Wild speculation on Opteron

I am wildly speculating here but could it be that the 4 in the middle refers to the DDR speed?

5 for 533?
What numbers for DDR II though?

Regards,
Andreas

We kicked an argument off today with an article all about drivers...

You are not to be flamed, because you are head on!
Why did Creative stop the Live!Ware program ?
Because people did believe that they would get a completely new product
just with drivers.
Maybe we where partly to blame, but what we sad people did read a lot more
in to than we where expecting, resulting in real flaming of us :-( and we all know how
sad that experience is.

So now its just called:
Sound Blaster Live! - LiveDrvUni-Pack English 10/03/2003

anonymous
[From a Creative email address]

Well, Samuel Clemens:

Good write-up, but I think you TOTALLY missed the 'root' issue.

The issue being that VIA chipsets operate so horribly, compared to the others 'out there', that owners are FORCED into the driver upgrade game in hopes that their problems will magically go away.

Just simply looking at the cpu utalization during data transfers to and from a hardrive (s) will tell anybody, with their wits about them, that all is not well in VIA land. I see cpu utalization of between 35-45%, that I have since cured with the addition of an Adaptec 2400 RAID controller. (utalization is currently 2-3% here now)

You mention drivers and the 'magic' that is played on hardware owners by several companies. Might I submit, that you clearly don't remember too well the 'magic' that VIA played on their customers in the past with, if memory serves, 2-3 4in1 driver releases a week. Seems to me that, rather than ATI, or Nvidia, cited in your article, that VIA was the originator of this driver BS, along with 'the weekly bios up-grade'.

I did read the mentioned article and all that was re-enforced was the high resource cost of hardrive read-writes. As you mentioned, the differences between the two drivers, fell within the margin of error. The question I have is; why release a new driver at all? Unless things are a bit slow in VIA land?

I have noticed that VIA has become a major advertiser at your website and trust that the 'honest' reporting will continue.

Be good folks, you are my first daily stop.
George

Via is a great chip Company they get overlooked for some unknown reason they have made more Chipsets than Carter has liver pills I don't know of a chip Company that has outdone them for a long period of time they just keep pumping out really great Chipsets but do it with little fanfare .

Freddy

An article last week on the Nforce 2 having some problems is still prompting possible work arounds...

I don't want to flame anyone but I pretty sure abit builds a safeguard into their boards or they did that allows you to boot at a safe or default setting or at least they used to because I only learned about it when I had a bh6 but I am pretty sure it was on a kt7 I had and my current leadtek nforce2 board does this also the manual doesnt always point it out but this what I do

1. shut the power off to the board with the switch on the powersupply if there it has no rocker switch unplug it.
2. turn the power back on after the led's go off or wait few seconds then turn it back on or plug it in.
3. Push and hold the insert key down and keep it down thru the next step.
4. While you are holding the insert key push the power button on the front of the case.

Doing this boots the pc with the cpu, memory, fsb, and voltage settings at a safe setting (in my k7ncr18dpro the fsb is lowered 100mhz even with the internal jumper set at 266/333) If you make no changes to the setting and reboot it will still have the settings which did not work. I was getting the impression that most boards have this feature or it is something that award bios has because it has been available when I have tried it on the boards I mentioned. Just wondering if you could check it out if it is award bios that is responsible then some boards that may not mention the feature could have it and not document it.

Email address supplied

You should look at the forums.sudhian.com, the nforce 2 SN41G2 boxes
will run only reliably with Corsair PC3200 if you use the integrated video. Otherwise, don't count on reliable operation with 2700 memory at 333mhz operation when using the integrated graphics. I couldn't get it to stay free of BSOD without having 2 dimms installed; with one PC2700 dimm and it running at 266mhz it would BSOD or linux would never install or boot. Simply adding a video card solved every crash. nVidia's website has the memory validation list and the usual suspects don't pass for 333
or even 266mhz operation with the IGP. Have a SPP or technically not use the integrated graphics and you are OK.

Bottom line: Run a lower end system with a 266FSB and 2 dimms of PC2700 memory in sync or run it at 333+ and a video card.

Email address supplied

Is Intel hiding functionality from end purchasers? And how does it do it?...

I used to work at Intel. A few years ago, I saw the instructions for making a Pentium into a Celeron. The main thing was to set certain "fuse bits". These fuse bits have a complicated method to be set, and may include applying a special voltage to certain pins. I suspect the bits can be reversed/unset, but I don't know that for sure.

I'm sure Intel is using the same process for HT.

Anonymous

In regards to Hyper-Threading existing on older Pentium 4s, I seem to
remember reading many months ago (over on www.anandtech.com I think) that
whilst Hyper-Threading did exist on older Pentium 4s, it wasn't as
"polished" as the new properly enabled Hyper-Threading. As I recall,
enabling Hyper-Threading in one of these older P4s could actually decrease
general desktop performance by about 10%, which is why it the chips weren't
sold "Hyper-Threading Enabled".

Or something like that, anyway :)
-Colin

Our Cebit coverage brought a poke in the ribs for our esteemed editor...

Personally I think Mageek goes to these IT shows just to hit on the babes. Evidently his "charm" must be failing him as I haven't noticed any of the "eye candy" latched on to his arm. :>))

And you can tell him I said so! [I know who you are, Ed.]

Email address supplied

Gigabit Ethernet going double speed brought in this mail chop-chop...

I'm glad Intel is doing something about poor GigE performance on their desktop platforms. The current 930Mbps figure used by Intel in your previous article I'd like to see the test data and configuration info on. Intel's not disclosed the PCI bus latency issues on their P4 chipsets. Creating CSA and claiming its better is the only confirmation that there was a problem to solve.

CSA is a brand new class of Band-Aid for PCI limitations. The first was AGP, then AGP2X, AGP4X, and AGP8X to improve graphics performance. PCI-X is a great improvement over low-end 32-bit PCI, but its yet to migrate down to the desktop. Bandwidth improves, but there are still latency issues with Intel's implementation. Again, mum from Intel on this.

Marchitecture hypes the peak bandwidths of FSB, PCI, and now CSA, but due to high latencies and small buffers, actual through put falls far short. The 64KB buffer space on Intel's NIC chips is not enough to bridge gaps in PCI data delivery. Its worse now than on the P-III. The P4 has double the cach-line width, so the PCI bus has to wait longer for that to complete before it can get at memory. Second, the P4 makes speculative memory references so the long, 20-stage CPU pipelines run dry less often. This competes with the PCI bus. The speculation is bad for hyperthreading and SMP getting enough needed memory access too, by the way. That's just fallout from the speculation being tuned for single-cpu, desktop platforms.

Intel claiming that CSA has priority over other memory access is admission of PCI latency issues and today's chipsets' bias towards CPU access. If they didn't do that, benchmarks might not look as good. So what, that real system performance (computes + I/O) would benefit - Quake is why people buy computers, right?

Email address supplied

Minimising the space that Windows takes up has its fans...

Although I never got my old system to 40 meg, I did manage to get the 96 meg 400 MHz duron system to boot 99% free system recourses. And that is after running the app to get the info. The system would boot in a couple of seconds. Removing all the crap at startup etc. also makes the system run very stable.

The guy makes a good point, all this hype about how Linux is smaller, faster and more stable may be somewhat true on the server front, but on the desktop that is horribly inaccurate.

You can run win95 code on XP without any trouble. You can also run 32 bit code on windows 3.1 with libraries installed. Linux breaks compatibility with just about every version release.

Kapps crash often on my system, my Red Hat install uses over 250 megs of RAM just to boot. My ram usage climbs over 400 megs after using it for a few hours of use. My XP install has been running all day with fairly heavy usage, I also have a number of apps open and It's using under 100 megs of ram.

It also takes for ever for most of my GUI Linux apps to load.

The Linux kernel is getting huge, growing about 25-30% a year currently. The kernel itself taking over 100 megs of disk space, and that is a monolithic system!

So Linux is slower, less stable, larger, and has less features that windows. It's also much less productive, and seems to be growing much faster than it's improving. Not to mention that GPL and difficulty of installing/distributing software scares away closed source commercial software vendors. The only real way to distribute software for Linux is to GPL it and let the distributors build and test it for you. If MS ever decided to sell XP Pro lite edition, they would end any hopes Linux has at desktop market share.

Email address supplied

The failure rate of PCs comes as no surprise...

I used to build PC's and them breaking a lot early on is no surprise to me. It's one of the reasons I tried to burn them in for at least 72 hours before pronouncing them fit for delivery to the customer, and I can say a lot of them broke during the burn in process, and the majority within the first hour.

The fact is if you where plot out the failure rate of PC's by time of use you would end up with an inverted bell curve. In other words they break a lot early one and then for a long time very little and at the end of there life spans a lot again. Because of that I recommend any who buys a new PC to leave it on 7X24 the first week, especially if they are light computer user. This way if it breaks it does so during the time manufactures are most flexible and responsive to their customers.

Email address supplied µ