Already a Bloomberg.com user?
Sign in with the same account.
Yahoo! Eats Del.icio.us |
| Calling 'Em Like He Sees 'Em
December 11, 2005
2006: The Year When Intel Catches up to AMD?
Recently, I attended Intel's manufacturing conference in Oregon. There, an Intel vice president, Stephen Smith, talked about Intel's roadmap for next year. I came away convinced that, unless Intel stumbles on execution, it will finally catch up to smaller rival AMD in chip performance.
Sure, investors have long hoped that Intel would catch up to AMD – and have been bitterly disappointed. Over the past several years, the chipmaker canceled several products. Its server and desktop processors have fallen behind AMD’s in key performance metrics. As the much smaller AMD began churning out higher-performing chips, Intel’s share of PC and server processor market sunk from 82.1% to 80.8% in the past year, according to semiconductor consultancy Mercury Research.
But, finally, Intel might be close to taking AMD head on. In 2006, Intel will release several key processors that, Smith claims, will, by the second half of 2006, place Intel's products ahead or on par with AMD’s chips. “[With these chips out], Intel is back in the ballgame,” says Jim McGregor, an analyst with consultancy In-Stat.
Intel’s rise won’t be instantaneous. “Intel has stumbled, and it will take it a few years to gain its premium reputation it once had,” says Peter Hofstra, an analyst at AIC funds. Industry insiders believe that AMD’s momentum might continue through the first half of 2006. In fact, AMD hopes to nearly double its server chips market share and grow its desktop processors share in 2006, says Marty Seyer, senior vice president of commercial business and performance computing at AMD. “We are gaining lots of traction,” he says.
But AMD could start feeling the heat from its larger rival come mid- to late-2006 as Intel rolls out chips that will be at least as good as AMD’s in performance. Let’s first look at server processors, where, today, AMD’s “Opteron would smoke [Intel’s latest server chips],” according to McGregor. When Intel introduces its chips code-named Sossaman and Woodcrest, in the first half and the second half of 2006, respectively, that “will get Intel back into the game, says McGregor.
To beef up its desktop chips, where Intel has fallen behind in performance and “has spent two years without innovation,” according to McGregor, the company will right the ship with Conroe, due out in the second half of 2006. The new chip should be as good as AMD's, according to McGregor.
What’s more, Intel’s much-anticipated Yonah chip on a so-called Napa platform (it’s due out in the first quarter), will be better than AMD's mobile offerings due out next year, believes McGregor. Today, notebook processors are the only area in which Intel leads AMD in performance.
Of course, the proof is in the pudding. AMD won't stand still, waiting for Intel to catch up. And Intel will need to keep to its products roadmap and to execute well to make 2006 the year when it finally responds to AMD's challenge.
TrackBack URL for this entry:
OK. Maybe INTEL can catchup with the Opterons' today, but, AMD is moving to socket F with two memory controllers and other advanced features, AMD will also have glueless 32P computing...
On semi-conductor process, AMD+IBM developed technology to increase speed by another 40%.
Posted by: sharikou at December 11, 2005 08:11 PM
I think Intel is sandbagging a bit on the advances they have made w.r.t AMD. A new architecture supporting embedded MRAM is going to put Intel MUCH further ahead of AMD.
The Yonah dual core is already being tested and prepped by Supermicro for servers.
There is no way a dual core Yonah running at 100% load can consume less power than some chips standing at idle unless it has disruptive architectural changes supported the disruptive qualities of embedded MRAM...the holy grail.
Question...ask Intel how the Viiv platform becomes instant on/off, after intitial boot up?
Intel....you guys are sandbagging and I know it!!!
By EDVIRTUAL from the Great White North.
Posted by: EDVIRTUAL at December 11, 2005 09:54 PM
OK, so maybe Intel has some new chips coming out, but are they really good enough to catch up to AMD? Lets dig into them a bit further.
Intel's new server chips, Sossaman and Woodcrest, will still use intel's FSB. Sure its faster, and they have 2 now, but no on-die memory controller and no direct connects between processors like AMD's 2 1/2 year old opteron. The FSB will still be a bottleneck when you run multiple chips/cores, and AMD will easily win at scalability with existing opterons. And AMD isnt sitting still, releasing socket M opterons in the sametime frame, supporting 2007 quad-core architectures and faster hypertansport links. AMD will continue to stay way ahead of Intel in server chips.
In mobile, Intel's Yonah/Napa is nothing more then 2 pentium M's glued together. Here's a news flash, its also only a 32-bit processor!! Microsoft is already saying Vista will require 64 bits (see barrons article), so if you buy a Yonah laptop, your out of luck. In the same time frame, AMD releases a real dual core (turion 64 x2), that can run vista. Therefore if you want to future proof your laptop, it better be AMD Turion based. Plus AMD is releasing a new mobile socket that will support dual channel memory controller.. oh did I mention Yonah doesnt have on-die memory controllers either. How the heck Yonah then better?
Instead of listening to Intel's marketing machine, which was regurgitated in this article, I suggest reading what the President/CTO of VooDoo PC says about these new intel chips.
Posted by: SidMan at December 11, 2005 10:11 PM
Why do you mention the reverse situation as it is now:
Before: Intel has the lead, AMD catch on and Intel will release something next gen stuff to kill AMD's slightest chance to gain momentum.
Now: AMD has the lead, etc., etc., etc.,...
Intel is two years late from now. Open your lovely eyes and make a real job of reporting what the REAL situation at Intel.
What you forgot to mention:
- Quad core
- Fab36 with 65nm (world best fab)
- HyperTranport 3 (not Intel HyperHype 35 ok!)
- Socket F2 and Socket M
- Dual core Turion 64
- and more surprise to come...
Don't you ever look at http://www.amdboard.com????
Posted by: Karishou at December 12, 2005 04:26 AM
Sidman is WRONG about two things:
First, Yonah is not just 2 Pentium M's glued together. Unlike the Pentium D, there is a lot of coordination between the cores. http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2627
Second, Windows Vista will run on 32-bit systems. There are concurrent builds for Vista for x86, x64, and Itanium. http://www.winsupersite.com/showcase/winvista_editions.asp
EVERY edition, from the Starter to the Ultimate edition, has a 32-bit build.
Posted by: anonymouse at December 12, 2005 08:46 AM
Intel's only hope in the short term is to create a new market based on VIIV using the Yonah where 32/64 bits doesn't matter. Unfortunately, their plan for VIIV smells terribly like BetaMax.
How many more times will people buy into Intel's repeated "Just you wait" marketing? Intel has been saying "Just you wait" for two years now. This while AMD has actually been releasing better and better technologies two and three times a year.
Intel has thrown out their roadmaps twice already, and the one they have now doesn't appear all that solid to me. Sure, it's possible to show off next generation CPUs one, and even, two years ahead, but that doesn't prove they can be efficiently manufactured.
As long as AMD is ahead of Intel, as has been universally asknowledged, they don't have any strong need to talk up their plans for year 2115 as Intel is doing with their InSb based transistors.
Sometimes too much money makes for too many cooks that spoil the broth. I think Intel is a case study in this kind of problem. Didn't Intel just have yet another shake up in their management? Not a good thing to have happen at a time when they are losing their grip.
It sounds a lot like the "Shoot the messenger" syndrome. The message being: "We can't continue this way. We need to start over from scratch."
Posted by: ArtII at December 12, 2005 09:02 AM
It's very believable to think that Intel will catch up to AMD. With all due respect, though I'm all about AMD, I hope Intel does catch up. The One thing that can't be denied is that AMD has gained it's position in the industry because of it's ability to drive technology itself, and hence, become a great company of competition. One thing I've noticed that is, even further, less conventional than the innovations of their chips, is how they play the Market.
Sixty-five Nanometer will bring industry changing benefits to processors starting next year. AMD has a fab ready to start rolling that out, and continually newer process shrinks in the future, but isn't going until the closer part of mid-2006. Although it may seem a little slow, especially since they've recently announced their performance enhancing technologies bumping up an impressive 40 percent peformance, I think there is a reason they're going "play the market" and hold off on it's release. AMD is not a wierd company, and certainly not slow, but a company that does deliver when it says it will and does it at the right time.
It's not a race. It's business. It's the forefront of technology. It's our future. AMD is doing it right.
Posted by: Shyquan F. of Albany,NY at December 12, 2005 09:03 AM
Sounds like a lot of speculation to me. Lots of "so and so heard" and "processor should be" comments. No hard facts, no benchmarks, no concrete results. I'm taking this one with a grain of salt; sounds like a lot of Intel fluff.
Intel has always been notorious for being more of a marketing machine than actually responding to market demand. AMD has responded to consumer (DIY/Overclockers) sentiment with the introduction of processors capable of attaining speeds ~50% over their stock values. Actively endorsing (not warrantying) forms of overclocking and really just feeding the grass roots marketing efforts that come with this area of the market. AMD also actively encourages third-party board and chipset manufacturers; something Intel has never really encouraged (more like discouraged).
Again lots of speculation, lots of heresay; until I see results in both performance and delivery (NO MORE PAPER LAUNCHES) then I'm chalking this one up to the Intel Marketing fluff.
Posted by: Pillage at December 12, 2005 11:43 AM
Sidman, you are incorrect on a few points: The Pentium M is actually quite competitive with anything AMD has to offer. Hell, a 5.0 GHz CELERON will match an FX-55 clocked to 3.0 GHz, not to mention the Pentium M has a 12 or 14 stage pipeline, and has beaten the FX-53 before in several benchmarks while running on an older generation i855GM chipset, with no support for dual channel, DDR400, or any of the goodies the FX-53 was able to take advantage of. If the Pentium M wasn't a worthy competitor, they wouldn't be using it with all of their future designs.
Secondly, Windows Vista is NOT, a repeat, is NOT 64-bit dependent. While it will run faster on a 64-bit processor, it is merely taking advantage of those capabilities and NOT reliant on them.
The reason they did not add 64-bit extensions to the Pentium M is that it is a mobile processor, and obviously does not need it yet, as Windows Vista is not currently available to the general public. DER. When it does become available, Merom is waiting in the wings with 64-bit support. Intel isn't about to kill precious battery life on your laptop so it can support something you'll never use with the hardware you have.
I think it also speaks a ton for the Pentium M that it has beaten the FX series WITHOUT using an integrated memory controller. As far as Intel not using that, I don't know why they haven't, but suspect they will whenever the architectures are no longer efficient or fast enough to compete wth AMD's offerings. Tom's Hardware Guide has a full lineup of Intel's offerings through about 2009.
So while an I.M.C. definitely helps AMD, it has to be said that the Pentium M series is still faster than most of AMD's offerings, clock-for-clock. This is why Conroe is so big nex year: 4 MB of cache, 64-bit, potentially HT, dual-core, and all of that with the efficiency of Yonah, not to mention reports of a 30% increase in efficiency over the current generation Dothan chips, which are faster han some of the FX series offerings AMD has, as well as twice the cache, which is part of what makes the Dothan so fast.
I honestly don't see the hype that the AMD fan-boys are spewing about the Dual-core turion. The turion is just not that great of a chip in the first place, so it won't be able to compete with Yonah and Merom next year as they come around.
Intel has never discouraged third party chipset and/or board manufacturers. It just so happens that Intel chipsets have continually offered beter performance than any of the other availabilities. OEM's buy Intel boards because they're dirt cheap.
Intel is coming out with tri, quad, and octo-cored processors as well, that's not an AMD exclusive there, son.
Yonah is based on the 65nm process, nothing really for AMD to brag about there, not to mention Intel is already developing 45nm and 32nm chips.
Besides, Sossaman and Yonah benchmarks have already been out for a little while, and Sossaman, which is a server variation of Yonah, pretty much rapes anything else in benchmarks.
That, coupled with the fact that those are pre-production models, and are by no means final showings of performance, as they are generally better at production time, show that Intel's offerings for next year are a force to be reckoned with. Just because AMD throws some new technology into their chip, doesn't necessarily make it better or faster.
Posted by: Eliott at December 12, 2005 01:32 PM
this maybe true if Intel adopts Internal memory controller, HyperTransport, DirectConnect cores, and 64Bit. Maybe they just Buy AMD.
Posted by: Paul at December 12, 2005 01:36 PM
intel will never buy AMD.... AMD like's kickin intel all over the technology world to much.... it's to much fun for AMD to sell out ti frickin intel.... and AMD will still be able to beat intel in performance next year as well even with there 2 1/2 year old K8 that intel is only able to mach with there next generation CPU's how pathetic..... make's me want to puke when people say look intel's next generation can beat an Athlon 64.... well i hope it would because if it can't and it's intel's next generation.... that would be sad..... you can't compare 2 1/2 year old AMD technology to intel's next generation that's just stupid.....and turion is a very competitive product to the pentium m and
Posted by: jameskirk at December 12, 2005 02:48 PM
Wow, by these responses, it just goes to show how effective Intel's marketing hype is! When you dont have a technology lead, sure pays to have a big marketing dept. You are comparing Intel vaporware products, to real AMD products. Notice how AMD doesn't have to sell futures and lay out their entire roadmap. It is also assumed that intel can make these over-aggressive schedules for the new merom/conroe core. It has been a long time since Intel released any new core on schedule, wasn't prescott a year late? How many years late was Itanic? This is an entire new unproven core, built on a new 65 nm process. It takes years to develop a whole new core, even AMD took several years to develop the K8 used in opterons/athlon 64's. And as I stated earlier, Intel's new core, being rushed out, and likely behind schedule, will still underperform AMD's current K8. Intel is now talking 2008 futures in servers now, that's when it is promising to be competitive. See: http://www.theregister.co.uk/2005/12/12/intel_csi_low/
Also if Intel's new technology is so much better, why is Dell about to release AMD based products, thereby hurting is lucrative deal with Intel? Couldn't Dell just wait another 6 months for Intel to "leapfrog" AMD which seems so obvious to some here? I would say Dell knows a lot more about how real Intel's new chips are then the rest of the market, and is tired of loosing server business to its competitors. I am assuming everyone has been following the Dell developments, like AMD stockpiling chips for Dell, and Taiwan manufactures all ready to go building AMD based boards for Dell, not too mention dell has AMD chips listed on their site.
Clarifying the Vista 64 bit issue, according to Barrons... "When it ships in August 2006, Vista will need a 64-bit microprocessor to perform acceptably". Ok sure a 32 bit version will exist, and you can run it on any centrino laptop, including yonah. But acceptable performance will only be on an AMD Turion 64, running Vista 64. Add to the fact, some of the newer versions of software, including those from Microsoft itself, will only be available in a 64 bit version. So again, why would i buy an obsolete Intel yonah, over a 64 bit Turion?
Basically, while AMD's management was out there really listening to its customers, and developing the useful products, Intel's management was instead listening to its marketing department. This marketing department put Intel way behind, with terrible calls like Itanium for their 64 bit solution, or not caring about real x86 performance, just chips that run fast clocks. Now Intel has fallen 1 1/2 years behind AMD. If it hadn't been for some rogue engineers in Hillsboro (who sneaked in the 64bit code into prescott), Intel would be in much worse shape. That is really what saved Intel from a total disaster, not their marketing and management teams.
Posted by: SidMan at December 12, 2005 05:05 PM
I did not find the 90+ watt power envelope promising in Yonah, but more of a disappointment. Think about it, 2 Dothan cores that would burn 21watts each, rather than Yonah burning 90 watts in one packaging. I almost wanted to suggest that Intel stay at 90nm and launch 2 Dothan cores in one packaging.
You guys are not thinking straight. At 90nm AMD is playing a lot of games. First of all if you take into consideration that the Turion processor and the Athlon 64 are the same with the exception of power saving transistors, then you are looking at a normal Athlon 64 running at about 89 watts and a Turion running at 25 watts. If you now look at the difference in power consumption with a single and dual core Athlon 64 you get 20watts. Theoretically at 90nm AMD could have a Dual core Turion running at 55 watts (I could be wrong, but I believe that AMD has already tested Dual core Opteron HEs at this power envelope). The Opteron HE processors are intended for Blades.
Imagine what kind of power envelope a 65nm Turion could run at? Are you starting to get my point now? In Anandtech’s article the Yonah was running roughly on par with a 3800 X2 processor, which runs at 2Ghz.
AMD has a production problem, not a product problem at current, until the new Fab starts ramping and by the end of next year AMD will have a production capacity problem again until 65nm chips start ramping.
You also need to take into account that Intel measures power handing of their processors at 75% peak performance where AMD measures at 100% peak performance. This is why the Pentium D while on paper doesn’t look like it’s burning that much more power than an Athlon X2, but it actually burns a lot more.
If you think back to the Dothan’s launch the fastest processor was 2.13Ghz. The processor performance of the Dothan core now hasn’t changed much. Do you really think that the Yonah core is going to scale any more? No I don’t think so because power consumption will increase dramatically. Also I think that Intel is also using low power transistors like the Turion and the downside to this is that the processors can’t run as fast.
On a last note the reason why Dothan’s looked more energy efficient than early Turion products was a time to market issue. Early Turions were sent out with immature laptop components. I think the story should have changed by now since second generation products have been launched.
Posted by: Shawn Henry at December 12, 2005 07:25 PM
"and Sossaman, which is a server variation of Yonah, pretty much rapes anything else in benchmarks.
If you take a closer look of those performance bars, it may become a bit clearer just who is raping who:
2xDual Sossaman 2.0GHz: 884 (linear "forecast" of processor performance as it says below the figure, this is not even a real benchmark result!)
2xDual Sossaman 1.5GHz: 663 (4 cores)
1xDual Athlon64 X2 4800+: 638 (2 cores)
The 4800+ runs of course at 900MHz higher clock speed but any way you look at those results, Sossaman isn't blowing anyone's mind just yet...
Posted by: msironen at December 12, 2005 11:04 PM
According to Olga's articles (never mind the reality), for Intel, things always go from good to better, for AMD, things always go from bad to worse.
That is until you realize that Olga is not a reporter, but an Intel advocate.
We are in the 3rd year of Intel being behind in chip performance in 2 of the 3 segments (desktop and server). But we find out from this "independent technology reporter" nearly 3 years late, coincidently, when given a go ahead to report this from Intel directly at Intel's manufacturing conference.
Posted by: jh at December 13, 2005 12:23 AM
Eliott, what are you talking about? AMD's chips ARE BETTER than Intel's today! Opteron is superb; Athlon-FX is unbeatable; Turion's performance is no less than any Pentium-M, with much lower price and slightly more power consumption.
Intel MIGHT be able to come up with better products next year, but they HAVEN'T. Some think they will, some don't. Talking about it as if it's already reality is kind of silly IMO.
If you are buying a computer today, buy AMD. Pick whatever best available now. Tell me Intel's products are really better when the better products are in production, not in plan, not in paper.
Posted by: Edward at December 13, 2005 01:26 AM
Shawn Henry - Yonah DOES NOT dissipate 90W. Look again at the AnandTech chart. 90W is for the WHOLE SYSTEM.
These comments are generally ridiculous. With literacy in the US at this level, no wonder India and China are catching up.
Posted by: anonymouse at December 13, 2005 05:02 PM
Speaking of the Yonah power consumption at Anandtech. Early in the article, they mentioned that the setup was just like the one used for the X2 3800 ( http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2484&p=6 ). Since the Max TDP of the X2 3800 is 89W, lets remove that from the power consumption at full load (144W) and we get 55W. Presuming that 55W is also used by non-cpu items in the Yonah setup, subtracting 55W from the 108W at full load gives you 53W CPU power consumption (39W at idle). Perhaps this is correct, but I suspect that the electronics on the Intel MB would use less power than the Asus A8N-SLI. If so, then the actual TDP of the Yonah would be even higher. Regardless, we have a notebook chip that uses 53W at full load sitting on our laps in mid '06.
Posted by: RiRedneck at December 13, 2005 07:39 PM
This isn't an article. It's not news.
It is nothing more than a fawning report of what Intel's PR people had to say.
The writer should be ashamed.
Posted by: James at December 14, 2005 04:32 AM
Read the Voodoo Blog at www.rahulsood.com, the president wrote about this subject. Seems to be a much more unbiased view.
Posted by: John Strong at December 14, 2005 02:14 PM
RiRedneck's "calculations" are misleading. We do not know the efficiency of the power supplies used, nor do we know the core voltage of the Yonah sample used. The AnandTech test used a 945G desktop board. This is not a mobile product and thus it is fallacy to assume it consumes less power than the Asus board. Hard drives consume about 8-9W at idle, video cards like the Radeon X850 series used burn 19-20 watts at idle and 60-70 Watts under load.
Here is an alternative way to look at it. The Yonah has a 12x multiplier (166.7MHz FSB quad pumped) to run at 2 GHz. Pentium Ms usually drop down to a 6x multiplier at idle. Assuming this is the case, the Yonah runs at 1 GHz idle at some voltage below 1 volt. If you look at typical power for 90nm Pentium Ms at 600 MHz you get figures of 3-7 watts per Sandpile. I think we can assume that this is about the same or better on a 65nm Yonah due to process and architectural improvements. Note that the power dissipation on the Yonah systen in the AnandTech article only goes up about 16W. If Yonah eats 5W typical and 16W under load, this would mean TDP in the low 20's.
I suggest to you that the AMD X2 3800+ is not as power hungry as you claim it to be. AMD seems to make a TDP for a whole line of X2 chips, not to the 3800+ specifically.
Appologies for poor grammar - it is late and I wanted to type this out quickly.
Sadly, I think that nobody will listen. The rabid enthusiast community will always look on AMD favorably no matter what.
Posted by: anonymouse at December 14, 2005 10:24 PM
You folks expected too much off INTEL. Let me put the Yonah fantasy to the rest. Anandtech has done some benchmarking and power consumption measures. Yonah core is far inferior to Athlon 64 core in performance. Power consumption wise, it is at least 53 watts.
Posted by: sharikou at December 19, 2005 01:32 AM
I see that there's a lot of hope, goodwill and optimism for Intel, however not a lot of realism. The reality is that every year for the past three years (since AMD launched Opteron), somebody or another has said, "next year, Intel will catch up and pass AMD". We're still waiting.
There's nothing magical about AMD's rise to technological superiority. AMD just thought out of the box, hunkered down, and worked hard for several years and came up with a design that was like nothing seen before in its marketspace. Meanwhile, it's able to run all of the same software better than the old-fashioned processors from Intel.
Posted by: bbbl67 at December 19, 2005 03:43 AM
Want to play H.264 1080p HDTV video? 32-bit Windows chokes on it, but 64-bit Linux plays it effortlessly (videolan.org). Go ahead, get one of the HD movie trailers from apple.com/trailers and try it yourself. Java is another example where AMD64 architecture makes an extreme performance difference. 32-bit CPUs, whether for media work, databases, Java, etc are rapidly becoming OBSOLETE. 32-bit Vista will be provided for "legacy support" but anyone buying new hardware is going to go straight to 64-bit Vista and a DirectX9 capable GPU (which Intel's integrated notebook GPU can't handle). No one with a clue is going to buy 32-bit-only Yonah. A Turion dualcore notebook, however, is on my shopping list, though it would help if HP would quit being cowardly and build AMD notebooks with decent GPUs!
Posted by: Technojunkie at December 19, 2005 10:07 AM
Intel is not going to catch up in 2006, and not in 2007 either.
Intel's new stuff at 65 nm does not quite match up to AMDs 90 nm stuff and will be way behind again in about 6 months when AMD goes to 65.
Leadership in 2008 is still arguably up for grabs, only because there is still time for Intel to do something radical. But I dont expect it.
Posted by: Jim at December 19, 2005 03:52 PM