[SI-LIST] Re: DDR2 at speeds faster than DDR2-533

  • From: "Douglas Burns" <dburns@xxxxxxxxxx>
  • To: <electronbob2@xxxxxxxxxxxxx>, <si-list@xxxxxxxxxxxxx>
  • Date: Tue, 7 Jun 2005 09:30:10 -0400

Bob,

Without seeing what your waveforms look like and understanding the
associated lab setup and probe locations, here are our thoughts on on the
issue.

Remember, the switching levels become tighter for DDR2-667/800 than for
DDR2-533(Vref +/- 250 vs vref +/-200, although some data sheets only show
the as vref +200/-250). Also, the more agressive device setup/hold
specifications are based upon the worst case of the nominal and tangential
DC-Vref and Vref-AC slew rates for ADDCMD, CTRL, CK, DQ, DM, and DQS.
According to specifications, the signals must swing between the appropriate
Vih/Vil DC/AC switching levels and then use the appropriate derating. This
provides a more accurate, but still conservative picture of the actual
timing margins.

In reality, receivers don't switch strictly based upon hard voltage levels,
but rather upon a function of the voltage and accumulated charge at the
receiver after crossing Vref. SiSoft has done extensive analysis on
receviers from multiple semiconductor vendors (including DDR2 receivers) and
found that by applying a charge model concept significant margin gains,
typically on the order of hundreds of picoseconds (up to a nanoseconds in
extreme cases) can be achieved over the current DDR2 derating methodology.
This is one explanation of why poor looking CTRL signals still result in
systems that exhibit stable operation.

SiSoft and Micron presented a paper at DesignCon2005 on DDR2 667/800 Mbs.
You can find a link to the paper and presentation at the following location:
http://www.sisoft.com/papers.asp

Also, we will be presenting a TechForum at Designcon East this fall. A link
to our abstract is below.

http://www.iec.org/events/2005/designcon_east/conference/tf-mp4.html

Regards,

 Doug

 Douglas Burns
 Vice President, Consulting Services/Chief Consultant
 Signal Integrity Software, Inc.
 6 Clock Tower Place
 Maynard, MA 01754
 978-461-0449 x14
 dburns@xxxxxxxxxx



-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx
[mailto:si-list-bounce@xxxxxxxxxxxxx]On Behalf Of Bob McNamara
Sent: Monday, June 06, 2005 1:50 PM
To: si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] DDR2 at speeds faster than DDR2-533


I bought a couple of "the ultimate game machine" DDR2
motherboards just to see how great they really were
(not because I wanted to waste endless hours playing
mindless video games).  ;-) One motherboard uses the
Intel chipset and the other uses the NVidia chipset.
I
decided to scope out some of the DDR2 signals to see
how much margin they have at various speeds.

I found that at DDR2-533, the signals look pretty good
and (at least for my sample of one) appear to meet
worst case timing margins.  Most of the
command/address
signals use 2T timing (i.e., they use two clock ticks
to become valid).  A few of the signals (CS, ODT, CKE)
must meet the more stringent 1T timing.  So I looked
at
CS as an example of a worst case signal.

At DDR2-667, the CS signal just doesn't hit the
Vihac/Vilac thresholds when it's toggling
high-low-high-low.  But there is enough of a swing
above and below Vref that you can see why the memory
might actually work well enough for the computer to
boot and even appear stable.  At least for my one
sample at room temperature and nominal voltage.

Of course as the speed increases to DDR2-800, the CS
signal only looks worse.  In fact it's so bad that I
wouldn't expect the system to pass a memory test or
boot.  But it *does* pass the memory test, boot, and
appears to be stable.

Has anyone else looked at this?  Maybe I'm not
supposed
to "look under the hood" when I buy commodity
performance hardware?

Can anyone point me at a system that really does meet
worst case timing at DDR2-800 and/or DDR2-667?

We've all heard and/or seen examples of how sloppy and
slipshod the engineering is for products in the PC
space.  So please don't waste bandwidth with a reply
unless you have something more to contribute than
"Yeah, it's really awful".

Thanks,

Bob

------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field

or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list

For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field

List FAQ wiki page is located at:
                http://si-list.org/wiki/wiki.pl?Si-List_FAQ

List technical documents are available at:
                http://www.si-list.org

List archives are viewable at:
                //www.freelists.org/archives/si-list
or at our remote archives:
                http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
                http://www.qsl.net/wb6tpu



------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field

or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list

For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field

List FAQ wiki page is located at:
                http://si-list.org/wiki/wiki.pl?Si-List_FAQ

List technical documents are available at:
                http://www.si-list.org

List archives are viewable at:     
                //www.freelists.org/archives/si-list
or at our remote archives:
                http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
                http://www.qsl.net/wb6tpu
  

Other related posts: