Intel strikes back

[Bozo]

http://www.bit-tech.net/hardware/2006/05/19/intel_core_duo_t2600_on_the_desktop


Bozo :mrgrn:

 

[LiamC]

In the same vein

http://www.xbitlabs.com/articles/cpu/display/core-duo-overclocking.html

 

[CougTek]

AMD doesn't seem to have an answer ready to fight Conroe when it arrives in July. Intel will certainly be able to fill the demand in term of volume, so AMD is almost sure to lose a lot of the recently gained market shares. The K8L, which still might not be able to match Conroe's level for desktop/workstation performances, won't arrive before many many months. I don't think it will be ready before the beginning of next year.

I think I'll sell a lot of socket 479 systems in the second half of this year.

 

[Mercutio]

I think the rational outcome on AMD's part will be a move to more reasonable pricing. I don't know if they'll do it but to me it looks like the logical way to go.

One of the problems I have with AMD at the moment - OK, the ONLY problem I have is that there's a big unappealing gap between high-end Semprons (3400s), low end A64s (3000s/3200s) and basically everything else. Is the A64/3200 worth $30 more than the 3000? Is the 3400/3500 worth another $50 on top of that? And if I'm looking at $200 CPUs, why am I screwing around with anything that isn't dual core? What's my motivation for buying a $300 Venice 3800 vs. an X2?

If the A64/3000 or the Sempron 3400 was a $75 CPU I could sell them day in and day out. Remember $35 Duron 1700s? Wouldn't it be nice to see Sempron 2600s in about that position? AMD's a little too proud of its less expensive components at the moment.

 

[paugie]

If the A64/3000 or the Sempron 3400 was a $75 CPU I could sell them day in and day out.

I'd buy one.

 

[Tannin]

As I understand it, AMD are capacity constrained at present — i.e., they can already sell every chip they manufacture, at the prices they currently charge, so where is their incentive to lower prices? In fact, if they did, three things would happen:

• AMD sales volume would stay static (because they can't make any more parts than they are already selling).
• AMD profits would drop alarmingly (equal sales volume at a lower average selling price = disaster for any business).
• AMD's reputation would suffer: with increased demand and no ability to supply, you get shortages, and when you get shortages everybody hates you. Consider, as an example, the period around 600 to 1200MHz days when Intel switched over to a new process and had a lot of problems: huge waits for Intel product, paper releases of new parts, and long-term true-blue Intel-only shops switched to Athlons in their droves - to the point where long-term AMD customers like us were struggling to get parts too. Intel have never really recovered fully from that bad customer relations exercise - they have been struggling to retain mindshare and market-share ever since. AMD, for their own part, have never forgotten the severe PR problems they inflicted on themselves when they were unable to deliver K5s and K6 Classics in volume.

Finally, consider the Apple switch. Why did Apple go to Intel when even Blind Freddie could see that, given a switch away from the Power PC, the AMD platform was streets ahead of the Intel one, and looking good to stay that way for quite some time? One reason only: Intel had the surplus capacity to guarantee on-time supply in volume (and, one assumes, were keen enough on the extra business to offer a firesale price as well), where AMD could only promise a strictly limited number of parts - and at a price that reflected the fact that AMD knew they were going to be able to sell everthing they could manufacture anyway, whether the cheques had "Apple" written on them or some other name.

So, if you are waiting for AMD to drop prices by more than a token amount just once in a while to make things look good ... well .... keep dreaming. You won't see any substantial drop in AMD pricing until the company needs to drop prices in order to keep the production lines humming. And that will only happen when Intel finally start providing world + dog with seriously fast chips at attractively low prices.
When will that be? Wait and see.

Right now Intel are very good at talking the talk, can they walk the walk? Wait and see.

If Intel can deliver the goods, do AMD have higher-performance parts waiting in the wings to overmatch them? Or will AMD, lacking a higher-than-Intel performace part to sell, be forced to compete on price, as they did back in K6 days? Wait and see.

 

[sechs]

As I've said before, AMD really only has the price/performance to hold out to people. Right now, they hold pretty good pricing power, but once Intel's Core chips hit the market in force, they'll begin to lose that.

While there will be some losening of prices, I really expect them to just move on to higher clock speeds and, to a greater extent, more cores to increase performance.

 

[Mercutio]

Do they have a 65nm fab yet? Isn't that the thing that makes the Core* chips so wonderful?

 

[LiamC]

Yes. FAB 36 (next to FAB 30 in Dresden) has started 65nm production with parts shipping for revenue due July/August.

Chartered Semi in Taiwan (FAB 7) has also started production of 65nm chips—1 000 wafer starts/week, which as near as I can figure is about 7% of AMD's total production, ramping to 3 000 ws/w within the next six to nine months. Both Dresden and Chartered use 300mm wafers vs 200mm in FAB 30 (Dresden, 90nm) which gives you about 2.2 times the total number of die per wafer.

The process isn't what makes the Core chips good, or at least not only. It is a superb design in itself.

 

[LiamC]

Actually, I did some digging and Intel's 65nm process is not bulk Si like all their other processes, but Si-Ge from what I'm seeing. Germanium in the matrix is supposed to enable faster switching transistors, the numbers I'm seeing thrown around are 30~40%. I'll have to do more digging.

 

[LiamC]

AMD's Conroe counter attack 3.4GHz?

http://www.theinquirer.net/?article=32026

 

[Mercutio]

Is anyone here really tempted at all by these new Intel chips?
For that matter, is anyone really interested in a faster processor at all at the moment?

 

[Groltz]

Is anyone here really tempted at all by these new Intel chips?
For that matter, is anyone really interested in a faster processor at all at the moment?

 

[time]

The process isn't what makes the Core chips good, or at least not only. It is a superb design in itself.

Given Intel's overall track record with PC 'innovation', I'll take that with just a grain of salt, thanks.

An alternative viewpoint:

http://sharikou.blogspot.com/2006/04/clovertown-scores-revealed.html

 

[paugie]

If I had the money, I'd be tempted. I don't know if I'm glad I don't have the money. I guess I'm glad. This way, I don't feel any bit guilty. Which means, I am really glad!!!

 

[LiamC]

Given Intel's overall track record with PC 'innovation', I'll take that with just a grain of salt, thanks.

An alternative viewpoint:

http://sharikou.blogspot.com/2006/04/clovertown-scores-revealed.html

Sharikou and credibility

http://www.google.com.au/search?hl=en&q=Sharikou+AMD+fan&btnG=Google+Search&meta=

Never read much of his/her stuff. People whose opinion I respect are quite dismissive of this poster however.

 

[time]

Never read much of his/her stuff.

Well, I included the link so people could read it. And I'm not interested in his opinion, let alone what some big mouths have to say about him.

I'm sure the new Intel CPUs are a huge improvement over the P4 architecture - otherwise there'd be no point in Intel continuing in the processor business. But the way some review sites are carrying on, you'd think it was The Second Coming. With the benefit of hindsight, Intel's much trumpeted 'innovations' have rarely deserved the PR frenzy of the time.

And where are the useful benchmarks? For example, it will take a lot more than a few percent improvement to make up Opteron's advantage in servers. What about real applications like Photoshop and professional 3D design tools (as opposed to the ridiculous synthetic tests)? Conroe may very well excel in these, and that will be great, but at the moment there just seems to be pointless wanking.

 

[LiamC]

Never read much of his/her stuff.

Well, I included the link so people could read it. And I'm not interested in his opinion, let alone what some big mouths have to say about him.

There are people worth listening to and there are others. The trick is to sort out the two.

I'm sure the new Intel CPUs are a huge improvement over the P4 architecture - otherwise there'd be no point in Intel continuing in the processor business. But the way some review sites are carrying on, you'd think it was The Second Coming. With the benefit of hindsight, Intel's much trumpeted 'innovations' have rarely deserved the PR frenzy of the time.

This is a carefully orchestrated media campaign. Overkill really, and I think it will backfire to an extent. IMO, this time it is different, too many people's opinion whom I respect say so. My reading of the benchmarks tells me so (see below).

And where are the useful benchmarks? For example, it will take a lot more than a few percent improvement to make up Opteron's advantage in servers. What about real applications like Photoshop and professional 3D design tools (as opposed to the ridiculous synthetic tests)? Conroe may very well excel in these, and that will be great, but at the moment there just seems to be pointless wanking.

Ask, and ye shall receive. From this thread:
http://www.storageforum.net/forum/viewtopic.php?t=5737

Woodcrest—server version of Conroe. If anything, Woodcrest will be slower than Conroe due to slower memory, conservative chipset timings etc.
http://www.techreport.com/etc/2006q2/woodcrest/index.x?pg=1
http://www.realworldtech.com/page.cfm?ArticleID=RWT052306090721

 

[Mercutio]

Interesting rumors of an ATI/AMD merger.

 

[sechs]

Whatever happen to those AMD-IBM and ATI-Intel merger rumours?

 

[LiamC]

Dunno about the IBM/AMD, but I thought the AMD/ATI rumour sprang from Intel/NVIDIA rumours

 

[LiamC]

These AMD rumours about their secret weapon just won't go away

<insert gratuitous Incredibles quote>
Bob: [laughing] He starts monologuing.
Lucius: He starts monologuing! He starts like, this prepared speech about how *feeble* I am compared to him, how *inevitable* my defeat is, how *the world* *will soon* *be his*, yadda yadda yadda.
Bob: Yammering.
Lucius: Yammering! I mean, the guy has me on a platter and he won't shut up!
</quote>

But apparently we are meant to find out today and the secret is:
http://blogs.mercurynews.com/aei/2006/06/amd_puts_pressu.html#more

A dual dual-core socket for 4 cores of goodness.
AMD: We r dah f1rsT 2 quad-core. We r da W1nn3Rs!

 

[sechs]

This feels soooo stop-gap.

 

[LiamC]

More "big mouths" on Sharikou:

http://www.aceshardware.com/forums/read_post.jsp?id=120058129&forumid=1

 

[ddrueding]

This feels soooo stop-gap.

Maybe. But it is quad-core, and I thought I'd read elsewhere that it would be compatible with any of their AM2 chips, so it will be the cheapest way to dual-core/quad-proc we're likely to see.

Me? I want a faster core, not more of them.

 

[sechs]

After additional reading, all that "4x4" appears to be is dual-socket without the necessity of registered memory. Whoopee.

 

[Pradeep]

It could be a big deal for many, if it allows 2 physical CPUs without having to pay the price premium for an Opteron 2xx or Xeon 5xxx system.

 

[Tannin]

Excuse me posting on-topic here, but I have to confess that I haven't bothered following Intel's progress with their new chips at all closely. Frankly, it's been so long since Intel had a compelling product that I simply stopped bothering to consider anything that wasn't AMD.

Hey, how long has it been?

If, like me, you are primarily interested in price-performance, and only mildly interested in raw performance, and not at all interested in the performance of anything that costs so much that no-one will buy it, it has been a very long time. Let's count back:

• The Pentium 4 has never been attractive. The current ones are red-hot and dog-slow.
• The last two or three versions (I stopped even keeping track of the available versions this last year or four) have all been much of a muchness: blown away by any of the Barton-core Athlon XP chips, and not even in the hunt when you look at the Athlon 64 and higher-end Sempron range.
• The only P4s to even look like serious propositions were the ones that were around back when AMD struggled to get the big-cache, fast 166MHz bus Athlon XP parts out - i.e., back around 2000 to 2400MHz days. But even these were so darn dear that no-one wanted them. And the XP 2500 buried them.
• The less said about earlier P4 designs the better. Outclassed in every department, with the very first ones setting a whole new standard in under-performance at a massive price.
• As for the P4-based Celerons, there has never been one worth considering for an instant. Hopelessly off the pace.
• That brings us back to the last of the P-III based Celerons, and these were a damn good part. Reasonably cheap, and good little goers .... but slower than a Duron and more expensive too. Close, but no cigar.
• Late-model P-IIIs were good, but Thunderbirds were better. Faster and cheaper too. Still no cigar.
• Mid-period P-IIIs - 666s and 800s and the like - were similarly outperformed by the Athlon Classics a lot of the time, and even when they weren't, they were horribly expensive, in very short supply because of massive production problems, and the motherboard chipsets for them were a total disaster.
• The first-gen P-IIIs weren't even a new chip, just a useless SSE unit grafted onto the moribund old P-II, And the prices!!!! Just as well the BX was good, because apart from the motherboard chipset, there wasn't anything else to like about them.
• Celerons of this period were, frankly, abysmal. From about the 400 or 450 through to the 766, they got slower, and slower, and slower. Maybe there would have been an excuse if they had been mega-cheap, but they weren't. And by the time the excellent Celeron 800 arrived, Durons were even excellenter (as Tea would say), and cheaper too.
• Many liked the P-II. I never did. Even the 100MHz bus ones were slower than you expected, especially where it counts, on the desktop, and the 66MHz ones were mega-expensive slugs for much of their lifetime. Hell, Cyrix outperformed most of them.
• Then there was the legendary Celeron 300A. A damn good chip. A brilliant chip if you were into overclocking, maybe the best overclocker of all time, but we weren't and still are not - not our market - so, in reality, this was a great chip that we sold in only modest numbers because the wonderful K6-2/300 was even better.

Whoah! We have gone a long way back, and we are still struggling to find a compelling reason to buy Intel. Let's be generous and say the last Intel chip that you would sensibly consider was the Celeron 1200, or thereabouts.

That's years.

And like my customers these days, I've simply lost the Intel habit. I don't really like the idea of working with Intel chipsets (give me VIA or Nvidia any day) but I'm sure that I could get used to them easily enough, and they would have to be better than SiS, you'd think. But why? What do I have to gain? Why shouldn't I continue just using the chips I hve come to know and trust so well over so many years?

Intel, if you want to claw your market hare back, you are going to have to come up with something very special.

 

[Mercutio]

For what it's worth, Industry Press is claiming that a $250 2.4GHz Conroe will bury a $1000 2.6GHz Opteron 285 on performance and on power consumption. That's... pretty special.

 

[paugie]

$250 2.4GHz Conroe

The equivalent of Pesos 15,000. The most expensive processor I've ever had the pleasure of owning was P4,500, a Celeron-D 2.66 which was the dogs!

 

[Tannin]

Interesting, Merc. Trouble is, I don't believe it. It would be an absolutely astonishing thing for Intel to do - astonishing two ways: for many, many years they have barely been capable of challenging AMD on the performance front, never mind trumping them; and they have never, ever, ever offered value-for-money in that class - not even with the Celeron 300A, which was as good as it was only by mistake, or so everyone believes.

But then, Intel have to do something radical if they are to have any hope of regaining mindshare - they've been running on bullshit for the last 5 years, and eventually that stops working - so maybe, just maybe ....

I will watch with interest.

Mind you, the US$250 price-point ain't where the action is. By far the greater part of the market volume is around $100 US, and that's the category that matters. Any news there?

 

[sechs]

For what it's worth, Industry Press is claiming that a $250 2.4GHz Conroe will bury a $1000 2.6GHz Opteron 285 on performance and on power consumption. That's... pretty special.

Is this before or after the supposed price cut later this month?

And how will it compare with the comparable Athlon 64 chip (to which we should really be comparing, anyway).

 

[Stereodude]

Interesting, Merc. Trouble is, I don't believe it. It would be an absolutely astonishing thing for Intel to do - astonishing two ways: for many, many years they have barely been capable of challenging AMD on the performance front, never mind trumping them; and they have never, ever, ever offered value-for-money in that class - not even with the Celeron 300A, which was as good as it was only by mistake, or so everyone believes.

But then, Intel have to do something radical if they are to have any hope of regaining mindshare - they've been running on bullshit for the last 5 years, and eventually that stops working - so maybe, just maybe ....

I will watch with interest.

Mind you, the US$250 price-point ain't where the action is. By far the greater part of the market volume is around $100 US, and that's the category that matters. Any news there?

Exactly... Why would anyone be foolish enough to believe that Intel could make a better CPU than AMD... :eekers:




Oh wait... they have before...

 

[CougTek]

Anyone knows which currently shipping Intel chipset will support Conroe? The 955X, 945G or 975X perhaps?

 

[LiamC]

975X is the only current one I know of. Intel have a new P965 chipset (which is supposed to be the volume chipset of choice), but it appears they are having difficulty shipping it in volume (and looks to be the reason for the release date slippage).

 

[Mercutio]

Possible reason for Conroe delays?

 

[LiamC]

Oh, by the way, the new 965 chipset looks like it will use the ICH8 southbridge which won't have any IDE ports

http://www.theinquirer.net/default.aspx?article=32855

 

[Stereodude]

Oh, by the way, the new 965 chipset looks like it will use the ICH8 southbridge which won't have any IDE ports

http://www.theinquirer.net/default.aspx?article=32855

That seems a bit hard to believe, but I suppose it's possible.

 

[Mercutio]

Nope. They've been saying that it wouldn't for quite some time, actually.

 

[LiamC]

Well tonight the bullshit stops. At 12:00AM US East Coast, 4:00PM Aus EST the NDA on Conroe (Core Duo 2 in Intel-speak) lifts. Expect a flood of reviews from the usual suspects.