Aberdeengate, Sledgehammergate, HP Drivergate

You know I may just be a dofus aussie (yes I know everyone sighs and says damn antipodeans) but I thought the point of the Sledgehammer processor was to hit Intel where it hurts, right in the Xeon. The last major market that Intel really has to milk dry (yeah the notebook market is there but I think notebook makers are starting to come round to AMD) is the Xeon multi processor market.

I think the other thing to remember about the Sledgehammer is that the market it is aimed at is never going to be a huge numbers market, big profit margins maybe but not like the number of desktop CPU's moved. Anyone who thinks a lot of Sledghammers will be sold is dreaming for that fact alone. Not to say that I think it won't gain market share quickly, in competition with both Xeon and Itanic CPU's I think it will do very well.

Industry support will come but the markets must be built just as they were in the notebook market. AMD does have a lot of work ahead of it but at least they're keeping Intel honest.. (now who's dreamin, me).

Anyway that's just my ramblings from down under.

Athol Courtenay

Servers are about stability

I had not really thought about it, but the Sledgehammer really does not make sense and AMD should stop wasting time and energy on it.

Servers are not about processors. Servers are about having a robust and stable platform which draws from powerful processors, excellent chipsets, and quality system design and manufacturing. The key issue is stability and reliability.

AMD is not taking the steps to establish a record of stability and reliability. They have an excellent processor in the Athlon series. The chipsets for Athlons lag far behind the offerings of Intel and Server Works. AMD based systems lag behind too: I cannot get a high quality major OEM business class desktop using an AMD processor. Mostly there are cheaply built AMD clone systems available. >

I agree with Jerry Pournelle at Byte (reference this article: http://www.byte.com/documents/s=2292/byt1011813192258/ ) where he states “the AMD Athlon and Duron, although excellent processors, have been hampered by their dependence on mediocre third-party chipsets from VIA and others.” VIA still has PCI implementation bugs. SIS may have some nice chipsets now, but they have a long reputation of horrible chipsets in very low end markets. ALI is not a contender.

When I want to build a system for a customer I go all Intel: Intel processors, Intel chipsets. I do this to avoid headaches and problems for my customers. You can argue that AMD offers better price/performance. And I will tell you that the cost of a flaky system or driver issues more than negates that benefit. I have mostly AMD XP systems in my lab - but I don't bill myself on an hourly basis, and I live with the compatibility issues that arise and my lab systems can be down. Of course my server is an Intel OEM dual processor server. I don't trust AMD for my server.

I have found consistently over the years that going with quality components and quality manufacturers saves time and money. I want quality for my customers because a little extra up front will cost them much less money. When my customers do not take my advice and pick up “great deals” on “value” systems it usually works out great for me. The customers save a $100 by getting a system made of low end parts and they pay me several hundred dollars getting it to work. So it really does benefit me. This market is dominated by AMD.

None of the major OEMs offer a business class AMD solution. The business desktop market is dominated by Intel processors and Intel brand chipsets (except some very low end systems which really are not business class PC's). The press sees it too. It has been reported that the i845 chipset is as stable as the BX chipset. Intel chipsets are the benchmark of reliability and stability. Nobody makes such claims for AMD chipsets.

The major OEM server market is entirely Intel processors with a mixture of Intel and Server Works chipsets. No other platform exists. These vendors have spent years establishing solid reputations.

AMD is still hyping and focusing on processor performance. It will take years of focusing on platform quality and stability for them to become a viable business desktop and server platform. Is AMD ever going to shift their focus to stability and reliability of their platform?

Nobody in their right mind is going to set up an eight processor AMD box. If you can afford an eight processor box, you cannot afford downtime.

Sincerely,

Mark Pells

Danger: 4GB memory barrier approaching

The analysts seem to have forgotten that we have a 4GB barrier in 32-bit architectures like the Pentium and Xeon, and they seem all to eager to stand behind Itanium as the future for servers, and claim that the consumer desktop market doesn't, and won't, need 64-bits. This seems silly to me.

Currently, workstation and server-level markets (which are what Itanium is aimed at) are either running Xeons, which have the same 32-bit limitations as I mentioned, or are using RISC chips (i.e. Sun, SGI, IBM, etc.). Xeons are useful at the moment because there is more x86 code out there than RISC code, even for servers. But in the long run, they don't really offer any architectural advantages over a regular Pentium or Athlon, and they don't effectively break the 4GB barrier (key word there is effectively).

And, Itaniums are not very useful, because 1) there isn't enough Itanium-native software to complete with the other RISC platforms which are well established, and 2) Itanium doesn't run x86 stuff very well at all, since it's only using emulation, so it's no replacement for a Xeon-based system where performance matters (and since when did performance in a server or workstation not matter; that's kind of the point, isn't it).

IMHO, this truly makes Itanium a bastard child. Hammer, on the other hand, solves both problems: it runs legacy code at full speed, and provides a migration path to 64-bits when needed. Also, the 4GB barrier is going to be approached much quicker in server/workstation space than in consumer desktop space, since these systems demand more RAM (many 64-bit servers already routinely use way more than 4GB of RAM), but within a few years the barrier will apply to all levels of computing if the memory growth trend continues.

I'm not just trying to sound pro-AMD here, either. I have looked forward to seeing IA64 for many years, ever since I heard rumors of its existence. I had originally hoped it would be the future of Intel desktop CPUs - I figured they would support x86 backward-compatibility in hardware, and going foward would have a new 64-bit instruction set for new code. It would then debut in the server market, and eventually trickle down into consumer desktop space.

At least, if I were running Intel, that's what I would have aimed for. When it finally appeared, I was very disappointed in what I saw. I think the Itanic has really missed the boat, so to speak. The Itanium is an impressive piece of hardware, but it is a solution for a problem that nobody has. It also doesn't solve the biggest problem, which is the efficient migration of legacy code. The market, not the analysts, will make the final decision, and by the current sales figures of Itanic systems so far, I'd say it doesn't look good for Intel.

I also question their lack of 64-bit x86 development, since if the trend I projected is true, then their desktop CPU line will hit the 4GB memory barrier in a few years - surely they can't expect to eventually sell Itaniums at the consumer desktop level? If the rumors about Prescott and Yamhill are even close to the truth, it may very well be that AMD has actually solved one of Intel's biggest problems for them.

By the way, Mike, since you personally responded, I'd like to take the chance to say "great website". I've been reading your stuff for years and I really appreciate the unbiased straightforwardness and intelligence reflected in your journalism. I wish there were more sites like yours.

Jonathan Lanier

Reliability in the corporate sector

Hi Mike,

I am writing this letter in response to your article "AMD To Eat Intel's Lunch" which featured a letter by Robert Munro. I do not agree with the conclusions drawn by Mr. Munro because I believe that he both over estimates the performance of the Claw/SledgeHammer and underestimates the importance of reliability in the corporate sector, an area where AMD is sorely lacking.

First off, lets start with performance. Since neither the Clawhammer nor the McKinley have reached final silicon, everything following will just be predictions.

In Paul DeMone's article "The Spider and The Mountain" at Real World Tech (1), he estimates that with clockspeed taken into account the 1Ghz McKinley will offer 1.9 to 2.1 times the performance of a 800Mhz Merced in Spec CPU 2000. With an 800Mhz Merced achieving an base score 701 in Spec FP2000 (2) and an 379 in Spec INT2000 (3), a McKinley running at 1Ghz should achieve a 1400 in Spec FP2000 and an unspectacular 760 in Spec INT2000.

Once again, I refer to Mr. DeMone for performance estimation for the Clawhammer. I am using the Clawhammer for reference, since it is the only Hammer CPU that AMD has given a concrete demonstration and launch date for.

In his post in Real World Tech forums (4), taking x86-64 performance increases into account, the Clawhammer at 2.25Ghz should achieve roughly an 800 in Spec FPU (roughly the same as a Northwood 2.2Ghz) and a 1150 in Spec INT2000. Assuming that these figures are roughly correct (+/- 20%) then the Clawhammer will make an impressive showing doing integer calculations, but in floating point it will barely muster half the FP performance of the McKinley at twice the clock speed.

This is one area where the x86 ISA has always lagged behind and IA-64 has made a good showing. These two processors will be competing primarily in the low end server and workstation markets. In the server market, inter-processor bandwidth, cache performance but most importantly overall reliability are the most important features.

With a single McKinley receiving 6.4GB/sec main memory bandwidth and 3MB on die L3 cache, it easily has the Clawhammer's 2.7GB/sec and 512k-1MB (once again a guess based on AMD's manufacturing processes) beat. Also, there isn't a single AMD motherboard that features Hot-Swappable CPU's nor hot pluggable PCI slots, nor multiple I/O channels (like Intel's E7500 or the ServerWorks GC-LE series), which is a big hindrance to making a dent in the world of 24/7/365 servers.

In the workstation market floating point performance is the biggest factor in the success of a processor. This is one area where the otherwise beleaguered Alpha's have always done well, since they have always had excellent performance compared to their competitors.

Again the McKinley holds the performance crown over the Clawhammer. However, the Clawhammer will have a much greater software base, since almost every major scientific and engineering program is available for an x86 processor. It is however very unlikely that many programs will carry the 64bit optimizations that will give the Clawhammer an extra boost.

And then there is the question of general support. Can I count how many times pundits have screamed for Intel death in the past eight or so years? First it was the clone market, with AMD and National Semi making 386 clones for less money. Then it was the CISC vs. RISC, with the PowerPC based chips wiping the floor with the P5, then it was the Athlon, the ultimate Pentium killer. Now it's the Clawhammer, a 64bit CPU set to throw the more advanced, more powerful albeit riskier IA-64 into the wind.

These pundits seem very quick to jump to conclusions because they forget the things that has made Intel the company it is today: reliability, compatibility, and availability. The Clawhammer will most likely suffer the same fate as the Athlon MP/AMD 760MP in the server workstation market because of the RAS and overall stability issues that AMD still hasn't addressed (5).

As for the overall market adoption of the McKinley, I do not expect it to dominate for some time to come. Its competitors are still far too powerful, especially the Power4 and the USIIIe for it to compete in those markets.

Casey Esworthy 1
2
3
4 ,br> 5

Aberdeengate 1

Mike,

Good day. If Mr. Kastner is any kind of analyst at all he well knows that MHz/GHz ratings are what the average consumer looks at when deciding whether to buy a PC with an Intel or an AMD processor. That reason alone is enough for AMD to have gone to the PR rating for their processors.

It was either that or watch the average selling price drop or sales tank because they were perceived as slower even though they weren't.

I applaud Mr.Kastner's apparent concern that the customer not be deceived, however I have to wonder where his concern was last fall when Intel was matching up 2.0 GHz processors with PC133 SDRAM. It was still a 2.0 GHz processor and as we already know GHz sells, but the performance certainly didn't meet the expectations of someone putting out that much money for what they perceived to be a performance PC.

Pat Mitchell

Aberdeengate 2

Over at HardOCP they contacted AMD and AMD was able to refute the "facts" in the article.

I don't get one other thing... he [Peter Kastner] says that Intel paid him nothing or whatever, yet you said that Intel told you that they did pay for it. Care to comment on that?

Billy Bryan

Mike Magee writes: Just after I filed my second piece on the Aberdeen Group, in which AMD responded to the report and filed midday UK time on the 27th of March, I received a call from Intel. I had asked the previous day whether Intel had funded the report and the representative said that he had investigated and his company had funded the research, although he claimed that made no difference as to the objectivity of the report.

Aberdeengate 3

Hi Mike,

I thought you might find this interesting. Notice the date it was written.

Regardless of what Aberdeen says, it's pretty obvious where/who they get their directions from.

Intel is already a market leader, Aberdeen says (Tech Web news).

Now, that's quick market penetration. The Itanium 64-bit architecture from Intel Corp. (stock: INTC) has yet to be formally launched in volume, but already one analyst group is projecting it to dominate the market.

The Aberdeen Group, Boston, projects the chip and its successors to control 42 percent of all server revenue by 2005.

Aberdeen analyst Gordon Haff predicts systems based on the Itanium architecture will be broadly deployed starting in the first half of 2002 and will span a wide range of operating systems and application environments.

Here's a little more.

Aberdeen's White paper on "Who benefits from early Itanium adoption" is Google cached here.

And this, another white paper that is almost hilarious. The link is to a 21 page PDF. here.

It seems as though Aberdeen is a sophisticated press release research firm.

Bill S.

Aberdeengate 4

Mike,

I am a faithful reader of your site and first, I'd like to thank you for it. Second, I normally sit idly by as we all watch the battle rage on between two fierce competitors like Intel and AMD but the points rasied by the Aberdeen writer is like the pot calling the kettle black.

If he has such heartburn about PR ratings "confusing" the consumer, then what about Intel "confusing" the consumer by introducing a new processor family, the P4, with a higher clock but worse IPC and hence lower performance for the same clock speed of a P3? Intel abused the notion of price/performance with the P4 by realizing that the average consumer equates this to price/MHz.

His very comment reeks so bad of bias that he would have been better off not saying anything at all.

Regards

Steve

HP loses $1 million business through lack of drivers

I too had a similar situation under Windows 2000 with an HP3300C scanner about 1 year ago. I was running Windows 2000 so when I purchased a scanner being Windows 2000 compatible was a primary concern. At the local mass-retailer I took a look at the various models and on the side of the HP3300C scanners box was a Windows 2000 compatibility logo so I purchased the product.

Upon getting home and starting installation I plugged in the scanner, Windows 2000 PnP detected new hardware and asked to insert the manufactures CD/floppy with the driver. At this point I noticed the disk with the HP drivers made no mention of Windows 2000 on the label, only Windows SE and ME.

At this point in time I wandered out to www.hp.com and did a search around for updated drivers. Imagine my shock when there was nothing available for Windows 2000. I then called HP technical support and after waiting for a good amount of time and being transferred around for awhile I finally was talking to a human. That human informed me that the Windows 2000 drivers were not yet available but were planned to be developed and offered to put me on a waiting list for them. I expressed my disappointment in this ‘bait-and-ditch' game and told her I would return the product.

Trying to return a product for lack of support that is claimed on the side of the box proved futile so I still had the bloody scanner.

I then called HP back and got myself on the nice little waiting list…..

About four months later I was contacted by HP that drivers were ready and if I sent them $29.98 they would send me the CD with them. I was again a bit shocked and asked why the price? I was informed that HP had decided to charge for the drivers to support the product as Windows 2000 development took longer then anticipated. I flatly told them I was not willing to pay for driver support for something that was claimed on the box when I purchased the product.

I then went on mailing rampage with email to everyone from “Queen Bitch Carly” herself to any support/service/customer rep's I could find. Nobody responded in a helpful way.

What is so disappointing about this entire episode is that I used to LOVE HP. My first job as a programmer was in the early 80's on HP85 and HP86 systems and then onto the HP9000's. HP used to standard fro two things: “High precision” and “High price” and could do no wrong. Support was wonderful in the past. Service was wonderful in the past. I was almost a religious zealot for HP and our business was exclusively HP product and many others that I consult for were also HP due to my suggestions.

We have since put a stop on all HP purchases. No more Vectra desktops. No more NetServers. No more OmniBooks. No more Journadas. No more ProCurve switches. We went from spending over 95% of out IT budget annually on HP equipment to spending ZERO. Dell gets all our business now, which last year was $980,000. This year we've got another $1.9M budgeted for SAN and storage consolidation infrastructure. HP threw this all away for $29.98 and a lie on the side of a box for a $140.00 scanner…… What lies would Carly encourage to sell a $250,000 server?

Best regards,

John Welter

PS: Wanna buy a HP3300C scanner for cheap?