comments in the page, a simple click on the Refresh/Reload button may solve the issue. If you still can't find it, you can e-mail me and I'll direct you to it but please do check around before e-mailing me.

Frankly, I have no idea where you get that idea from. I did not drop any prior comments because there were none to drop in the first place. This is the first page of the Comments page for the XStore Pro Comparison and your e-mail is the second to be posted here. I posted all the first 5 comments at once. Before your first comment was posted, this was just an empty page.

If you did send me a third e-mail, then I did not receive it. This is the third e-mail that I have received from you. You can resend the third e-mail if you wish.

Jim, you have got to be joking. :-) I did not change any of the posted replies due to new info, if any. If there is new information, it will be posted in the Updates section of the article. I most definately did not use whatever you wrote to me. In fact, this article was written and posted (27/01/99) before I even received your first e-mail. It would be ludicrous to say that I have used your "knowledge" to write this Comparison.

About the relative performance of the IBM Deskstar 14GXP and the Maxtor Plus 2500, it's best that you perform both the tests for XStore Pro 1.20 and Win98's bus mastering driver and compare the results. You cannot just consider both hard disks to be performing equal just to suit your opinions.

Regardless of your opinions, my only aim is to provide comprehensive and fair reviews of various computer hardware and software with no nonsense tests. If it's working, I'll say it's working but it isn't, you can expect me to tell everyone about it. Thus, I still stand by my articles, irrespective of your opinions. Just as I dare to publish the truth in my articles, I don't mind publishing your e-mails. I have always published the e-mails in their entirety, only editing them occasionally for grammatical and spelling mistakes or to remove overly irrelevant text.

Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/

Adrian:

Since you "claim" to have *not* received my "third" E-Mail, which you did as I sent it with a Return Receipt required and I got the receipt, I am attaching my Winbench 99 scores for the HighPoint IDE driver (WB99_NCE336v2) and the standard Microsoft IDE "Controller" (Standard DUAL IDE Controller) (WB99_STD336.txt). I am also sending this one "Return Receipt" required as well. The systems tested upon were identical, period.

Your not posting these results previously doesn't say much for you.

Now your lack of honesty as well as your lack of knowledge is simply amazing. You did, in the original comments in the first review, make SEVERAL comments about DMA and UDMA speeding HD performance. This is just false, period. You DID "change your tune" between your "initial" "review" and your reply to me, period. Go back and compare the two sets of "Comments" and your replies. You just seem to think that the average reader won't make the connection.

You also have *no* idea about benchmarking. SisSoft Sandra is junk, period. If you had EVER looked at the results with a modern, fast system on a comparison basis, this would be blatantly obvious to even you, and for MUCH more than their HD tests. Similarly, Norton and WinTune (any version) are also junk. OTOH, HD Tach 252 is a decent *synthetic* benchmark, but doesn't apply to real world HD performance, especially with something like the HighPoint IDE driver. The results that I get with and without the HighPoint IDE driver using the HD Tach 252 benchmark are essentially the same due to the fact that HD Tach 252 is a synthetic benchmark, period. If you understood how the HighPoint IDE driver works versus how HD Tach tests things, this would be obvious to *even* YOU.

Also, the HighPoint IDE driver doesn't have anything at all to do with "vcache", period. OTOH, letting vcache be set by Windows *does* improve HD performance. *I* have done exhaustive tests to confirm this, which you *obviously* haven't! There is a "common" "old wive's tale" about setting the parameters for vcache improving performance which is absolutely incorrect.

You should also understand that people using the HighPoint driver with a non Slot 1 MB and using more than 64 MBs of DRAM with an Intel chipset *will* experience a decrease in performance. This is simply due to the fact that almost all Intel chipsets do *not* cache the DRAM in the L2 cache when the DRAM exceeds 64 MBs (the only exception is with certain MBs that have the HX chipset and enough TAG RAM to support the additional DRAM). Again, your ignorance knows no bounds.

Now, the HighPoint IDE driver *does* use more DRAM. But, again, this has *nothing* to do with vcache, period. The HighPoint IDE driver does set up a separate cache, totally different than the vcache, for its performance improvement. In fact, Win9X works better (faster) if you set the "Read Ahead" for vcache down to 4KB (Device Manager - I'll let you find the place to change it). Not a LOT faster, but faster. Again, I have done exhaustive testing of this *fact*.

OTOH, the "average access time" with the Winbench HE Disk Mark does *not* make as much difference, *if* you understood what average access time really measures and thus does on a non fragmented HDD (which EVERY HDD tested *should* be Defragged *before* each and every test).

You also obviously do *not* understand about HD Tach 252 and the amount of data on a HDD. HD Tach, a *synthetic* benchmark, *does* show the decline in HD DTR the farther "in" you go with the HDD. Thus, with more data on the HDD, the slower the HDD DTR will be for a given "real world" test, such as Winbench 99. Your scores should therefore have been MUCH faster than mine, period.

With the HighPoint IDE driver and the *unique* cache that it establishes, the more the DRAM the better the speed improvement (given that you are using fast DRAM properly set up, of course, which *you* may well not be doing). Thus your 128 MB of DRAM should *significantly* outperform my 64 MBs in my test system, especially with the HighPoint IDE driver, period. This is due to the cache size enabled, which the HighPoint IDE driver *does* vary with the amount of DRAM available.

I am *not* going to go into how the HighPoint IDE driver works in the real world, but your "knowledge" is abysmal. But it *does* work, *IF* the system is set up properly. Your system is obviously not set up properly, period.

You, and the majority of your readers, are very ignorant of computers. The lack of understanding about benchmarks, how they work, how reliable various ones are, the lack of understanding of a proper set up, and the lack of knowledge about how the HighPoint IDE driver works is simply astounding, given your claims about your test system and your results. If you don't think that the HighPoint IDE driver works (as I *know* that it does), then why do you think that so many companies have licensed it? You don't think that these companies tested the HighPoint IDE driver before they licensed it? That's just DUMB!

Another thing, you clearly also do *not* understand business, either. When a company licensing something, they control when and what versions are made available to the public. HighPoint IDE drivers are *not* "interchangeable" between different MB mfgs, nor even between all MBs made by the same company. This is due to "BIOS" check that the HighPoint IDE driver does to make sure that the version (MB mfg's version) that is being used matches the *specific* MB that it is licensed to be used upon. But that's probably beyond your dense skull to understand. Otherwise, anyone could download a HighPoint IDE driver from any MB site and use it with their similar chipset MB. Thus HighPoint would not only lose money, but they would also have to charge a LOT more for their licenses!

Yet another display or your ignorance, as well as that probably of your readers!

I could go on and on, but the ignorance displayed by you and your readers doesn't make any sense to me to do so. There are MANY more factual errors in your "statements", but they are all equally "baloney".

If you do not "fess up" and admit ALL your errors, then I will take other action.

Cheers;

Jim Navas
NCE Enterprises

Jim,

No. I do not claim to have not received your third e-mail. I know I have not received your third e-mail. In any case, the results (minus unnecessary details) are posted below your e-mail in a table.

Yes, I did mention that checking the Enable DMA box in the hard disk's properties page enhances hard disk performance. I have already detailed the reasons in my previous answer. I can't help it if you don't believe me. The two graphs aren't even from me though I have personally noticed the same phenomenon with my hard disks. I have read all my replies (I wrote them after all) but I have not seen a discrepancy. Since you seem to have found some form of discrepancy, why don't you tell me which comment it is? It's not that hard. Just the comment numbers will do.

Who mentioned anything about SiSoft Sandra? Or Norton System Benchmark or WinTune? All my benchmarks were done using WinBench 99 and Winstone 99. Norton System Benchmark is a benchmark of overall system performance. While it's not very accurate, it is affected by any software that enhances a system's performance or otherwise. So, it can still detect any change in system performance due to XStore Pro. You still have a misconception of what a synthetic benchmark is. A synthetic hard disk benchmark uses programmed sets of data to stimulate average hard disk usage in order to test the performance of a hard disk. Thus, Disk WinBench 99 is a synthetic hard disk benchmark. HD Tach isn't a synthetic benchmark per se because it tests the actual throughput of the hard disk and is thus, unaffected by any software caching utility.

Here's what HighPoint Technologies has to say about XStore Pro and vcache - "XStore Pro will modify the activities/operation of vcache". I guess that pretty much rebuts your statement. As I have mentioned in the previous reply, yes, letting Win98 dynamically set the vcache size improves hard disk performance but that doesn't necessarily mean it will improve overall system performance. I will be running some tests later on to check that out.

Actually, I have not tested the XStore Pro on the Intel 430TX or older chipset motherboard for obvious reasons. Thus, I did not mention anything about XStore Pro performing poorly in a i430TX motherboard. I have no idea why you are hammering on this. FYI, the Intel 430TX will still cache the first 64MB of DRAM if you exceed 64MB. Also, from the 440LX onwards, the cacheability of the RAM depends on the CPU, not the chipset. The Celeron and the new Pentium II (stepping dA1) can cache up to 4GB of RAM while the older Pentium IIs can only cache 512MB of RAM. Still, that's more than enough, right?

Please read HighPoint Technologies' comment above on the relationship between XStore Pro and the vcache. XStore Pro most definately does not set up a separate cache. Rather, it adds its own read ahead algorithim. Again, I have this in writing from HighPoint Technologies themselves. I can't really say I agree with your reduction of vcache's read ahead optimization but it's your system, not mine.

Again, you have a misconception of HD Tach 2.52. HD Tach doesn't care how much data you have. Your hard disk can be full or empty for all it cares. It just reads from the outer tracks to the inner ones, just like the disk transfer rate test in Disk WinBench 99. Disk WinBench 99 isn't a real world test. Winstone 99 is a real world benchmark. That's the reason why I used Winstone 99 to evaluate the actual performance difference with/without XStore Pro. In any case, you are conveniently assuming that my hard disk is empty while yours is filled to the brim.

Again, there's no unique cache associated with XStore Pro. And SDRAM is synchronized to the memory bus speed. Whether you are using 6ns or 7ns SDRAM doesn't change a thing - both of them are still running at i.e. 100MHz. You can't compare a K6 system with a Celeron system. It's like comparing apples with oranges.

Speculatively, HighPoint's XStore Pro could be just good conware. It's faster than Win98's default bus mastering driver because bus mastering is disabled by default. So, HighPoint and its clients can post stupendous results. That's good for HighPoint. For the manufacturers, if one of them bundles XStore Pro with its motherboards (for hype value), others will follow suit as well, irrespective of whether it works or not. After all, would a buyer choose a normal no-frills motherboard or one with a new hard disk accelerator that speeds hard disk performance? This is just one possibility that I can think of right now but any budding conspiracy theorist can probably come up with something better.

I did not mention anything about the different versions of XStore Pro. As far as I know, they are all the same. Only the BIOS checks are different. But, all motherboards under a supported manufacturer can use the same version of XStore Pro 1.20. I don't believe I said otherwise.

Let me save you the trouble. Head over to Anandtech and SysOpt.com. There, they have a BBS and a Forum. Please feel free to post any flames (which is what these e-mails are) over there. From all the e-mails you sent me, you have said a lot of things, most of which have been proven inaccurate. In addition, while you claim the other visitors are "ignorant" and "do not know how to benchmark, set up a proper system, etc..", you have also conveniently refuse to tell us your "proper" methods of doing such things. I have been very transparent, posting my articles and research including everyone's comments on my site. Perhaps it's time you did so yourself.

Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/

Adrian:

Your "reference" was, once again, incorrect. Don't you even know your *own* "site"? It referred to the comments for your "first" review, *not* the second review.

Also, I have sent you, Return Receipt required, a final E-Mail about the HighPoint IDE driver. Your ignorance is simply astounding as well as the ignorance of your "readers", such as they may be. And your disingenuous replies are simply absurd, period.

You should "get out of the business" before you REALLY EMBARRASS yourself!

Jim Navas
NCE Enterprises

Jim,

The first link pointed you to the Comments Index while the second pointed you to the first Comments page of the XStore Pro comparison. Due to the nature of your previous e-mail, I made sure that the link was correct before I e-mailed you. Therefore, the link could not have sent you to the Comments page of the XStore Pro review.

I'm surprised that the thought hasn't occurred to you yet.

Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/

Why even bother with psycho Jim Navas...

Thanx for the good site and info

jaqtdj

I think you are right. I have wasted too much time entertaining him.

Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/

Adrian:

As the subject says, you are simply ignorant. And if you want to "keep company" with "Jaqtdj", that just shows how ignorant you are. All "jaqtdj" can do is get himself into a LOT of TROUBLE by violating Copyrights, as he will find out.

The two "test" examples that I sent to you were run on *IDENTICAL* systems, everything "ceterus paribus", PERIOD.

I have ALSO tested the "read ahead" cache size with identical MS as well as HighPoint IDE drivers, which YOU haven't!

The HighPoint IDE driver *DID* outperform the MS IDE "Controller", PERIOD.

The HighPoint IDE driver does *NOT* just "play with" "vcache". You are simply stupid.

You ALSO have *NO* idea what a "synthetic" benchmark is. HD Tach 252 *IS* a "synthetic" benchmark in that it *DOESN'T* use *REAL* files. The jerk who sent you the "graphs" from HD Tach is also obviously "screwed up", just as you are. There is *NO* "real" HD Tach 252 that would produce such graphs, period. Why don't you read the HD Tach "comments"?!?

There is a REAL difference between HD Tach 252's "synthetic" benchmarks and a benchmark, such as Winbench98 or 99, that uses *real world* files and file names. But that is something that YOU just "don't get", obviously!

As to SDRAM, it depends on *how* the system is set up. If you *knew* anything, you'd know that Win9X always loads into the TOP part of memory. Thus with a TX MB, if you have more than 64 MBs of DRAM, your system *WILL SLOW DOWN*, PERIOD. Why don't you ask Intel rather than "mouthing off"?!? The DRAM above 64 MBs is *NOT* cached with a TX MB, PERIOD! And, thus, when Win9X loads into DRAM above 64 MBs, it *DOESN'T* make the system as fast, in fact it *SLOWS THE SYSTEM DOWN*! Get real.

Your ignorance is simply astounding. I don't believe that you can "figure your way out of a paper bag", so to say (at the LEAST).

Your ignorance is simply too much to waste more time on.

Good bye!

Jim Navas
NCE Enterprises

Err... Jim,

Copyright? What has jaqtdj's statement got to do with violating copyright?

So were my tests, Jim. All done on the same system "sekali gus".

Then perhaps you would like to share with the community your discovery? I don't mind allocating some space for your article.

Yes, it did but a further test showed that with XStore Pro installed, the vcache size was also 60% larger than normal. Thus, the performance increase in hard disk speed could well be due to the larger vcache rather than XStore Pro's read ahead algorithm.

Well, if you are right, it's the guys at HighPoint who are stupid because I have been quoting them directly on the vcache issue. As I mentioned earlier, I have their statement in writing.

LOL! If you don't like HD Tach 2.52, try Disk WinBench 99's Disk Transfer Rate test. It will show you the same graph, only more thorough. If you want a "real" HD Tach 2.52 (whatever that means) that will produce such a graph, jump over to the Binaries Storage Area. There's a copy of HD Tach 2.52 there.

Actually, Win95/98 uses memory from top down but the 430TX chipset caches only the first 64MB of RAM. That means that in a 430TX system with i.e. 96MB of RAM, the first 32MB used by Win95/98 isn't cached but the rest are. Whether the system will slow down or not depends on a lot of factors. During memory intensive use, the 430TX system with 96MB of RAM will be faster than the 430TX system with 64MB of RAM. You shouldn't depend so much on WinBench. Try Winstone once in a while.

I didn't figure my way out of any bag. I just let the facts speak for themselves. Feel free to rebut but try to do it with facts, not claims. As I mentioned earlier, take this issue to the forums and BBSes out there. I don't mind. In fact, I challenge you to do that. Since you are so confident of your facts and knowledge, present your points on those forums. If you are right, the numerous experts online will support you. If you are wrong, well, we are all ignorant, right?

Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/

Adrian:

As a final comment, attached is my Winstone 99 score using the HighPoint IDE driver. As you can
see, it outdistances your CeleronA at the same MHz. And you have both more SDRAM (128 MBs vs the 64 MBs used in my test) as well as a faster video card (TNT vs my i740).

If this doesn't "show you the light" about both your poor system setup as well as your patently ridiculous comments about the HighPoint IDE driver, then nothing will.

And I *DO* think that "nothing will".

Jim Navas
NCE Enterprises

Hello again, Jim,

Again, you are comparing apples with oranges. Check out this table of the differences between our test systems :-

I think you can see that everything in both our test systems is different. Granted, my CPU and graphics card is faster than yours. But your K6 has twice as much L1 cache and 8 times as much L2 cache as the Celeron I use. This helps a lot in office applications. Also, you need to remember that the architecture of the K6 and the Celeron is different. You'll note that even the Cyrix MII can outpace the Pentium II in Winstone at the same clockspeed.

In addition, you ran your tests at a low resolution of 800x600, which needs 39% less processing and bandwidth than the 1024x768 resolution I used in all my tests. I also noted that you have partitioned your hard disk into 3 different partitions with 3GB in the partition that you use for testing. That 3GB partition has fewer FAT entries than my 14.4GB single partition. Thus, that introduces another possible advantage in your case. Fewer FAT entries mean lower FAT overhead.

Lastly but not least, your motherboard's chipset is VIA's MVP3. Microsoft's bus mastering driver for the MVP3 in Win98 is most probably not as optimized as the Win98 bus mastering driver for Intel chipsets because the Intel south bridge has been around for a long, long time (since the 430TX) while the MVP3 south bridge came out much later with the introduction of the VIA VP3 chipset. Needless to say, Intel is a much more important partner to Microsoft than VIA. That's probably why VIA has released so many bus mastering driver while Intel has done nothing of that sort.

Jim, I frankly think you should post your results on the forums and BBSes. Then you will be able to compare your scores with other VIA MVP3 users. Perhaps XStore Pro helps improve the MVP3's performance but as far as I can tell, it won't do that for Intel 440BX motherboards. It's pointless to continue this apple vs. orange argument.

Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/

Adrian:

Your comments are both absurd and extremely ignorant. The following points pertain:

1) You "fessed up" to the fact that your CeleronA CPU was running at 450 MHz vs mine at 336 MHz. That's a 33.9% speed increase for your CPU vs mine. Where is your 33.9% speed increase in Winstone99? And, remember, that the CPU used in MY test was *NOT* an "XT" core AMD CPU, which would be .8 Business Marks faster at the same MHz!

So why does my system get a BETTER Winstone99 Business Mark that yours? Try and explain that! Instead, your Winstone99 score is LESS than mine despite a 33.9% faster CPU!

2) The CeleronA runs its L2 cache at CPU MHz (i.e. in your case 450 MHz) vs mine at 112 MHz. The speed of your L2 cache *MORE* than makes up for the larger cache size of a Super 7 MB, period.

3) Your 128 MBs of SDRAM is MUCH faster than my 64 MBs as used in my test system. If you don't understand this then you are even more ignorant. Further, the Intel "TX" chipset, which does limit the cacheable DRAM to 64 MBs, *does* load Win9X into the highest memory. This is due to how
Microsoft set up Win9X to operate, not due to anything to do with the hardware itself. Thus any Intel socket 7 chipset, with the exception of a few HX chipset MBs, will *NOT* run as fast if they have more than 64 MBs of DRAM installed, period. Again, your ignorance is astounding for someone who promotes themselves as "knowledgeable".

4) As for 1024 x 768 resolution, I get a Winstone99 Business Mark of 18.3 when using that resolution. Big deal. Your faster video card more than makes up for the .1 less Business Marks under Winstone99!

5) Your comment about my partitions on my HDD again just show your ignorance. The number of folders and files determine the size of the FAT "directory", not the physical size of the partition itself. As *ALL* of my programs and data files are loaded into the "C" partition, your comment is both irrelevant and ignorant.

6) The default driver used by Win95 OSR 2.1 is *not* a VIA driver as is used by Win98. Again,
ignorance on your part. I clearly stated that the driver that was used in the "WB99_STD336.txt" file
was the Microsoft "Standard DUAL IDE Controller". Thus it *IS* a Microsoft IDE driver, *not*
anything from VIA! Again, your ignorance is astounding, especially since I specifically stated
this.

As a result of ALL of the above, you should "hang your head", at a minimum. You do *NOT* know
nor understand the HighPoint IDE driver, period. Nor do you know how to set up a computer system
properly. This is apparent from your abysmal scores.

I now understand that your ego and ignorance will prevent you from facing the truth. I just hope that not too many people are mislead by your "reviews"!

Cheers;

Jim Navas
NCE Enterprises

Jim,

Perhaps you might want to take just a second to check out the 2nd page of this comparison. It states very clearly what processor I'm using for the tests. From your previous letter, I can only deduce that you did not read the article or you just like arguing over nothing.

Now, as I have mentioned earlier, the architecture of the K6 is different from that of the Celeron. Look at the Cyrix MII. It beats the Pentium II flat in Winstone 99 at the same clockspeed. Now, does that mean that Pentium II users are idiots? No! AMD's and Cyrix's CPUs have always been strong in business applications, especially since they all have 64KB of L1 cache and recently, 1 to 2MB of L2 cache. But most people buy Celerons and Pentium IIs for a different reason - strong FPU performance. So, you can't compare your K6 with the Celeron. Is that a good enough explanation for you?

How do you know the 1MB of L2 cache doesn't offset the faster speed of the Celeron's 128KB L2 cache? Have you done any tests? If size doesn't matter (which is what you are saying), pray tell why is everyone so interested in increasing the size of the L1 and L2 caches?

Actually, I have no idea why you suddenly brought up the 430TX's cacheability issue a few e-mails ago. It had nothing to do with anything being discussed at all. But let me repeat it for the third time. The Intel 430TX chipset caches the first 64MB of RAM but Win95/98 accesses memory from top down. So, if the system has 96MB of RAM, the 430TX will cache the first (bottom) 64MB but Win95/98 accesses the RAM from top first. Naturally, in this manner, benchmarks will show a drop in performance. However, real world performance will improve regardless of benchmark scores if you use memory intensive applications. I hope this will be the last time I need to repeat myself.

Hey, Jim, there's a newsflash! My pal, Adrian Tung, has just pointed me to a page you should find extremely interesting. Don't take my word for it. Just check it out.

P.S. Sorry, everyone there don't want you there either. So, I removed the link.

Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/

Adrian:

Despite my getting a "Return Receipt" with my VERY accurate rebuttal to your comments regarding my and your Winstone99 scores, my E-Mail has *NOT* been posted.

Your ignorance, disingenuous "replies" and plain deceit just go to show who you really are.

Cheers;

Jim Navas
NCE Enterprises

Jim,

Now that I know who you really are, let me lay it to you straight. You have been extremely disruptive to this site. I have tried to be as helpful as I could to you. But you have countered that with nothing but hostility and well, a load of nonsense. I believe I have rebutted all your points and you keep coming up with more of them. The buck stops here.

Go and bother someone else or I'll be sure to let your brother, John Navas, know that you have been up to your usual pranks again. You have wasted enough of my time and now you, Jim, have the dubious honour of being the first in this site's history to be considered persona non grata and banned for life.

Congratulations, good day and good bye.

Adrian Wong
Adrian's Rojak Pot
http://www.rojakpot.com/
http://www.adriansrojakpot.com/