DC Athlon FX in June!

[Mephistopheles]

...according to <A HREF="http://www.xbitlabs.com/news/cpu/display/20050410210218.html" target="_new">Xbitlabs</A>.

I wonder about its clock?... Maybe 2.4Ghz? Wow, this would make a kickass product. This will probably make Intel look rather silly with its hacked-up jobs...

 

[slvr_phoenix]

I don't think I'd care one way or the other about a 'hacked up job' so long as it worked just the same. What I would care about however would be the heat from an Intel dual core. If they haven't fixed their heat issues by then... :O

<pre> <font color=purple>I express to you a hex value 84 with my ten binary 'digits'. </font color=purple></pre><p>@ 185K -> 200,000 miles or bust!

 

[Rugger]

Take it on campouts and cook your hotdogs over the CPU.

__________________________________________________
<font color=red>You're a boil on the arse of progress - don't make me squeeze you!</font color=red>

 

[Mephistopheles]

I don't think I'd care one way or the other about a 'hacked up job' so long as it worked just the same.

Actually, you're quite right there. <i>So long as it works the same,</i> who cares? I mean, lotsa people keep whining about Intel not being technically dual-core, but personally, who gives a damn? It will work in just E-X-A-C-T-L-Y the same way! Who cares? Presler is even two chips on a die! But who cares? I certainly don't. I don't buy the product because of the aesthetical values (heck, I don't buy apple)! Who cares what the chip looks like if it works just the same? It simply doesn't matter. Period.

So I couldn't really agree more, slvr_phoenix. Thermals and bad clock rates and performance are the problem with smithfield, not the fact that it's just two prescotts on a die. Exactly.

Think about it, dual-core has no real added benefit if compared to dual-CPU Opteron and Xeon setups. Well, maybe a <b>very few</b> percentage points in performance because of chip integration and signal travel distances. But negligible at best. <b>It's most likely that a dual nocona 3.2Ghz setup will perform exactly like a Pentium EE (3.2Ghz) in most cases; in fact, I'm kinda surprised noone compared them both yet. </b>It should have been the number one test! I mean, all dual-core does is make 2-CPU systems more affordable, that's all. In technological terms, it's not exactly breaking news.

I'm also rather curious to see how a dual-core Opteron at 2.0Ghz compares to a typical dual Opteron 246 setup that has been out there for two years. I'm willing to bet that there is actually surprisingly little difference between them both. Except, of course, price.

<b>Why doesn't anyone test that?</b> Come on, I can't be the only one thinking this way about dual-core... Isn't there a single review site that has been given a Pentium EE to test and <i>also</i> happens to have a 3.2Ghz Nocona somewhere?... :frown: <P ID="edit"><FONT SIZE=-1><EM>Edited by Mephistopheles on 04/12/05 03:02 AM.</EM></FONT></P>

 

[endyen]

Well, that "hack job" may not get the job done as well as you think. There are penalties for external connects, and gains from internal connects. Oh well. looks like we will see sooner than later.

 

[P4Man]

Here is a comparison if dual core opteron versus dual cpu:
<A HREF="http://www.hkepc.com/hwdb/dualcore-opteron-5.htm" target="_new">http://www.hkepc.com/hwdb/dualcore-opteron-5.htm</A>
Note opteron "866" is the old code name for the 875. The article begins here:
<A HREF="http://www.hkepc.com/hwdb/dualcore-opteron-1.htm" target="_new">http://www.hkepc.com/hwdb/dualcore-opteron-1.htm</A>

As you'll notice, dual core opteron performs just the same (within a couple of percent) as dual cpu opteron. Still, this is pretty good, as memory bandwith per core is cut in half going from single do dual core. Now, the benchmarks they ran are pretty silly for a server setup, but my guess is that the decreased cache coherency traffic (which no longer has to go of die) makes this happen.

For Intel dual core chips, memory bandwith per core will also be cut in half, BUT it will not have the AMD crossbar advantage, and to keep the caches in synch, all snooping goes over the FSB which already has to feed 2 cores. There will be an impact. Maybe not big, but under real world loads, which also require I/O, I'd expect to see a smaller speedup on DC P4Ds than K8s. You indeed should not care much about the architectural "cleanness", but only about the results.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[Mephistopheles]

You indeed should not care much about the architectural "cleanness", but only about the results.

Point is, the difference being minimal means that the architectures of each core are far more important. This way, if things were reversed and P4 was built for multi-core setups, this still wouldn't be enough to reverse the fact that the A64 core is much better designed. How much worse is Intel's arbiter chip if compared to AMD's crossbar? We don't know the answer to that just yet, but in the end, it won't matter as much as the fact that the A64 cores are more sophisticated than the P4's.

Therefore, the problem with this whole story is that P4s are very ill-suited for dual-core because of their very high power dissipation. Because of that, they can clock even lower than they could on single-core (low for their initial ambitions, read). In any case, AMD's approach will make for better end products, and, regardless of product cleanless, a better result...

 

[slvr_phoenix]

Therefore, the problem with this whole story is that P4s are very ill-suited for dual-core because of their very high power dissipation.

And remember that this likely mostly due to Intel's refusal to use SoI so far. Once that changes the thermals of Intel chips may become much more akin to AMD's.

<pre> <font color=purple>I express to you a hex value 84 with my ten binary 'digits'. </font color=purple></pre><p>@ 185K -> 200,000 miles or bust!

 

[Clob]

People also forget the fact that this Dualcore setup means less space needed for two cores. We only need one HSF setup and we can have two cores on our favorate gaming motherboards! I would like to see Dual Opterons fit in a small ATX case!

"If youre paddling upstream in a canoe and a
wheel falls off, how many pancakes fit in a doghouse? None! Icecream doesn't have bones!!!"

"Battling Gimps and Dimbulbs HERE at THGC"

 

[Clob]

Maybe one day we will have quad channel memory controllers for the high demand of dualcores?

That would give me and excuse to have 2gb of ram!

"If youre paddling upstream in a canoe and a
wheel falls off, how many pancakes fit in a doghouse? None! Icecream doesn't have bones!!!"

"Battling Gimps and Dimbulbs HERE at THGC"
<P ID="edit"><FONT SIZE=-1><EM>Edited by Clob on 04/12/05 11:12 AM.</EM></FONT></P>

 

[P4Man]

> would like to see Dual Opterons fit in a small ATX case!

You already can:
<A HREF="http://www.iwill.com.tw/product_2.asp?p_id=36&sp=Y" target="_new">http://www.iwill.com.tw/product_2.asp?p_id=36&sp=Y</A>
Its not really bigger than your typicall SFF, but its a dual opteron workstation nonetheless. Imagine plugging DC opterons in there, and you'd have a 4 way powerhorse the size of shoe box !

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[Clob]

Your a Meanie!

"If youre paddling upstream in a canoe and a
wheel falls off, how many pancakes fit in a doghouse? None! Icecream doesn't have bones!!!"

"Battling Gimps and Dimbulbs HERE at THGC"

 

[P4Man]

>Maybe one day we will have quad channel memory controllers
>for the high demand of dualcores?

Maybe, but as it is, I think it would be too hard and too expensive to produce the motherboards. More likely is eventually seeing something serial/P2P like Rambus' Yellowstone '(XDR) or FB Dimm to provide the cpu's with more bandwith. There are limits to what you can realistically do in parallel at high speed, the way like our current DIMMs work.

= The views stated herein are my personal views, and not necessarily the views of my wife. =