Slashdot Mirror
AMD Desktops Outsell Intel
prostoalex writes "For the week ending August 21st AMD managed to capture 54% market share among new desktops sold. Intel's share during the week was 45%. While Intel leads the U.S. CPU market with 82.7% market share, folks from AMD are proud to announce this is the second week this year - they also outsold Intel on the desktop market one time in April 2004."
Its for desktops selling within the retail channel. And Intel does have around 80 market share overall - its just that this past week AMD machines outsold Intel for some reason.
Maybe for enthusiast and home gaming PCs, but if you include business desktops I'd venture to say that Intel still carries somewhere around 75%.
The blurb itself says that despite AMD's share of new CPUs, Intel have 82.7% of the US market. Which is close enough to 72%.
The article itself admits that AMD's market is 'constrained' such that these results are very impressive. Intel indeed makes AMD a clear underdog for businesses and (at least up until very recently) notebooks.
ATI mainly outsold Nvidia because of Nvidia's shoddy manufacturing of early Geforce 5 series cards - poor drivers, drivers that lied, and late to market hardware that looked distinctly weak by the time it was public. This was a direct mirror of the emergence of Nvidia over 3DFX as a major graphics card force a few years ealier, with the exception that this time around, Nvidia had a lot greater cash reserve than 3DFX ever did, so could actually afford to make the mistake.
As it is, I'd be very surprised to find out that the ATI share was more than 55/45 in their favor (remember - a LOT of people outside of hardcord gamer circles are still using early Geforce / TNT cards - I have even seen Geforce 2 *MX* cards still being sold as low cost no frills acceleration) and with the new 6600 cards coming out, this is going to be a firm kick to the nether regions of ATI. There just isn't a card on the market that can hold a candle to it, and when you combine this with Nvidia's far superior Doom 3 performance, I'd certainly not bet against Nvidia becoming a dominant 3D acceleration force over the next few years.
Sunday you're Thinking Different, Monday you're a huge tool, paying too much and waiting to think like everyone else.
The VIA KT600 chipset is definitely slower than the exceptional NForce 2 chipset.
It is also cheaper.
Anything Intel blows goats compared to a decent Nforce 2 board.
That was total bullshit, not insightful at all. Anybody doing serious audio work on an X86 PC, even at the low end of the market, is using a breakout box for I/O and conversion.
Whether quality or no, the major thing AMD has going for it is price. For those of us non-gamers, who are fine with a pretty darn good machine, even sacrificing a little quality for a major difference in cost is worth it.
Snazzier than a Three-Piece Suit: http://kf.rainydaycommunications.net/
One of the major complaints I have about my XP2500+ is that the thing runs hot, like really hot. We hit a heatwave locally and temperatures were up to about 40 celcius at peak. My CPU actually hit 95 celcius (for those that use Fahrenheit, 100 celcius is boiling temperature).
I have a bigass thermaltake fan in there now, which I can turn down when the weather is cooler. The computer is still rather noisy.
My point to all this is not AMD bashing however. Apparently the 64-bit CPUs do much better for heat dissipation. The CPU die is much larger (the actual die is small on an 32-bit Athlon), so heat dissipates much more nicely into the heatsink due to the increased surface contact area. When I do upgrade, I'll be going AMD64... more power (in 'nix anyhow) and cooler running than my current CPU.
Professional Audio applications aren't running with onboard AC97. You'll have added a high-end card, or two, to your system. The only way the N/S bridge chips could add crackles would be if they weren't exchanging data properly, in which case nothing would be running correctly on your computer.
"It's the height of ridiculousness to say for those 9 lines you get hundreds of millions."
AMD numbers are based on (mostly?) retail sales.
All of Dell sales are direct.
Most of HP sales are direct.
Most of IBM sales are direct.
Most of Intel sales are direct.
I am referring to desktops in the gov, and corp market, as well as direct to customer sales.
So yes, AMD sells more retail.
Retail sales overall are a decreasing percentage of the desktop sales figures.
Makes for a great headline, but it is not true at all, not even close.
AMD does not have anywhere near the production capacity Intel has, and both are cranking out full steam ahead.
So do the math yourself.
if AMD has 20% of the capacity of Intel, and both are maxed out, who sells more?
wake up and hold your nose
I wish he was trolling, but I find his assesment to be quite accurate. I'll never buy another VIA-based mobo due to all the problems I've had with audio. It's a constant battle between various 4-in-1 drivers and sound card drivers, amongst other things. Troll my ass, he should be modded informative. I'm certainly not going to swich to a Mac, but my next computer will have an Nforce sticker inside of it someplace.
I wouldn't say that Intel has everything wrapped up. comparing a Xeon to a 3800+ is hardly fair as you are comparing a server processor to a desktop one. Now if you compared it to say, an Opteron (a much more fair comparison), well then you'd see AMD still wins or pulls up even.
What's more, the more processors, the better. Hypertransport gives each processor it's own bus.
That said, comparing an FX-53 to a 3.8 GHz Intel would also be a more fair comparison. And while it's true that the Intel wins it's share of benchmarks, keep in mind: You are comparing a 3.8 GHz Intel chip to a lousy 2.6 GHz processor (the FX-53). Theoretically, the Intel should totally kick it's ass - but it doesn't. That's some good chip design there my friend!
I just got a 3800+ last week. All I can say is: WOW!
"...Well, there's egg and bacon; egg sausage and bacon; egg and spam; egg bacon and spam; egg bacon sausage and spam..."
" I'll never buy another VIA-based mobo due to all the problems I've had with audio."
Soooooo, , don't buy another Via based mobo then. It's quite simple really.
What has this to do with AMD CPU's again?
"You can't fight in here, this is the war room!"
While AMD's x86 stuff is better than Intel's x86, its a bit sad that Itanium has lagged so far behind Opteron. Itanium's architecture is vastly superior to Opteron's, as it marks a break from the 20-year accumulation of old designs and legacy crap. It would be nice to see people embracing a new architecture for once.
Or course I can't claim superiority, having purchased Opterons myself. I guess software availability will always win out over good chip designs. Just ask those poor Alpha designers over at Compaq/HP.
Don't you hate meta-sigs?
where are all those intel favourable benchmarks?*
lots of amd favourable ones
in my personal experience, Intel's always have a small lag that is quite noticeable. Although this is comming from the same person who can tell a 85hz refresh rate from a 75 so its probably not something most people have to worry about.
and THEN there is the huge price difference
*(i wouldn't personally count office benchmarks like word but i know intel has a weird history of doing well in those)
I'll just use my special getting high powers one more time...
Actually, the lead paragraph mistakenly says the 82% figure is US -- it's not, it's ~82% worldwide.
The AMD > 50% figures are specific to US Retail sales, so they are totally uncomparable numbers.
Raw CPU speeds are fairly meaningless.
Its like the RPM guage on your car. Lets say that a Corvette has a lower RPM per mile per hour than a Porche and it also costs less. Now lets pretend that they both top out at 165 mph. If all you're worried about is how fast you get from point A to point B (and what else is there when talking about CPUs?), then the Corvette obviously gets you more bang per buck. Who cares if the Porche has higher RPM per MPH (its actually a bad thing!).
(hint: they're actually innovating)
honnold.org - sometimes-rock band, all the time awesome forum
Whenever it does a jump, many instructions are wasted
No. Branches generally do not cause a pipeline flush. (This is why branch prediction is a hot topic.)
HT does not exist to operate only in pipeline stalls. HT exists because analysis demonstrates that most x86 programs do not exhibit enough parallelism to fully utilitize all of the multiple execution units in a modern Pentium. You've got a lot of silicon devoted to peak performance that isn't used all the time, because you don't happen to have (for example) a bunch of full-width add instructions going on at the same time. HT allows a second thread to use those chip resources.
HT is cheaper than building two processor cores, as lots of the instruction fetch and decode logic is shared. Putting two complete cores on the same die does not increase the efficiency of utilization of the resources in either core. Dual core is much more of a brute force solution to the problem (a complaint AMD fans usually lodge against Intel). In this case, execution units in both cores will often be idle, as neither thread alone happens to need the full capability of a single core.
Since you've spent more silicon on the problem, dual core can have performance advantages -- specifically whereever you actually need that duplicate logic that would be shared with a HT design. Often, however, that extra fetch/decode logic is going to waste as well.
HT is an elegant optimization for a modern superscalar processor. It is not, however, the same thing as a dual processor, nor does it solve exactly the same problem.
*I don't like KDE all that much either, but then again I'm also not a big fan of GNOME
"[Regarding the 'cloud,'] ownership was what made America different than Russia." -- Woz
this doesn't include systems that people build for themselves, which more often than not are built around AMD processors.
easy, they know something you don't know. Mhz is just frequency and not speed. Marketing aside (which is what intel's overinflated mhz rating comparative to actual performance), the AMD K7 and K8 series CPUs do a lot more per clock cycle so a lower clocked chip competes with intel's higher clocked chips. This is much the same way that the Pentium M (Banias/Dothan) compete favorably to the Pentium 4. A 1.8Ghz Pentium M beats the 2.8Ghz P4 down.
I think CPU comparison vs car engines is a valid one. Engines just like CPUs have a number of different features each of which are important.
You can't just choose an engine based on one measurement, there are an assorment of features: horsepower, max RPM, torque, gear ratios, fuel economy; same with CPU you have GHz, cache size, memory controller, power efficiency.
A well engineered product will maximize all these features, or emphasize certain features over another depending on application. For example Pentium-M power consumption, or torque in a truck engine.
Higher-revving engines, properly engineered, also allow better fuel economy because of their smaller frictional loss.
Wouldn't higher revving = higher frictional loss, and less fuel economy? If you are moving something faster against a surface, and increasing the number of times it rubs against I would assume the friction would be higher. Isn't why cars are more fuel efficient at higher gears, since the revs are lower, more power/stroke. I would think in F1 the engines are high revving for faster acceleration
I'm not a car guy so was just curious about that.
D6 63 0D 70 89 81 BB 8E 7B 7C 5F 5D 54 EA AB 73
Yeah, HT works better on deep pipelines, but it's still a win for shallower ones. I couldn't find any more recent figures on AMD's pipeline depth, but duron was using a 20-stage pipeline.
That's not shallow.
I saw a graph based on HT simulations, it was CPU utilization versus degree of hyperthreading. Utilization didn't start to level off until you have something like 6-8 execution contexts on the chip (that's assuming you have work for all of them, of course.)
That's probably geared towards an Intel-style pipeline, but surely AMD would benefit from adopting the technology. Though since they're already going with dual-core, I'd rather see them apply it towards speculation to absorb cache misses than expose those contexts. That way single-threaded apps could get some love.
What about power consumption?
Get an MSI K8T Neo motherboard with an Athlon64. It can automatically vary its CPU speed from 800Mhz to its full rated speed (2Ghz+). So if you have its throttling control turned on, you don't worry about the Athlon64's maximum power consumption because it rarely runs at max speed. Best of all, although you turn this on or off in the bios, its controlled on the motherboard, so it works in either Windows or Linux (no software drivers).
There's no way in hell SMT gives you an 80% speedup; on the P4's it gives 10-15% at best. Thankfully, it comes at little-to-no-cost in terms of silicon, IIRC. SMT is nowhere near actually having two physical processors, it just utilises the power of the P4 more efficiently by just filling in the gaps in the execution stages. Some apps actually show a performance decrease under SMT.
Dual cores won't give you 200% either, even with the Opteron arch it'll still be 190% at the highest.
Moderation Total: -1 Troll, +3 Goat
At work we had this Dell XPS running a 3.4 ghz p4. That thing ran hot as hell. We had problems with the machine when it was rendering for 3dsmax and when we opened the case the heatsink was very hot. Actually, it was a pretty crappy heatsink considering the cost and thermal needs of th 3.4 ghz p4. Anyway, I'm assuming your joke was to point out how hot AMD's can get, well Intel chips can get very hot themselves.
So this totally random sales statistic is more interesting than a story about the Intel Developers Forum ?
.. Stuff that matters " , yeah right .
How about the showing of Montecito , a chip with 1.7 Billion transistors . "News for nerds
Intel chips, since at least the intro of the Pentium 4, have HIGHER max thermal output than AMD chips.
It's gotten so bad that the P4's performance is getting hurt... Some of the newer processors they introduced don't perform any better than the older/cheaper ones, because when they are at full-speed, they output too much heat, and the CPU has to slow down to keep from burning up.
Now, AMD has some power management issues, but even with that, they aren't any worse than Intel.
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant
VIA's Intel boards might be cheap and crappy as well, but they do not have the well-documented bug that several of VIA's AMD chipsets have that makes high-bandwidth PCI usage a fucking hassle at best and impossible at worst. Again, I DO NOT USE A SHIT SOUNDCARD. I said I speak from experience, so kindly believe me... Even if this is Slashdot, infamous for people talking out of their asses.
First off, yOu are quoting an article that is 6 months old. Second the article is noting the previous years sales numbers making the numbers at least 18 months out of date. You are trying to compare that to last weeks numbers? As far as production capacity. You have to look at it differently than you are now. I have worked at both places building FABS. Sure intel has more production capability, but not to produce the latest chips. Both companies can only produce their latest chips in the newest FABS. That levels the playing field for both of them. When you take into account that intel is on a smaller die size, that is gennerally going to mean that they are producing a lower percentage of 'good' chips. ones that pass the QC checks. You make things smaller, they are more prone to error until you iron out all of the kinks in the process. So FAB per FAB on Latest processor runs, AMD probably can produce more QC passing chips than intel right now. Sorry friend, but your wrong again.