Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
AMD marketing has deceived the ENTIRE WORLD, except for folks like you and Tenchu.
Far from true but it is always a good laugh when even well informed people mix up
QHz and MHz when talking about AMD chips. The remarkable thing is even with a
good product and reasonably effective obfuscation scheme AMD still can't make
money. Perhaps that's why so many otherwise apparently easy to offend and outrage
people give AMD a free pass on their lies - they aren't making ill gotten gains, LOL.
arbitrary transistor clock speeds
What is this? How does transistor clock speed differ from wire clock speed, resistor
clock speed, or capacitor clock speed? And how can clock speed be arbitrary? You
may not have a firm grasp on the timing measurement of periodic signals but the
practical art has long been well established in this area.
For instance, here is the auditor's report for 2800+:
http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_756_3734^3746,00.html
Whatever happened to Arthur Anderson? Maybe QuackHertz was too sleazy even for them. ;^)
You can see the actual tests run. Elsewhere the tests are defined in general, I don't have that page at my fingertips, but this list is in this report.
And how does this dog's breakfast get converted into the 2800+ number? If the process is
so "scientific" there must be a clear formula right? So tell me why:
1) If QuackHertz is methodically based on performance how come it always has two zeros
as least significant digits? Go to the SPEC web site and see how many CPU2K scores end
in two zeros. The answer is obvious, the number was picked out of thin air.
2) If this scheme was based on relative performance compared to a 1 GHz Tbird as AMD once
claimed why isn't an Athlon 2800+ not rated a 2.80+? Or a 280+? Have you notice that with the
SPEC92 -> SPEC95 and SPEC95->SPEC2k transition the scaling factors were changed to
avoid confusion? In this case it appears that AMD deliberately chose its scaling factor to *cause*
confusion with real clock rates. Just like the Rolecks watches and guchi handbags sold by
back alley street vendors in Hong Kong.
The Athlon is a fine product and it is too bad AMD feels it must resort to deception to sell it. BTW
I wonder how AMD designers and engineers feel about the end result of their hard work being
sullied by association with such sleazy tactics. I know I wouldn't feel very good about it.
Perhaps it's time for me to re-establish my TMTA position.
P.T. Barnum re-affirmed.
What is your point? What were you thinking? Are you insinuating that Dew and I are a gay couple?
No, I merely observed you both seem to fantasize about wrestling so perhaps you share other
common interests. Not that there is anything wrong with that.
Depending on your response I will be reporting you as I strongly feel you are in full violation of the
rules of this board and feel that your post has crossed over the line as a full-fledged illegal personal
attack.
The lady doth protest too much, methinks.
If your hypocrasy and insecurity compels you to complain then don't let me stop you. Keep in mind
that this nonsense goes back to DewDrop's unfounded and paranoid accusation that wbmw and
myself were the same person. You were anxious to chime in with a "me-too" with his latest ad hom
so you are hardly an innocent party.
Dew, Just had a strange dream about tag-team wrestling. Can’t figure out why, but I
have an odd feeling that it has something to do with this message board…
I had a similar dream. How weird...
Perhaps you two "like minded guys" should get together and watch some WWF
and gladiator movies. Remember - don't ask, don't tell.
ASPs for Q1 were "down a tad from Q4, as we sold a lot of lower grade CPUs in Asia".
Kind of worrying. "Down a smidgeon" would have been better "news".
Please no more QS BS.
Please direct your comment to AMD.
At the very least they should change the "+" suffix on the QS rating to "-"
and add a little "your actual performance may vary" in fine print under
every usage.
For the first time, TMTA might really be a buyout candidate.
There are less expensive ways to acquire used office furniture.
I don't know where you are getting this 2000 number from but you could probably easily
exceed that by counting the CPUs in the system sales announced on the newswires
over the last month.
Every new architecture, even highly successful ones like S/360, VAX, and SPARC had
very modest sales until a critical mass of applications was achieved to drive sales
into production environments. IPF is just starting to achieve a core of applications
that can drive hardware sales. The uptake of IPF into the market has also been slow
because of the huge drop in business investment in IS since the dot com bust. New
hardware purchases are mainly replacements and upgrades for existing systems
(which favours the status quo) and transitions to new platforms, which are complex,
disruptive, and expensive, are have been almost universally delayed.
I'll bet you a number
of OEMs, who've never been thrilled with the Itanium family in the first place, see this as much
more than a "don't care". Rather, something more like a "last straw". Imagine being a sales guy
at a tier 1 OEM trying to sell Itanium-based products to customers. Your task just went from
difficult to nearly impossible.
Catch your breath and try to get a grip on reality. Most server processor families have had bugs
found in the field at one time or another that had to be remedied by a SW workaround or field
replacement. No field of human endeavor is free from error. Not computer design, aircraft
design, or biochemical engineering. Yet we all rely on servers to calculate our paychecks, jets
to fly us across the continent, and the safety of drugs and accuracy of diagnostic tests our
doctors may prescribe. This isn't the first bug to be found in a server processor and it won't
be the last. Any Intel competitor that tries to exploit this is setting itself up for a hard fall when
what goes around comes around. No buyer but a fool would accept an assurance from any
vendor that a bug like this couldn't turn up in *their* hardware. What an OEM or buyer does
expect is that the uP vendor will remedy any problem that turns up as Intel has reportedly
done.
Obviously not many. It's hardly more than 15,000 of Itanium II's in the wild.
Intel made more than that during sampling and it has been in full production since
last July.
No big deal, it was not a very promising platform anyway.
LOL, If all IPF did was replace MIPS, PA-RISC, and Alpha as intended that is about 350k
devices per year. With an ASP of say $2k that is a base of about $700m per year. Take a
big chunk of market share away from Sun and that can close to double. What was AMD's
revenue last year?
All non-trivial chips have bugs, it is a matter of how many and how severe. It is curious
that it escaped escaped detection for so long, the silicon has been around for more
than two years. BTW, it seems like a test coverage issue more than anything since
it is reportedly only occurs in a fraction of chips under specific code sequences with
specific data. Just like the P4/3.0 bug. Perhaps the test coverage gating process for
release to manufacturing at Intel needs to be revisited.
Probably because without PC3200 DDR or dual DDR channel chipsets in the mainstream
there is no real reason to push the Athlon FSB to 400 MHz.
YB, maybe a 400MHz EV6 bus is harder than it might seem.
Compaq had 0.18 um Alphas running with 533 MHz EV6 bus a couple of years ago. If there
is a problem with a 400 MHz FSB for 130 nm Athlon the problem is with the use of PC class
chipsets and circuit boards. The fact that Intel can push the P4 FSB to 800 MHz suggests
that it isn't a problem for PC class hardware.
As you see the Intel compiler properly picks the options automatically, thus making base score equal to peak.
Wrong, look at the submission file for I2. The tuning section says that they didn't bother tweak compiler
flags for individual SPEC programs.
+FDO: PASS1=-prof_gen PASS2=-prof_use
Base tuning:
C programs: -ipo -O3 +FDO -ansi_alias
Fortran programs: -ipo -O3 +FDO
Portability:
178.galgel: -FI
Peak flags same as baseline (basepeak=true set globally)
Note:
EFI shell utility 'cpuconfig' used
to deconfigure other processors
Now compare that to the Opteron submission tuning:
+FDO: PASS1=-Qprof_gen PASS2=-Qprof_use
icl and ifl are the Intel C/C++ and Fortran compilers
f90 is the Compaq Fortran compiler
shlW32M6.lib is the SmartHeap library V6.0 from MicroQuill www.microquill.com
Portability:
178.galgel: -FI -Fe$@ -link -stack:32000000
Baseline: C icl +FD0 -O3 -QxW -Qipo
Baseline: Fortran ifl +FDO -O3 -QxW -Qipo
Peak tuning:
168.wupwise: ifl +FD0 -QxK -Qipo -Ow
171.swim: f90 -Optimize:5 -alignment:dcommons -alignment:records
-alignment:sequence -architecture:k7
-assume:noaccuracy_sensitive -math_library:fast -tune:k7
172.mgrid: ifl +FD0 -O3 -QaxW -Qipo -Oa -Qprefetch-
173.applu: ifl +FD0 -O3 -QxK -Qipo -Qscalar_rep- -Zp8
177.mesa: icl +FD0 -O3 -QxW -Qipo -Oa -Qscalar_rep-
178.galgel: f90 -Optimize:5 -fast
179.art: icl -Qipo -Oa -Qunroll4 -Zp4
183.equake: icl -O3 -QxK -Qipo -Oa shlW32M6.lib -Zp4
187.facerec: ifl +FD0 -O3 -QaxW -Qipo -Qscalar_rep- -Qunroll1
188.ammp: icl -QxW -Oa
189.lucas: ifl +FD0 -O3 -QxW -Qipo -Qprefetch-
191.fma3d: ifl basepeak=1
200.sixtrack: ifl -Qipo -Oa -Zp4
301.apsi: f90 -Optimize:5 -fast
ONESTEP is used for all base and peak runs
In other words Intel doesn't see fit to play stupid compiler flag tuning whack-a-mole games to squeeze
out higher numbers for peak scores.
Or maybe, as I suggested once, they needed the extra space to investigate the technology taken from a crashed UFO...
MIght be something to it since AMD's profitability has crashed too since setting down the road to SOI technology.
We're finally back to the days where the front side bus can send data at 1/3 the processor frequency. I haven't seen this since the original Pentium.
Not quite. The original Pentium, the 5V, 0.8 um Pentium, ran its bus *at* the
CPU frequency, either 60 or 66 MHz. We will never see that again
In virtually every case, when I have seen an Athlon system tag, it says something
like Athlon XP XXXX+ at Y.YY GHz.
At the discretion of the computer maker and/or retail outlet. They don't want to be
in a legally compromised position when a customer figures out QS is a bogus
number, not a clock frequency.
That blows a few holes in your conspiracy theory, doesn't it? The actual frequency
is right there, on the tag, every time.
Wrong. When AMD brought out QS it insisted that vendors only report the bogoHz
figures. But AMD was too weak to make it stick and vendors and stores reported
both bogoHz and real Hz in contradiction to AMD's wishes.
Even with HP pushing Itanium, I just don't see how that architecture will go
up against IBM.
1) IPF isn't going "up against" IBM, it is competing with POWERx in the high end. IBM
itself offers IPF based hardware as you probably know by now.
2) The important thing is that it isn't just HP driving IPF up against POWERx. There is
SGI on the HPC side, and Unisys, NEC and others driving IPF in the commercial area.
So the battle for the big iron is between IBM (Power4+) and HPQ (Itanium). IBM has a much stronger hand here, IMHO.
IBM scores bigger here on the basis on a very expensive system level infrastructure surrounding its CPUs.
Compare the price and price/performance of these systems. Also compare the HP system with other IPF
systems on commercial workloads. It is obvious that HP has a lot of room to improve Superdome. The
POWER4x is an impressive system but achieves it at very high cost that doesn't scale with the silicon. IBM
will be increasingly squeezed into the very high end and this will keep CPU volumes low and costs high.
And, no selling millions of 970s to Apple won't help a bit.
PC World is heavy behind AMD with top 15 list:
PC World is in the business of selling magazines. Saying that Intel
products are better than AMD products is not news. But saying the
opposite is controversial and garners attention, and increased sales,
web site hits etc from both Intel and AMD supporters. Newspaperman
call it a "man bites dog" story.
So what was the Opteron scaling factor that you don't like?
Nothing, they're really good. What I didn't like was the BS that AMD was
shovelling last year comparing Hammer and Xeon scalability with the
Hammer shown at about 100%. No, I didn't believe it for a second but
but it was taken by some of the more credulous AMD partisans as the
God's honest truth. Sheesh, I think some of those people would prepay
to buy a perpertual motion machine from JSIII.
While they may be impressive
These figures are very good. But as people at ACEs and other forums have pointed out
the scalability with clock rate falls short of the 100% that AMD showed in its pre-launch
puff material. Maybe AMD can come up quanti-scaling to bring it up to the miraculus 100%
level.
they still don´t mean anything to me
Then it is in good company. Quite crowded but good company nevertheless.
Athlon has shown that real world application performance and SPEC performance
don´t have too much in common in many instances
Don't hold your breath waiting for the rest of the computer industry to drop SPEC
in favor of superpi.
That doesn't address my question. The 970 and the POWERx series devices are quite distinct.
Even if Apple buys millions of 970s a year the POWERx devices will still be manufactured in the
tens of thousands per year for IBM's system division and will remain just as costly.
Won't sell it in high volume? What is Apple then? They will get desktop volumes which
will push down the unit cost of manufacturing the Power4 server chips.
Please explain how IBM Microelectronics manufacturing and selling PPC970 devices to Apple
can lower the manufacturing cost of POWER4 and POWER4+ paid by IBM's server division?
SPEC scores have to be reviewed internally by other members, i.e. competitors, before posting
and this reportedly takes weeks. Apparently AMD either 1) didn't want to tip off Intel early, or far
more likely, 2) was fine tuning the peak scores to the last possible moment. The SPEC web site
is usually updated late on Tuesdays so Opteron scores could possibly be up as early as 5 or 6
hours from now.
If you don't mind the price difference you can buy Intel processors all day but
don't slander AMD just because they charge less.
I buy on price and performance, in other words value. BTW, my current main home
system is an Athlon because, unlike earlier PC purchases, when I bought it the
AMD based system represented the best value.
I don't "slander AMD just because they charge less". I criticize AMD because they
use a marketing strategy based on deceit and justify it by assuming their main stream
customers are too ignorant to consider any information other than clock rate, real
or fictional.
I think it is reasonable to assume that the most important factor for the OEM will be cost over
performance.
Transmeta better hope so, cost over performance is the
cornerstone of their business model. Such as it is.
So do you beleive that an Athlon XP 2400 is somehow lower quality than a P4 2.4 GHz?
Lower quality? As in higher failure rate or shorter MTBF? I do not have information on
this either way.
However the two devices are not fully interchangeable commodities likes 2 by 4s or
pork bellys. The P4 will run some codes much faster than the XP while the XP "2400+"
will run some codes much faster than the P4. Labelling the XP "2400+" to make it
appear equivalent to the P4/2400 in the eyes of consumers at best smacks of hubris
and elitism, and at worst fradulent misrepresentation.
Intel is the big dog, and you know darn well they'll never allow a TPI effort to succeed, because
it would not reflect well on them.
Why would you assume that unless TPI was a crooked game slanted in AMD's favor? There
already exists an industry wide effort for accurately, reliably, and with repeatability, quantifying
CPU performance in a platform neutral and fair manner. It is called SPEC and both Intel and
AMD have been part of it for a long time. AMD has reported SPEC scores for Opteron, which
although quite good, are close to existing Intel products that have been in the market since
last year and will soon be surplanted by faster devices.
And as you say, PAE doesn't cut it. Ask any programmer what he'd rather code for and he
will say 64 bit flat memory model. Since machines with 64 bit flat memory model are now
available for no premium it's a no-brainer what to use for new projects.
Since when has the likes and dislikes of individual programmers mattered more than
business interests? Programmers absolutely hated the 16 bit x86 segmentation model
and many gushed over 68k computers like Amiga and Next computers. But where are
Amiga and Next now?
About $16B of Xeon servers sold last year. That is a absolutely huge incentive for ISVs
to code for PAE when server class application needs more than 4 GB of memory. Very
few of us, programmers included, have the luxury of deciding to work only on what is
considered easy, elegant, or our idea of cool; we work on what earns a paycheck.
I am sure that when Intel stops confusing the market with diluted MHz, and adheres to a more objective measurement of performance, AMD will follow suit.
Curious how DEC wasn't demonized for selling 500 MHz processors that were roughly
equivalent in performance as the 200 MHz processors from HP that they competed
against in the market place.
I'll type this slowly so maybe you can follow: M H z i s a m e a s u r e o f c l o c k
r a t e n o t pe r f o r m a n c e.
chipguy, it makes no sense. Most of consumers know nothing about "benchmarks"
and buy based on "Mhz". AMD will have to price Barton 3000 at the price of P4 2.2 Ghz
(i.e. crappy discontinued cpu), which will be wrong from all points of view except yours.
Too much of wishful thinking from your side.
So making up a fictional number that looks like a competitive clock rate to trick
consumers for their own good is better?
A lot of bad things can be done with this lame justification that "consumers are
just too stupid to understand so we will do the thinking for them".
Asian counterfeiters could argue that their Rolecks watches and Guchi purses are
just as high quality as the true upscale brand name goods and by tricking consumers
the consumers themselves benefit by paying somewhat less and still getting a measure
of status symbol prestige from their just-as-stupid acquantances.
AMD should drop the QS subterfuge
I don't believe that AMD has ever been convicted of any illegalities in connection with Quantispeed.
Subterfuge does not imply illegality. Webster is your friend.
This may or may not have something to do with the fact that AMD is not legally classified as a
monopoly
I am not aware that any uP vendor has ever been "legally classified" as a monopoly.
Why don't you try an experiment. Stay away from commenting on QS for one year and see
how often QS is questioned on the boards that you participate in.
I'd like to see a different experiment. AMD should drop the QS subterfuge for a year, classify
their processors by clock rate like everyone else, market its processors honestly based on
their competitiveness as shown by a variety of benchmarks and see if it continues to lose
money and market share.
It was just months later that AMD beat them to 1 GHZ. Ahhh, those were the days to be a droid.
I hope you enjoyed it. I don't see AMD beating Intel to any clock frequency milestones
in the foreseeable future. Even though I bet K9 has more than 12 pipe stages. A lot
more.
OK, how 'bout this: MHz sucks tailpipes. QHz sucks tailpipes, but not as badly.
This only works if you buy into the premise that Intel misrepresents clock frequency
as a performance measurement. I think it is a mistake to buy into this AMD partison
style argument but even if you do then I still think QHz is even worse than MHz. Why?
Because neither represents performance accurately (a single number never can)
but at least MHz has a physical objective basis in reality while QHz is a completely
arbitrary and fictionalized value.
But I don't buy the charge that Intel misrepresents clock rate as a performance. Intel,
like most other microprocessor vendors, uses clock frequency as a unambiguous
and objective way to classify different speed of devices of a given processor model.
The fact that Intel clearly distinguishes performance from clock rate is demonstrated
by the large variety of benchmarking data that it typically releases when it introduces
a new speed grade of processor.
P4 was the first CPU in the history of the industry that performed
significantly worse, clock-for-clock, that its predecessor.
Wrong. The 40 MHz MIPS R3000A was succeeded by the 100 MHz MIPS R4000PC. The
clock for clock performance of the new chip was about half of the new chip.
And I thought you actually knew something about this industry.
Aside from the fact that this is clearly demonstrably wrong (remember the PIII 1.1 Ghz mess?)
Ummmm that was three years ago. I am talking about current policy. You don't suppose the
two could be related somehow? ;^)
LOL, did you not see the smiley? I think you're taking yourself a bit too
seriously today.
Yes I know you can do that, I've done it myself. This thread started off about cooling and noise
of workstations. You buy a workstation to stick at *somebody's* desk - even if many other
people use it remotely. If you buy it to stick in a closet or server room and use it only remotely
then you should have bought a cheaper machine without a graphics head instead. And AFAIK,
everyone else in the world but wbmw calls those machines servers. :-P