AMD Hits Milestone in Server Market

DontClickHere writes "According to data from Mercury Research, AMD has finally cracked the 10% mark in x86 instruction set server CPUs. AMD's Chairman had hoped that their server sales would hit 10% at the end of 2004, but they had only reached 5.7%. Some of this gain can be attributed to AMD's introduction of dual core chips in April this year. With Intel only due to ship dual core chips for low end servers later this year, AMD has been handed a golden opportunity to take a larger share in the server market."

Main Reason

[dsginter]

The main reason is that they sell the only 64-bit consumer chip. Yes, I understand that it is mainly marketing but the Athlon 64s are hot sellers. They need to crack Dell now.

Laptops?

[MarkByers]

I applaud them for their server sales, but I hope that they will soon develop a power efficient chip for laptops. At the moment they have nothing that can compete with Intel's M chips. Do they have plans to compete with Intel for this market or are they happy to stay in the server market?

How are dual cores counted?

[mev]

When calculating the percentage of processors, is AMD counting a dual core as one or two processors?

Dell is the decider

[soma_0806]

Everyone knows that AMD's share would seriously change if Dell could be persuaded away from their holdout status.

The two main reasons generally cited for Dell's allegiance to Intel is the millions in advertising and marketting (hard for AMD to compete when they're sitting on a little over a billion and Intel is sitting on something like 11 billion) and early notification of new developments.

The second one I just don't get. I mean, Intel annouced the Itanium in 1994 which consumers didn't see until 2001, two years later than projected and seven after the announcement. Really, how much notice does Dell need? Wouldn't they rather a company that actually gets things out in reasonable time frames?

Re:a thought...

[jurt1235]

Well, just overclocked an AMD3500+ (about 25% overclocked), and on some programs running 100% CPU (especially RC5-72), it outperforms the intel 3Ghz Xeon by a factor of 2! That way that Mack truck is pretty zippy too!

Re:a thought...

[tomstdenis]

What?

The K8 processors are way more power efficient then the K7s were. Keep in mind the K7 design came out as a competitor for the P3 processor not the P4.

The K8 is basically one-generation ahead of the P4. I'm sure Intel will catch up though as their Pentium-M is a good design in terms of efficiency.

A dual-core 64-bit Pentium-M would definitely give the AMD a run for some money I'd think...

But anything in the P4 camp and you're basically not making a rational comparison.

Tom

Re:Main Reason

[/ASCII]

I seriously doubt that the reason for AMDs sucess in the _server_ market is their 64-bit _consumer_ chips. As to whether the G5 and the 64-bit Pentium 4 are consumer chips or not, that distinction is pretty arbitrary, but since the 64-bit G5 can be found in the $1299 iMacs, I don't think you have a very strong case.

AMD should subsidize a switch

[ChrisF79]

I lease an Intel based server right now because back when I first signed up for it, that's all that was offered. Now the datacenter offers AMD based machines and I would love to switch. However, the company will charge me the setup fee on the AMD server and I will have to move everything over myself. It's not a steap fee, but it is enough for me to just stick with what I have. Switching isn't necessary by any means, just something I'd like to do. At any rate, I'd bet there are a lot of other people out there like me. If AMD would subsidize the costs for the datacenters to switch (pay them $20 per user that switches from the Intel machine to AMD) or whatever amount is fair, they could claim a bigger portion of the market.

Maybe it isn't feasible... just my $0.02.

Re:Laptops?

[stinerman]

Agreed.

I'm nearly an AMD fanboy, but I would have a hard time buying a notebook with their mobile processors in it. I think Via might have a good chance at cracking the notebook market with their new C7-M [via.com.tw]chip. Its max power output is 20W, while its idle output is only 100mW.

Re:Apple?

[the eric conspiracy]

Has the situation reversed?

Yes, at least on the 90m parts. I just built a dual core Athlon 64 system BECAUSE the power consumption is lower than anything Intel can offer in the same class. The Athlon 64 X2 was more expensive than the Pentium CPUs too, but i figure I will make the price difference back on power savings.

Re:Apple?

[AvitarX]

My guess is that like Motorola IBM is not focused on Desktop chips. IBM is focusing heavily on server and console chips while Motorola was working on embedded (I believe).

Apple know that with Intel there will be a long term focus on desktop/laptop chips weather they were there or not. This saves Apple from requesting their supplier to look at their needs (like gettin laptop G5s, which are applerently comin from IBM but after a lon time).

AMD would have met this requirement too, but Intel probably ave Apple sweetheart deal. Also AMD is still behind in the laptop area and may very well be forever now that Intel is shaping up.

Re:Laptops?

[creeront]

"Pentium M's are good because they are primarily based on P3 technology, not P4..." Pentium M's are based off of Pentium Pro Technology, arguably the best chip Intel has ever produced. http://en.wikipedia.org/wiki/Pentium_M [wikipedia.org]

Tier-1 supply sabotage?

[DamienMcKenna]

This week I had to spec out a replacement server when one of ours disappeared thanks to a delivery company, and I really wanted to get a HP DL145, HP's entry level Opteron server. As it turned out CDW's site said there was a two+ week delay in shipping the servers, whereas I needed one pronto. Given that other OEM [monarchcomputer.com]s have no problem with supply, I can only guess this may be part of AMD's case against Intel putting undue influence on the OEMs [bbc.co.uk].

Damien

Re:Main Reason

[laffer1]

I think it depends who you ask if AMD has won on the campus scene. I agree they have some great products, but how many large institutions actualy buy AMD? I think most companies and universities buy from large vendors like Dell, IBM, and HP/Compaq who mostly sell Intel based systems. In fact, most computers at my university are Dells. All dells are intel based... the remaining machines are HP/Compaq systems in a few labs or Macs. Thats why they only have 10%. On the server front, most machines are Sun sparc or Dell. There's a few legacy DEC boxes too.

To me, the best server offering for AMD systems is the Sun machines. I'd like to see more variety and price/feature ratio in the AMD server lineup. I know you can build servers, but i'm not in favor of that. When I think servers, i look at sun, dell, ibm, hp and apple.

Re:Turion

[bluGill]

Don't forget that the AMD chips have the memory controller built in, while the Intel chips require a separate chip sucking power (normally part of the North Bridge) to do this task.

Sadly I know of nobody who has measured who much the separate memory controller costs in power. Could range from insignificant to nearly as much as the CPU.

Re:Main Reason

[oconnorcjo]

But Pentium D doesn't have "Xeon" in the name so it obviously isn't for servers. Intel should know better. AMD was wise enough to come up with a new name for their chip to indicate that it was appropriate for use in servers.

The problem with InTel is that they are in a dilemna. They don't want x86-64 to take off because it was 1. AMD's idea and 2. Intel spent a fortune on the Itanic and were hoping to nudge out the competition (due to the fact that they patented the Itanium's instruction set). Intel knows that they have to sell x86-64 chips or let AMD run away uncontested but on the other hand they are not going to advertise that. Intel is in a terrible quandry. If they ever heavily endorse the x86-64 then that gives AMD a lot of credibilty and credit for being the leader and if x86-64 takes off big (as it seems it will) then Intel had better have a product to sell. This causes Intel to create these chips but intentionally obscuring the product line. They don't want to push this kind of chip. Intel would be estatic if x86-64 turned out to be a bust.

Long way to go.

[TheLink]

I still can't seem to find opteron _tower_ servers from IBM, HP or Sun (not even talking about Dell). And I don't think it's AMD's fault...

You can get 1U rack servers from those 3. And a 3U model from Sun. But if you want a "lowish end" cheap tower server you can't get it from any of the big names. Talking about something like one of Dell's PE 1800 servers.

The "bang for buck" sort of stuff. Say what you like, but you do pay a fair bit more for rack stuff and you can't stick 4 normal-sized SCSI drives in a 1U, or stick a fair number of NICs or other stuff in them. Tower servers generally make better "swiss army knife" servers.

Sure, one can get stuff from the "whitebox" manufacturers, but often there aren't enough PCI-X slots, or the frigging cooling/power isn't good enough[1], or you can't get 3 year next business day support with parts and labour (around the world would be good too)...

Sure us geeks can build servers. But most of us aren't paid to build servers for our companies - we have better things to do than to build, test, repair, and retest servers. At most we order a bunch, test them when they arrive, and tell the vendor - "This one is broken. Not paying. Swap it for something that works, and do by tomorrow".

[1] At my workplace we got three 1U servers from a noname manufacturer - and the CPUs _regularly_ throttle down due to heat (they use P4 class CPUs - nope I wasn't the one who ordered them).

Re:idiot

[jiushao]

Sure AMD is ahead in a lot of ways at the moment, the difference is far from dramatic however. People make a huge deal about 10%-20% differences in some benchmark (and the Pentium 4 still holds the crown in some areas, typically SIMD-friendly stuff).

Nothing against AMD (quite the opposite, haven't owned anything else the last decade), but their superiority was much more obvious to me with the K7 then the K8. The K7 and P4 were fairly equal in performance, the K7 won a few and the P4 won a few. The big difference was that the K7 was incredibly cheap, easily half or down to a fourth of the price of a comparative P4. The K8 does not really offer the same deal, we get slightly better performance overall but the prices are no longer the bargain-bin that AMD used to offer. It makes sense for AMD of course, but as a consumer I do feel a bit worse off than during the K7 days.

Intel on the other hand is working hard to get around their misstep with the P4 (and it is a real testament to Intels strengths that even what most people consider a failed architecture has stayed decently competitive over so many years), they have lowered their prices and are listening to market demand (making a very cheap dual core CPU and adding the 64 bit instruction set). I don't really think that Intel should be considered terribly evil, they listen to consumer demands where they could have harmed AMD greatly by making a sufficiently different 64 bit instruction set. Also; make no mistake, the P4 was not a marketing chip, it was just an attempt at a very innovative take on chip design. It did not pan out, but it deserves a lot of respect both for Intels guts to make it and the engineering that went into it. If they had wanted to make it a marketing chip they could easily have doubled the clockrate in marketing, the P4 ALU's actually run at double the advertised clockrate.

Overall things are looking good on the x86, decent competition between two companies who both really push the envelope in technology. Intels deals with OEMs should be looked into, but really, there are much much worse companies than Intel. I for one look forward to what Intel cooks up for the next generation.

I don't like coming to Intels defense over and over, but I feel that Slashdot is giving them less credit than they deserve. The P4 was an really interesting move (compared to the K7 for example which was just a solid take on very tried designs), and as a technological community I can't help but feel that Slashdot should appreciate Intel's attempt to try a somewhat different path.