Article View: comp.sys.ibm.pc.hardware.misc
Article #203Re: Octec DCA2 DynamiCache??? Advice and info needed
From: John Saunders
Date: Thu, 11 May 1995 00:00
Date: Thu, 11 May 1995 00:00
47 lines
3193 bytes
3193 bytes
Anssi Saari (s106275@cs.tut.fi) wrote:
> In article <1995May10.030123.2574@rd.scitec.com.au>,
> johns@rd.scitec.com.au (John Saunders) wrote:
> :>Alan Smith (cs547091@lux.latrobe.edu.au) wrote:
> :>
> :>> I am upgrading my motherboard and am concidering getting an Octek DCA2
> :>> motherboard which uses 15nS DymamiCache. Does anyone know much about
> :>> these?
> :>I have run Norton SI V7 on 3 486DX2-66 machine, 1 a DCA2 board, the other
> :>an SiS based VLB board and the third an Intel Saturn based PCI board.
> :>Guess what, exactly the same speed.
> Of course it is, Norton SI runs entirely inside the CPU's internal cache.
> Having all the memory run at 15ns does improve performance but not for
> applications that rarely access main memory.
If Nortons SI runs entirely within the CPU, then why do I get lower readings
when the L2 cache is disabled on a MB with a L2 cache?!? I would suggest
that a large percentage of Norton SI is CPU intensive but at least 35% of
the test involves external memory. If I remember rightly SI drops from
144 to around 90 on my MB with L2 cache disabled.
> The DCA2 is nice, but EDRAM is very expensive, you can only put 16M of
> it on the board and there is no Pentium motherboard which could use it.
That's one of the negatives about it. You are investing money in SIMM's
that can't be used when you upgrade to a pentium.
The only benifit of DCA2 over a normal chipset that I can see is that
a cache miss gives 35nS access rather than 70nS. Also some chipsets
might use fast page mode and give the same performance levels with a
standard configuration.
Does anybody know what the cache memory to normal memory ratio is on
EDRAM SIMM's? We might find it's equivalent to 512K cache for 16M
of main memory (or higher). That would account for more performance.
I don't think the performance comes from where the cache is situated,
it's more to do with other factors like cache to DRAM ratio and 35nS
for a cache miss.
--
-- . +------------------------------------------------------------+
,--_|\ | John Saunders - (Work) johns@rd.scitec.com.au |
/ Oz \ | - (Home) john@pacer.apana.org.au |
\_,--\_/ | SCITEC LIMITED - Phone +61 2 428 9541 - Fax +61 2 428 9933 |
v +------------------------------------------------------------+
Technical Details
Message-ID:
<1995May11.031421.3551@rd.scitec.com.au>Path:
arc1-ca.pugleaf.net!unknown!not-for-mailReferences:
<1995May9.034546.13283@lugb.latrobe.edu.au> <1/NsvokI5k5X083yn@cs.tut.fi>Show Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: 10d683,ce3f1e9962c7994a X-Google-Attributes: gid10d683,public X-Google-Thread: 163e5a,ce3f1e9962c7994a X-Google-Attributes: gid163e5a,public X-Google-Thread: 109e38,ce3f1e9962c7994a X-Google-Attributes: gid109e38,public From: johns@rd.scitec.com.au (John Saunders) Subject: Re: Octec DCA2 DynamiCache??? Advice and info needed Date: 1995/05/11 Message-ID: <1995May11.031421.3551@rd.scitec.com.au> X-Deja-AN: 102444222 distribution: inet references: <1995May9.034546.13283@lugb.latrobe.edu.au> <1/NsvokI5k5X083yn@cs.tut.fi> organization: SCITEC LIMITED followup-to: comp.sys.ibm.pc.hardware.misc,latrobe.compsoc,aus.computers.ibm-pc newsgroups: comp.sys.ibm.pc.hardware.misc,latrobe.compsoc,aus.computers.ibm-pc