Intel Hyperthreading discussion.

[HellDiver]

My word people talk a lot of horseshit about VIA drivers. Updates every week. Huh? Constantly changing. What was that again? Two years to get it right. Speak louder, you're not making sense.

How quickly we tend to forget... Remember VIA GART, Tony? Remember 4.17->4.20->4.24 story? Remember hybernation and 4.25? Remember 686B and 4.31? And that's just the big issues that received full confirmation from VIA and were widely published! But hey, Tony just sells systems, he doesn't care whether for example the 4-in-1 he installed "before you do the sound and video" drags FPS on customer's system down by 25-30% in games! And you know what? 25-30% lower performance is equal to getting a CPU a couple of grades slower! And that's a lot of money wasted, Tony, don't you think?

Or how about the PCI? VIA's Apollo Pro line of chipsets remained pretty much the same PCI controller -wise since the days of Apollo 100/133 and all the way till today : yeah, the FSB was changed once in a while, the memory interface clock was bumped here or there, the S/B was finally blessed to sit on its own bus - but the PCI controller... For God's sake, a 3rd party, independent programmer had to write a patch to improve the performance of a chipset! How embarassing must that be! How many PCI RAID controllers did you sell with VIA systems, Tony? Do I hear "zero"? Good for you, but you're not the only shop owner out there!

You say you never had any problems with VIA due to your high "VIA handling skills"? Fine. I'm not making any sense? Fine. Everyone's entitled to an opinion, and that happens to include me!

 

[HellDiver]

"I were" can be perfectly good English. For example: "If I were an elephant, I'd have big feet."

Can be. In statements using conditional future such as "If I were a..." the subjunctive "were" is even a requirement (if we're talking about proper English).

But I happen no use "I were" in completely wrong instances, including the ones Coug was talking about. And so do a lot of those folks whom Google search brought up. We are illiterate...

 

[time]

Remember VIA GART, Tony? Remember 4.17->4.20->4.24 story? Remember hybernation and 4.25? Remember 686B and 4.31? And that's just the big issues that received full confirmation from VIA and were widely published! <snip>... the 4-in-1 he installed "before you do the sound and video" drags FPS on customer's system down by 25-30% in games!

AGP is an Intel invention. Via was the first competitor to implement AGP, so I don't think anyone was surprised when they struggled to reverse engineer GART and other AGP issues. But that's ancient history, and it worked fine for countless thousands of MVP3 owners.

From memory, 4.17 was the most problematic release. The first public 'fixed' release was 4.20, which wasn't enough so they pushed on and released 4.24. These are their internal builds, not service pack numbers. I think I found 4.23 quite acceptable, but it was withdrawn in favour of 4.24.

I don't remember the hybernation issue because there are so many damn hibernation issues that no-one uses it. Hibernation is for laptops, despite what the EPA might claim. This so called problem wasn't even a blip on the radar for most people.

The "686B" issue did not affect 99.99% of users. Maybe more. The fuss was because of the possibility of data corruption. Many of us happily copies hundreds of megabytes without a glitch. Don't misunderstand me, it was serious, but nowhere near as bad as i820 with MTH.

I'd like to know which 4-in-1 cost you 25-30% in games. I've checked most releases for exactly this sort of improvement/degradation, but never seen it. You wouldn't be exaggerating by any chance, would you?

 

[time]

In statements using conditional future such as "If I were a..." the subjunctive "were" is even a requirement (if we're talking about proper English).

Actually it's 'present unreal conditional', nothing future about it.

We are illiterate...

Speak for yourself! We was learnt real well here in Orstralya. What happened to you?

 

[HellDiver]

Hi there, time! Nice to know your name's Tony too, time! HellTroll, you say? Then you would be feeding me, no? And doing so in an extremely poor way at that - you're really not up to speed on what you say. Next thing to clueless, in fact. How do I know? Because I've had to deal with those issues myself on many machines.

AGP is an Intel invention.

PCI, GTL+, EV6, USB, IDE and every other bus VIA and all other chipset manufacturers use or have to I/F in their chipsets (spare for N/B<->S/B bus in latest chipsets which is mostly proprietary) weren't invented by them. So what? As an engineer - I fully sympathize VIA, they indeed are facing a lot of technological challanges. But VIA doesn't sell its products to engineers only, and regular customers don't want to know the intricate details of who invented what - they want a working final product.

From memory, 4.17 was the most problematic release. The first public 'fixed' release was 4.20, which wasn't enough so they pushed on and released 4.24. These are their internal builds, not service pack numbers. I think I found 4.23 quite acceptable, but it was withdrawn in favour of 4.24.

May I suggest some B12 vitamin injections, time? Those were releases, not internal builds. 4.17 was the stable release, which in fact did better than all subsequent 4-in-1s till 4.24 AGP-wise, on some boards by a huge margin. 4.23's AGP performance sucked bigtime.

The "686B" issue did not affect 99.99% of users. Maybe more. The fuss was because of the possibility of data corruption.

You're kidding me, right? "Possibility"? How can you use a system that "possibly" corrupts your files? This isn't even a serious discussion - system that "possibly" corrupts data is a no go. Period. And BTW, how the MTH issue was worse than this? All the people I know who bought i820cc (spare one who got it exchanged) are/were using them without ever noticing anything and never bothered to replace them (for free, BTW).

I'd like to know which 4-in-1 cost you 25-30% in games. I've checked most releases for exactly this sort of improvement/degradation, but never seen it. You wouldn't be exaggerating by any chance, would you?

Don't think so. 4.20 vs 4.17/4.24. On my P3V4X (Apollo Pro 133) which I happen to use to this very day those were the results. If you look hard enough on the web - you'll find enough evidence of this, somewhere between 25% and 30%, don't remember the precise number, that's why I used a range. I don't have to look for it - I've seen those on my own monitor!

Look, VIA is a great manufacturer as far as cheap-ass chipsets go and in fact forced Intel to lower their own chipset prices in the past. But they also have a nasty history of bugs, incompatibilities (SB Live! anyone?), and "get it right on A revision". And I don't think this should be downplayed. "Lest we forget"! Especially as this whole VIA issue was brought up due to AMD's partial reliance on VIA for support for its CPUs.

 

[GIANT]

. .

There was a pub called the Fox and Hounds. One day they were getting the big name on the front repainted. The landlord left the painter to do their thing, but when he came out a bit later on he had to stop them and point out their error. He said "You need to leave a space between Fox and and and and and Hounds".

...the painter_ to do their thing,...


...between Fox and and and and and Hounds.


Pradeep, you are in serious trouble! Expect a knock on your door soon by the the the the the Tasmanian Ministry of Language.


. .

 

[HellDiver]

In statements using conditional future such as "If I were a..." the subjunctive "were" is even a requirement (if we're talking about proper English).

Actually it's 'present unreal conditional', nothing future about it.

Come think of it, I did give the wrong example. "If I were (simple past) ..., ... would + <verb>" is indeed Present Unreal Conditional, including ""If I were a...". "If I were to + <verb>..., ... would + <verb>", on the other hand is Future Unreal Conditional. Yes, despite having similar sentence structure those two fall under Present and Future respectively. Why? I don't know, I didn't invent English, I'm only participating in attempts to distort it - same as a hefty portion of population in English speaking countries around the world, including Aussies!

We are illiterate...

Speak for yourself!

Actually I was (DAMN! I typed in "were" again but this time caught myself in time... :twistd: ) talking about myself and "a lot of those folks whom...". (Crapola! What was I doing during my school days anyway?!)

 

[CougTek]

Hi there, time! Nice to know your name's Tony too, time! HellTroll, you say? Then you would be feeding me, no? And doing so in an extremely poor way at that - you're really not up to speed on what you say. Next thing to clueless, in fact. How do I know? Because I've had to deal with those issues myself on many machines.
(...)

Come on HellDiver, this was the first thread I read where you participated and that you had not been an ass (up to the post I quoted) in a loooonng time. Please don't stop your medication and don't let your testosterone level rise too much. Calling you HellTroll was uncalled (I think), but replying with the style that made you infamous (extremely poor ... not up to speed ... next thing to clueless ... etc) won't arrange things. Time isn't Palestinian and you aren't Sharon you know (oh shit, hope I didn't open another can of worms).

Now make peace, don't throw in more fire and stick to the topic (be it the language or the AMD/VIA/Intel thing).

 

[HellDiver]

Coug's right (Not! Ok, ok, at least partially right... ).

My apologies, time.

 

[Cliptin]

Pradeep, you are in serious trouble! Expect a knock on your door soon by the the the the the Tasmanian Ministry of Language.

Aww, Come on now. Don't make fun of the man's stutter.

 

[Pradeep]

LOL, don't worry Coug, I blame the Absolut for my horror grammar.

 

[CougTek]

Err...Actually, it was Cliptin and not me.

 

[Pradeep]

and for the loss of all my mental functions as well obviously

 

[HellDiver]

Hey, Coug!

Speaking of Bartons, T-Breds in March and my eyebrows rised high up in general, I just came across this one on x-bit Labs (from Apr 11th). Looks like no Barton after all...

XP 2400 only in Q3? WTF are those folks up to? They better know something I don't!

 

[Pradeep]

Well we are in Q2 now with not yet a 2200+ in sight. So 2400+-2600+ by the end of Q3 is reasonable. I guess it all depends on how hard Intel pushes the p4 upwards.

 

[CougTek]

I saw the news you linked (X-bit Laboratories is one of my daily visits) and I have a theory about the delayed Thoroughbred and presumably cancelled Barton : AMD combined both.

Thoroughbred will only come in June (the 5th IIRC), so maybe AMD is tweaking its core with a few improvements that were originally scheduled for Barton. My guess is that it won't be anything major (bigger L2 cache for instance), but I wouldn't be surprise iff there would be subtle architectural differences. I'm probably wrong, time will tell (so David, what is it gonna be? ;-)). I hope (dream) that AMD would use SOI into Thoroughbred, but I don't really believe about it.

It will be sad if they don't, at least, increase the FSB up to 333MHz. Many argue that the Athlon wouldn't benefit very much of a higher FSB, but I disagree. While it's true for some applications, others show a 6-7% boost with the higher FSB frequency (some games especially). 6-7% is two processor grades nowadays, nothing to spit on IMO.

Let at least hope the lower 0.13µm models (2000+ and 2200+ that will replace the current Palomino) will be overclockable enough to reach 333MHz FSB as easily as the current Northwood 1.6A can go at 2.13GHz.

 

[CougTek]

Sorry for the small mistakes (I screwed a "if") in the above post. My mouse didn't listen to me when I told it to click on "Preview" and it clicked on "Submit" instead. I'm sure it did that just to piss me off, but it will pay!

 

[Pradeep]

If a chip is going to be released in June, it would already be taped out and proceeding thru validation/testing, if not already being produced to build up stocks fo release.

 

[HellDiver]

When AMD just introduced the PR system [this time], Athlon XPs were a bit more than on par with P4s, winning in most meaningful benchmarks by a notch. By the time 0.13u 512KB Northwood A was introduced, the PR system began pushing it - Athlon XPs did better in some benches and worse in others. As I already stated somewhere on SR at the time, a weighted choice between Athlon XP and P4 would be based on specific tasks the rig was projected to be running (as long as money was taken out of the considerations list. And since we're talking about performance here, money should be taken out of the equasion). Today top of the line P4 is running at 2.4GHz on 400MHz FSB with 512KB cache. Sometime mid-May, 533MHz parts should be announced. Which means that by mid-June Intel is likely to have 2.4GHz Northwood Bs in volume.

Unless AMD does something to the PR scheme by the time T-Bred is introduced, I think those numbers will lose any meaning when compared to 533MHz 512KB Northwood Bs clock frequency. On the other hand, I don't think T-Bred will be introduced in 2400+ (whatever that index may mean at the time) - usually it takes time to ramp up clock speeds on a new process (which 0.13u happens to be). In a word, I think AMD can kiss the performance crown goodbye for some time, especially if we are to look at projected model intros in Q3.

And now - for the monetary considerations (i.e. price/performance). Intel CPUs were never particularly cheap compared to the competition. So, generally speaking the price/performance advantage is AMD's for the time being, IMHO. But the price differences are to shrink somewhat due to today's (AFAIK there was a price drop confirmed today by Intel, at least according to a couple of retailers I'm in touch with) and mid-May's price drops on P4. This, coupled with the performance-related ramblings above results in a very grim outlook of AMD returning to a position it occupied throughout the majority of its history - a position of company cranking out cheaper but lower-performing x86 compatible CPUs...

Yes, I'd say at the moment things are looking grim... :x

 

[HellDiver]

Speak of the devil... Ermm, VIA chipsets reliability I mean, and here you go : Mike Magee being his usual pleasant self greets VIA today with his "A year ago today" writeup. To eliminate possible misunderstandings - this is a re-post of a year old article, not a new discovery.

P.S. I swear I didn't talk him into it, time!

 

[Pradeep]

I believe there will be an AMD price cut either today or tomorrow.

 

[HellDiver]

Yup. AFAIK it was supposed to be today, dunno if it actually happened. The nubers should have been approx as follows : XP1800+ ~= $130, XP1900+ ~= $160, XP 2000+ ~=220, all prices in US$ and in quantities of 1k and up, I believe - i.e. for retailers, no sure way to extrapolate consumer prices from those.