On Sun, 11 Jun 2006 15:50:57 -1000, NoNoBadDog! wrote:
> "DRS" <> wrote in message
>> Intel also got no initial benefit when they shifted to DDR2. The shift
>> from 939 to AM2 per se was not intended as a major architectural
>> overhaul although it provides the platform for one. Conroe will have
>> an initial performance boost over AMD but by early 2007 we should be
>> seeing the 65um K8L 2nd generation AM2s. AMD can and undoubtedly will
>> make the necessary improvements such as supporting more flexible
>> reordering of Loads (a known K8 weakness), doubling the bandwidth of
>> the on-chip caches, increased SSE throughput, etc., etc. K8L won't
>> need to be revolutionary to compete with Conroe.
>>
>>
> There are some serious errors in your statements.
No there isn't. But there's some serious errors in yours.
> Intel benefited dramatically from the introduction of DDR2.
No they didn't. At least not dramatically as you put it.
> Since Intel does not have an on-die memory controller, the proc must
> still use the Northbridge chipset for all CPU-Memory I/O. Regardless of
> what the FSB rating is,
> the Northbridge chipset on an Intel motherboard operates at 400 MHz.
No it doesn't. It operates between 100MHz and 266MHz (266 on the latest
offering)
> (Intel plays number games to claim 800 or 1066, but the Northbridge is
> still 400 MHz). Since each operand requires 4 clock cycles on the
> Northbridge chip to complete, the effective speed of *any* Intel NB is
> 100 MHZ. DDR2 immediately gave a nearly 100% increase on an Intel based
> system.
>
You need to understand the difference between the FSB, which runs between
the CPU and northbridge at 100-266MHz on Intels and the memory bus that
the ram is attached to, which runs between the ram and the northbridge. If
the FSB isn't fast enough to support the bandwidth of the ram bus where
does that leave you? It leaves you with one big (actually small)
bottleneck. So with the early P4's running a 100MHz QDR bus, how does the
extra bandwith of DDR2 help? It doesn't at slow FSB speeds. And even with
faster FSB speeds the actual access to external ram plays only a small
p[ercentage of the system performance roll. And I don't have a clue where
you came up with the FSB of any Intel MB being 100MHz. The officially
support 100, 133, 200, and now 266MHz They list these at 4 times the
actual clock rate.
> Since AMD had an on-die memory controller that can communicate directly
> to the RAM at the core speed of the processor, and that bandwidth
> exceeds that of DDR2, DDR2 is of no benefit for AMD. In order for DDR2
> to benefit a system, that system must be using the NB FSB for CPU-RAM
> I/O.
>
Wow, are you screwed up.
> AMD chose to migrate to AM2 for several reasons. One of them is that as
> we go along, DDR2 will begin to replace DDR,and the RAM Manufacturers
> would rather produce DDR2 than DDR, so DDR will eventually suffer the
> same fate as RAMBUS and other RAM types that preceded it.
>
While what you say here is true, it's not the major reason. The major
reason was the fact that the cpu's are approaching speeds where ram
bandwidth starvation will occur, thus slowing down performance.
> In addition, the Conroe is not living up to it's hype (as everyone
> anticipated it would not). Intel originally claimed a 40% power
> advantage over AMD. Then it backed it off to 20%. Now that independent
> testers have Conroe, it looks like it's real world advantage is more
> like 8 to 9%. This is based upon non-synthetic testing, using the same
> apps you and I would use. Intel, when it tests it procs for numbers,
> uses a proprietary suite of tests that have no bearing on real world
> use. Intel also only uses tests that it knows will provide inflated
> values.
>
Well, I can tell you I've been a long time AMD fan. The last Intel I used
was a 486SX20, but the introduction of the core 2 cpu might change that if
Intel regains the price/performance edge. IOW's i go where the bargains
are. If it's Intel then it's Intel. I don't own stock in either Intel or
AMD. Bang for the buck for me. The name on the cpu doesn't matter to me. I
am partial to AMD, but if the margin gets too big, I'll be an Intel
customer again.
> All things considered, I think AMD does not have too far to go at all to
> better the Conroe. As for Merom, AMD already has the Turion64 X2, which
> is true dual core (Intel...NOT), is a 64 bit processor
> (Intel...NOT)...you get the picture.
>
It's nice you have that much faith in AMD. Time and Core 2 prices will
tell.
> Intel is fond of the smoke and mirrors games.
>
You got that right for sure.:-)
--
Want the ultimate in free OTA SD/HDTV Recorder?
http://mythtv.org
http://mysettopbox.tv/knoppmyth.html Usenet alt.video.ptv.mythtv
My server
http://wesnewell.no-ip.com/cpu.php
HD Tivo S3 compared
http://wesnewell.no-ip.com/mythtivo.htm