Umm... whatcha think this means?

[SammyBoy]

While browsing for motherboard reviews, I noticed <A HREF="http://www.xbitlabs.com/news/story.html?id=1025349410" target="_new">this</A> little article at X-bit.

As you see, the doubling of the L2 cache size provides a 600+ increase of the AMD rating. So, the rating of Athlon XP (Barton) can be obtained from Athlon XP (Palomino or Thoroughbred) working at the same frequency by simply adding these "pluses" to the rating of the latter.

Lets just say I nearly $h!t my pants. Yeah, I know it's a pretty unverified site, but x-bit seems to be putting a lot of stock into it. If this is true, a 2GHz Barton will have a 3100+ rating! Sounds rather insane to me. Too bad it isn't planned to be release until Q402, if the source is to be believed. Also, it mentions that the Barton production will all be done at the UMC fab, seeming to leave all of Dresden to produce Hammers. I still can't believe, though, that the extra 256kb of cache will result in that large of an increase. Thoughts anyone?

-SammyBoy

 

[TheMaverick]

Wow that is very nice if that is all true of course. That processor have a very nice performance increase right there lets see if we will notice it or not =)

So if that Barton is going to be a 3100+ at 2GHz how is the rating going to be with the Hammer?

AMD XP 1900+
MSI Ultra ARU (MS-6380E)
Geforce 4 Ti4600
PC2100 512MB

 

[eden]

Well if sources say the truth, an 800MHZ Hammer could compete a 1.6GHZ Wilamette, both having 256K L2, in Quake 3. One can imagine, a 512K L2 ClawHammer vs a NW B, would mean that a 2GHZ CH can compete or outrun a 3.6GHZ NWB. Pretty powerful eh?
So a 2GHZ CH should be using a PR 3600 or above.

--
Intel and AMD sitting under a tree, P-R-O-C-E-S-S-I-N-G!

 

[jihiggs]

that would be a site to see!

how do you shoot the devil in the back? what happens if you miss? -verbal

 

[ritesh_laud]

If this is true, a 2GHz Barton will have a 3100+ rating!

There is no way that adding another 256KB cache is going to improve performance by ~30% over the current AXP, unless they're redoing other parts of the core. Given that AMD's spending its time on Hammer, I doubt they're doing such a major revamp of the Athlon core. So if they're really going to use this new formula for the Barton's PR, it would be massively over-inflated and would lead to really bad press for AMD (similar to what happened to Cyrix and their PRs).

The other tantalizing possibility (more plausible too) is that this formula is actually used to compute the PR for ClawHammer!

Ritesh

 

[eden]

It could also be that they are improving prefteching... Just wondering though...


I don't get your last claim though.

--
Intel and AMD sitting under a tree, P-R-O-C-E-S-S-I-N-G!

 

[ritesh_laud]

I don't get your last claim though.

I'm speculating that the new formula is actually a leaked formula for Hammer's PR! In other words, I'm guessing that the Chinese site got word of the formula somehow but is misinformed about it being for Barton, it's actually for Hammer.

Just a wild guess from yours truly but remember that the ~30% boost in IPC (should be equivalent to boost in PR/MHz ratio) is, according to AMD from March 2002, <A HREF="http://aceshardware.com/read_news.jsp?id=55000468" target="_new">about what we can expect from Hammer.</A>

Ritesh

<P ID="edit"><FONT SIZE=-1><EM>Edited by ritesh_laud on 07/01/02 09:21 AM.</EM></FONT></P>