Anandtech's Conroe vs. FX60

On Thu, 09 Mar 2006 12:32:00 GMT,
(The little lost angel) wrote:

On Wed, 08 Mar 2006 22:15:58 GMT, Robert Redelmeier
wrote:

I'm appalled Anand would accept Intel's build of an AMD FX60.
Too much room for sandbagging. Can't he build one himself?

I don't think they have a choice. Intel set up the two systems and
gave them an hour without advance notice. However since then Anand has
updated that article with a revisit.

He *did* say that he thought it was all honest - he just wanted/needed the
"scoop".shrug We'll see how much of a chump he feels in a few weeks.

Of course, not everyone is convinced. Rahul Sood has a counterpoint:

Rahul Sood's Weblog
http://voodoopc.blogspot.com/2006/03...ine.html#links

Some good points, esp BIOS & CnQ.

http://www.anandtech.com/tradeshows/showdoc.aspx?i=2716

The updated BIOS crept AMD upwards by 1FPS here and there. The main
killer was Anandtech configured the AMD settings for FEAR wrongly,
1280x960 vs Intel at 1024x768. But overall Conroe is still sitting a
comfortable overall 20% ahead of an overclocked FX-60.

That's a huge difference: 40% vs. 20%... and not too surprising when e.g.,
the single core executing FEAR has 4MB of L2 cache vs. 1MB for the AMD.
There are also things about the configuration of memory which are difficult
to know about - AMD has lots of options there (rank interleaving, bank
swizzling) which are not easy to cater for in BIOS Setup and difficult to
know where they "fit" for performance. Most BIOSs just hide them from the
user and take some hoped for average "optimum".

So even with AM2/DDR2 improvements, it wouldn't be unreasonable to
think that Conroe is likely to be seeing at least a 10% lead on the
competition at launch?

I don't see 10%, if that's what comes out of it, as that big a deal for the
few months it's going to exist. OTOH it could get AMD moving on
accelerating a part with bigger, maybe even shared, L2 cache. IMO they've
been playing too many games with cache sizes to differentiate parts and the
number of parts (s939 Opterons) has gotten out of hand - they need to
rationalize their offerings.

--
Rgds, George Macdonald

On Thu, 09 Mar 2006 10:59:47 -0600, Gnu_Raiz
wrote:

On Thu, 09 Mar 2006 04:57:28 -0500, George Macdonald wrote:

On Wed, 08 Mar 2006 16:51:51 -0500, Yousuf Khan wrote:

Anandtech did a test on a couple of Intel-supplied boxes one containing
a pre-release Conroe vs. an FX60. And exactly as Intel said, their box
beat the FX60 in every test by about 30%.

AnandTech: Spring IDF 2006 Conroe Preview: Intel Regains the Performance
Crown
http://www.anandtech.com/tradeshows/...spx?i=2713&p=1

One thing I don't get here is the large difference in game scores - I
thought that at those resolutions, 1280x1024, the CPU was overwhelmed by
the GPU in frame rate scores - has this changed?

Of course, not everyone is convinced. Rahul Sood has a counterpoint:

Rahul Sood's Weblog
http://voodoopc.blogspot.com/2006/03...ine.html#links

Seems like Intel has really taken the bull by the horns here. If they
screwed with the benchmark system setup, even unwittingly, they're gonna
get gored... and of course, the truth "will out".

Also, I have to ask: what the hell are they going to do about the umm, P4
inventory over the next 4 months? They really expect people to buy them
after this?... maybe put all the fabs on furlough?:-)

I was thinking the same thing but then I looked at the prices of the
PD and all my questions were answered. I bet Dell will have a lot of good
sales on desktops for the next few months.

I also see this a a job well done for AMD, just look at how long it took
Intel to catch up? When the release of AM2 happens, which people knew was
only a ddr2 update, AMD is only 10% behind. I would say that is one big
accomplishment. The real question is where are all the Intel server parts?
We all know that bread and butter is in the server parts, lets see some
benchmarks of those parts.

Well they *did* a pretty good job on the Dempsey/Bensley "advance"
benchmarks a few weeks ago where they got tecchannel to fall over and lick
boot: http://www.tecchannel.de/server/hardware/432957/ :-) (translates
reasonably well with Google Language tools). I'm still gob-smacked by
Intel's sudden interest in "showing" performance of parts which are 3-4
months away from intro - this has to be a watershed change in their
marketing philosophy... or just desperation?:-)

Also like the fine article mentioned, Intel was flogging chips but where
were the motherboards? Manufactures were showing AM2 boards but no AM2
chips. I think when June comes around you will be able to pick up an AM2
chip and have a good choice of boards to choose from. I also bet SM, Tyan
will have boards available, we all know how people love getting their
hands on product when its released. I don't know about you but if a
company has a choice of boards, and availability is good, then I see it as
good for AMD. I also see it as problematic for Intel as who else is making
the chipsets? I see this as a supply demand problem, the demand for the
new chipsets will be greater, and will encourage people to go another
route. If your chip is only 10% slower on some benchmarks, and price is
good then I don't see the gloom and doom that some predict of AMD.

The fact that Intel actually acknowledged the existence of a competitive
offering and had it present (for comparison) at IDF is, again, a momentous
occasion.

Would you want to buy a first generation Intel chipset given the last few
years, and Intel's current track record? All I can say is it's going to be
good for the consumer.

We'll see - the gamer market is fickle and loves overclocking and appears
to have deep pockets. If Conroe overclocks well, they could easily swing.
For the rest of us, yes, it's grrrrreat.:-)

--
Rgds, George Macdonald

wrote:
AMD down to under 30 by the end of summer, and back to at least around
40 about a quarter (maybe a little longer) after 65nm release,
assuming they don't screw up on it. If they do... well, they better
don't. Unlike Intel, AMD doesn't have such a luxury as being able to
afford to screw up...

Too geeky, most investors couldn't tell the difference between 65nm and
a barn door.

Of course it is - Conroe is at least a quarter away, so Intel will
lose a few more percentage points of market share by the time it is
released officially, and then some before it is awailable in
quantities from PC makers (AAPL is not one of them). Or alternatively
Intel will lose quite a bunch pushing Netbust at firesale prices. Or
a combination of these 2.

Back in the early days of PCs, Intel was able to successfully counter
Motorola by releasing the specs of its next-gen parts even before they
were taped out (or even designed for that matter), along with suitably
optimistic benchmarketing. It's trying that old tactic again. I don't
think it's going to to work this time, people are already pointing the
flaws with the benchmark setup here.

Yousuf Khan

The little lost angel wrote:
The updated BIOS crept AMD upwards by 1FPS here and there. The main
killer was Anandtech configured the AMD settings for FEAR wrongly,
1280x960 vs Intel at 1024x768. But overall Conroe is still sitting a
comfortable overall 20% ahead of an overclocked FX-60.

So even with AM2/DDR2 improvements, it wouldn't be unreasonable to
think that Conroe is likely to be seeing at least a 10% lead on the
competition at launch?

Well, the wierd thing is Sood's blog pointed out that one of the game
benchmarks showing Conroe beating FX, had it at something like 190fps
vs. 160fps. But a previous Tom's Hardware benchmark on the same game
but on a lower-end X2 (rather than an FX), showed 180fps for the X2. So
the X2 also beat the FX by 20fps!

It's likely that Intel just found the cruddiest chipset available for
AMD and pulled a magic trick. Wasn't this supposed to be an SLI setup
too? Wasn't there word recently that Nvidia disables SLI on non-Nvidia
chipsets? There was also word recently about a crack for the Nvidia
drivers that enables SLI on non-Nvidia chipsets. If this crack were
applied to the Intel system and not the AMD one, we'd see this kind of
result.

Yousuf Khan

On 9 Mar 2006 17:17:06 -0800, "YKhan" wrote:

Well, the wierd thing is Sood's blog pointed out that one of the game
benchmarks showing Conroe beating FX, had it at something like 190fps
vs. 160fps. But a previous Tom's Hardware benchmark on the same game
but on a lower-end X2 (rather than an FX), showed 180fps for the X2. So
the X2 also beat the FX by 20fps!

You can't compare this way because the benchmark used are different.
i.e. when Anand revisisted and retested Quake4 with their own usual
benchmark demo, Intel got an even bigger advantage.

For the one cited by Sood, Unreal Tournament 2004, it's pretty much
the same thing. The fps depends on the demo used.


It's likely that Intel just found the cruddiest chipset available for
AMD and pulled a magic trick. Wasn't this supposed to be an SLI setup
too? Wasn't there word recently that Nvidia disables SLI on non-Nvidia
chipsets? There was also word recently about a crack for the Nvidia
drivers that enables SLI on non-Nvidia chipsets. If this crack were
applied to the Intel system and not the AMD one, we'd see this kind of
result.

Erm, it's a Crossfire setup, ATI :P

More theoretical correlation can be done by looking at Xbit lab's
2.6Ghz FX-60 tested here
http://www.xbitlabs.com/articles/cpu...on64-fx60.html

While most of the benchmarks cannot be compared directly due to the
difference in test materials, there is one common synthetic benchmark.

In Hexus's PCMark CPU results, we have Conroe 6.8k (or so, they didn't
give numbers on their chart) and FX-60 at 2.8Ghz is 5.56K or so.

On Xbitlabs, the FX-60 @ 2.6Ghz shows 5340 (2053/Ghz), @ 2.9Ghz shows
5912 (2038/Ghz). Assuming simple linear, 2.8 should be around 5700 so
the Intel setup FX-60 @ 2.8 is performing a little below expectations
in terms of computational power but only very slightly around 3%.

Hexus did theirs before Anand updated the BIOS. Comparing with Anand's
figure after updating the BIOS, the biggest jump was in Fear around 4%
increase which is around the same ballpark. So I have to conclude that
Intel didn't appear to have done anything fishy as far as the hardware
setup is concerned.

The concern that Intel may have optimized the ATI driver when making
it work with Conroe but that would not be possible to verify since
nobody installed and ran 3DMark or something :P

--
A Lost Angel, fallen from heaven
Lost in dreams, Lost in aspirations,
Lost to the world, Lost to myself

On Thu, 09 Mar 2006 19:05:43 -0500, George Macdonald
wrote:

I'm still gob-smacked by
Intel's sudden interest in "showing" performance of parts which are 3-4
months away from intro - this has to be a watershed change in their
marketing philosophy... or just desperation?:-)

It is a good marketing move IMHO. They know they have a good 20~30%
lead on AMD's expected best at the launch of Conroe, hence comparison
with an overclocked FX-60 instead of just plain vanilla FX-60.

They know the only change for AMD until then, barring surprises, is
AM2, which so far have either performed poorer than S939 or on par at
best. So releasing performance figures now won't give the competition
any chance of countering unless that counter is already in place and
just bidding for the right time.

Bear in mind those who know about these are mostly going to be
1. Stock Analyst/Media
2. PC Hardware savvy folks
3. Net savvy Stock investors (non-Net ones would eventually get to
hear this from others)

Hence with these figures out in public, they put a damp on the AMD
success story, possibly push up Intel share prices, down AMD, relive
pressure from investors. Importantly, this would likely cut down on
AMD's sales for the next six months since non-time-crucial purchases
would probably be held back seeing this. These people are also the
people most effective in word of mouth marketing, if they buy into
Conroe later, they will become a very effective indirect marketing
tool.

For mainstream, these aren't likely to make any difference since how
many Joe Sixpack buying from Dell/HP and Walmart would be checking up
on Cebit news for latest advances in processor arch?

--
A Lost Angel, fallen from heaven
Lost in dreams, Lost in aspirations,
Lost to the world, Lost to myself

On 9 Mar 2006 17:00:27 -0800, "YKhan" wrote:

wrote:
AMD down to under 30 by the end of summer, and back to at least around
40 about a quarter (maybe a little longer) after 65nm release,
assuming they don't screw up on it. If they do... well, they better
don't. Unlike Intel, AMD doesn't have such a luxury as being able to
afford to screw up...

Too geeky, most investors couldn't tell the difference between 65nm and
a barn door.

But they understand stuff like market share. I expect AMD to be _at
least_ competitive vs. Intel's then-latest and greatest at 65nm
because at that point AMD is going to do some things not really
feasible at 90, such as quad core and double cache size, to name just
a couple. A good clock speed bump also might be expected. Greater
performance results in better benchmarks, and at least some of
investors, and a good proportion of the consumers, can interpret
things like Doom3 frame rate, or even SPEC (OK, maybe I overestimated
the brains of securities anal ysts here). Greater consumer
appreciation - more sales - (higher market share + higher revenues).
That's why I estimate about a quarter needs to pass before the numbers
understandable for investors (market share, revenues, profit, future
projection, etc.) are available.
And I have no magic chrystal, so my predictions are about as good as
anyone else's. If I could predict the stock market direction for
sure, I would've been a billionaire by now.

NNN

On Thu, 09 Mar 2006 04:57:28 -0500, George Macdonald
wrote:

On Wed, 08 Mar 2006 16:51:51 -0500, Yousuf Khan wrote:

Anandtech did a test on a couple of Intel-supplied boxes one containing
a pre-release Conroe vs. an FX60. And exactly as Intel said, their box
beat the FX60 in every test by about 30%.

AnandTech: Spring IDF 2006 Conroe Preview: Intel Regains the Performance
Crown
http://www.anandtech.com/tradeshows/...spx?i=2713&p=1

One thing I don't get here is the large difference in game scores - I
thought that at those resolutions, 1280x1024, the CPU was overwhelmed by
the GPU in frame rate scores - has this changed?

Well, the system was setup with 2 ATI X1900 XT video cards, which is
about the most high-end gaming setup money can buy these days. At
$500+ a piece for each of those two video cards the CPU definitely is
being overwhelmed by GPU in terms of cost, but perhaps not so much in
terms of frame rates.

Of course, not everyone is convinced. Rahul Sood has a counterpoint:

Rahul Sood's Weblog
http://voodoopc.blogspot.com/2006/03...ine.html#links

Seems like Intel has really taken the bull by the horns here. If they
screwed with the benchmark system setup, even unwittingly, they're gonna
get gored... and of course, the truth "will out".

Also, I have to ask: what the hell are they going to do about the umm, P4
inventory over the next 4 months? They really expect people to buy them
after this?... maybe put all the fabs on furlough?:-)

There are more then enough people who will still see 3.6GHz and think
that everything is well and good in the world. The retail box chip
market and DYI folks are probably going to turn away from the P4 in a
big hurry, but then again, this market is pretty much ignoring the P4
already. When it comes to HP, Dell and the other big OEMs though,
their customers will quite happily pick up some P4 chips for quite
some time to come.

-------------
Tony Hill
hilla underscore 20 at yahoo dot ca

On Fri, 10 Mar 2006 03:32:08 GMT,
(The little lost angel) wrote:

On Thu, 09 Mar 2006 19:05:43 -0500, George Macdonald
wrote:

I'm still gob-smacked by
Intel's sudden interest in "showing" performance of parts which are 3-4
months away from intro - this has to be a watershed change in their
marketing philosophy... or just desperation?:-)

It is a good marketing move IMHO. They know they have a good 20~30%
lead on AMD's expected best at the launch of Conroe, hence comparison
with an overclocked FX-60 instead of just plain vanilla FX-60.

Hmm, they seem to have you on the hook already;-) - you could at least wait
for "clean" independent testing results; their "lead" of 20%-30% is
basically on games and for more general computing appears to be 10-15% at
most. Hell we have no word on EM64T on the thing yet... and with Vista
coming up a year or so from now. As for whether it's a good marketing
move, time will tell - it's certainly a different approach for Intel...
which was the point I was making.

They know the only change for AMD until then, barring surprises, is
AM2, which so far have either performed poorer than S939 or on par at
best. So releasing performance figures now won't give the competition
any chance of countering unless that counter is already in place and
just bidding for the right time.

Oh, you have AM2 results too? Where?... prototype chip results?Ô_ö

Bear in mind those who know about these are mostly going to be
1. Stock Analyst/Media
2. PC Hardware savvy folks
3. Net savvy Stock investors (non-Net ones would eventually get to
hear this from others)

Add retail store clerks and people who read anal...yst babble. You think
there are still people who do stock gambling without Net now?... I don't
think so.

Hence with these figures out in public, they put a damp on the AMD
success story, possibly push up Intel share prices, down AMD, relive
pressure from investors. Importantly, this would likely cut down on
AMD's sales for the next six months since non-time-crucial purchases
would probably be held back seeing this. These people are also the
people most effective in word of mouth marketing, if they buy into
Conroe later, they will become a very effective indirect marketing
tool.

No need to spell it all out... but IOW potentially kill off PC sales for 4
months or so - quite a risk with their stock nearing bottom already.

For mainstream, these aren't likely to make any difference since how
many Joe Sixpack buying from Dell/HP and Walmart would be checking up
on Cebit news for latest advances in processor arch?

The tide has already turned there - AMD is ahead in retail (HP - Walmart
sells very few PCs AFAIK). As for Dell, they are already on shakey ground
because of their bottom-feeding prices. They badly need a good current
quarter, which is already shortened by a week due to lengthening their last
quarter - last thing they need is to have a fire-sale on their premium
systems.

--
Rgds, George Macdonald

On Fri, 10 Mar 2006 18:52:10 -0500, George Macdonald
wrote:

It is a good marketing move IMHO. They know they have a good 20~30%
lead on AMD's expected best at the launch of Conroe, hence comparison
with an overclocked FX-60 instead of just plain vanilla FX-60.

Hmm, they seem to have you on the hook already;-) - you could at least wait
for "clean" independent testing results; their "lead" of 20%-30% is

hehehe, I'm just bored with the state of things as they are. Eversince
A64, AMD hasn't really pushed ahead due to the lack of competition and
Intel hasn't really put something innovative on the table since. Plus,
I was trying to exercise the marketing cells the uni gave me :P

They know the only change for AMD until then, barring surprises, is
AM2, which so far have either performed poorer than S939 or on par at
best. So releasing performance figures now won't give the competition
any chance of countering unless that counter is already in place and
just bidding for the right time.

Oh, you have AM2 results too? Where?... prototype chip results?Ô_ö

Yup, prototype chip general results have been posted on various online
resources with earlier prototypes showing significantly poorer
performance, the latest one almost on par with S939 and supposedly a
new spin this week or next might make it good with S939. With another
3 months to go, I'm guessing that AMD could at best push it another
5~10% max.

3. Net savvy Stock investors (non-Net ones would eventually get to
hear this from others)

Add retail store clerks and people who read anal...yst babble. You think
there are still people who do stock gambling without Net now?... I don't
think so.

Which was why I said they will eventually get to hear this from others
:P

But the people who would be most interested and act on this
information are still mostly stock investors and those who already do
a lot of their own research.

PC Retail clerks aren't likely to start telling their customers not to
buy now for a good reason

No need to spell it all out... but IOW potentially kill off PC sales for 4
months or so - quite a risk with their stock nearing bottom already.

Only for AMD IMO, because the ones interested in max potential
performance are likely all buying AMD now. Those buying Dell's bottom
end model aren't going to care or be aware enough to care.

For mainstream, these aren't likely to make any difference since how
many Joe Sixpack buying from Dell/HP and Walmart would be checking up
on Cebit news for latest advances in processor arch?

The tide has already turned there - AMD is ahead in retail (HP - Walmart
sells very few PCs AFAIK). As for Dell, they are already on shakey ground
because of their bottom-feeding prices. They badly need a good current
quarter, which is already shortened by a week due to lengthening their last
quarter - last thing they need is to have a fire-sale on their premium
systems.

I know AMD is supposedly ahead in retail with some 80%, but that's
only in US. The rest of the global market is still pretty much
Intel's. This early release will only dampen the sales of AMD based
systems simply because again these people are buying AMD because of
the performance (top end AMD are as expensive if not more than Intel).
So my bet is the impact of sales delay on Intel is probably marginal.

Given their inventory level supposedly at pretty high level, it
wouldn't be surprising if Intel is already finalized on the Conroe and
ramping up production on that instead while waiting for the existing
inventory to clear, would it? And the potential double whammy by
launching it on the smae day or perhaps 1 day ahead of AMD AM2 at the
upcoming Computex? :P



--
A Lost Angel, fallen from heaven
Lost in dreams, Lost in aspirations,
Lost to the world, Lost to myself