Posting anything I can to keep HK's post from the top :lol:

Edit: seriously though, I understand how enthralled we all are with c2d but there's been no discussion of how little the 65-nano's slower cache affects performance; when the 65-nanos debuted it was this huge scandal. I feel even better about my Brisbane now - not only is it basically at performance parity clock for clock, but it is a much better overclocker than the 90-nano chips.

Surely you should have waited for the new AMD architecture to come out seeing as Athlon 64 is quite old compared to C2D.

So AMD still kicks ass. Wins easily in benchmarks that matter ( like 3D rendering and folding ). , while Intel wins in benchmarks which make little to no difference.

ROFL, Damage is getting addicted to Supreme Commander!

Just remember, love and respect the Mavor.

Nice article - I appreciate all the work that had to have gone into it...

-Martin <><

Hmm, I picked up a 90nm A64 X2 4600+ for only $125 shipped from the Egg last week. The 65w version, too. Seems like that's a far better deal than the 4400+ at $170, and a better comparison to the Core2Duo 6300.

I assume it was excluded because this was a 65nm to 65nm comparison article, but those who need a fast, cheap, cool, energy efficient CPU should consider it.

I recognize that such a broad review can't cover all areas, and generally I found it a very interesting and informative review, but it failed to address overclockability. I've been a long-time AMD user and AMD fan (however, not a fanboy), but I recently built a C2D system, and one of the major factors for choosing C2D is the remarkable overclocking ability of this line of chips. I have a 6600 (2.4GHz default) running at 3.25GHz with almost no effort. This is by no means unusual, as many many other C2D owners report the same amazing overclocks.

So while it may be a daunting task, maybe in another review Scott can compare how likely and how well C2D vs. AMD chips overclock, for IMHO these characteristics certainly factor into the overall assessed value of a chip.

Hellokitty and Proesterchen in the same thread... it's like André the Giant vs Hulk Hogan in the 1988 WWF title fight for the belt. All we need is Shintai (Macho Man Randy Savage) and the fight is on.

Intel blue spandex v.s. AMD green spandex. Which fanboy will be declared the winner and take home title belt for *ULTIMATE TOOL*..........

..........stay tuned fans

A possibly interesting observation about the Oblivion benchmark: if you look at the two leftmost performance graphs captured from Task Manager (i will call them graph1 and graph2), you will notice a suspicious coincidence: when graph1 is high, graph2 is low. Overlaying them in Photoshop will convince you that the correlation is basically 100%.

I suspect that if you were to convert the graphs to two columns of numerical values in a spreadsheet and add the columns, each row would end up totalling to 100%. That is, Oblivion's multi-threading sucks, and it's completely random as to which CPU any given thread runs on but on average Oblivion is really only using one "full" CPU at a time. I have seen this in the past with other applications, but generally only with apps that were totally clueless about 'proper' multithreading techniques. The worst example was Adobe Premiere from a few years back. When I ran it on a true dual-CPU box, it used both CPUs at exactly 50% each, and was absolutely no faster than running it on a single CPU at the same clock speed. Nice.

This observation is borne out by comparing the performance charts for Oblivion with the performance charts for the Valve particle simulator. Oblivion shows no clear trend of the quad-core CPUs beating the dual-core CPUs; in fact the only correlation is clock speed (check it in Excel -- if you list the Intel CPUs in descending order of clock speed, you get _exactly_ the performance chart from Oblivion). The Valve code, which clearly utitlizes all cores fully, shows the quads blowing the doors off the duals.

Conclusion: using Oblivion as a benchmark for multicore CPUs is a waste of time unless you just want to show how little gain you will get from running a quad-core CPU on some apps.

Cheers,
D

My X2 3800+ feels so inadequate. :o

Running the same CPU (albeit slightly overclocked at 2.2GHz with a 1/1 CPU/RAM ratio), I have to disagree; every real-world gaming benchmark I've seen still leads me to believe that the GPU is far more critical for gaming performance. Hell, given the comparative lack of difference in gaming performance, I could stay on my current system for at least another year, only changing the graphics card to a midrange DX10 card from either ATI or nVidia. I can live with that pretty easily.

Thanks for the E6600 vs. 6000+ numbers!

AMD's current weaknesses manifest themselves most fully in its high-end models, like the Athlon 64 X2 6000+, which draws more power at peak than the Core 2 Extreme QX6700 yet is often outperformed by the less expensive Core 2 Duo E6600.

And this is definetely a much more reasonable conclusion on the 6000+'s merrits!

How do you change detail settings in the perf test in supcom ? I just ran it with the default settings and got 16217 or something like that. I would be interesting to see what I get at other detail settings.

Hmmmm.
Page doesn't load.
Check Slashdot; TR is at the top of the page.

Put 2 and 2 together.

A sleuth is I!

AMDs K8L had better be good... better than Penryn.

The latest Penryn information from Intel states that the server variants will be available in Q3 2007 and hints at the Penryn desktop CPUs both dual and quad to be available by the end of 2007. A very low-power variant of Penryn for notebooks has just been announced to join the 45nm lineup, so Intel must have considerable confidence in their 45nm process. If K8L turns out to have any significant advantages over Core2 at the same price-points, expect the relevant Penryn introductions to be accelerated.

I am sorry, this article is fundamentally flawed..a CPU comparison must be run with no other bottlenecks in the system, a single GeForce 7900 GTX 512MB running on inefficient nVidia Vista X64 drivers is limiting your results way before the CPU differences can come into effect.
You need to get a pair of 8800's in SLI, and run with a more mature driver set (ie under XP), in order to draw out the differences in CPU performance.
Compare your results with Legionware's GPU/CPU benchmarks on SupCom, and you'll see what I mean: http://www.legionhardware.com/document.php?id=620

*Slashdoting in progress*
feels like I'm back on my 33k modem :)

Nice summation! I'm going to sit tight with my A64 3700+ San Diego, clocked to 2.7GHz, until later this year when AMD's 65nm Barcelona variants ship. That's going to settle the "To Core or not to Core" question for me once and for all--as I have a feeling it will for most people.

I am fairly impressed, though, with just how well AMD's 90nm, previous-gen cpus stack up to Intel's 65nm, current-gen cpus. Frame-rate differences of 10% or under just aren't perceivable by the user, and such performance differences can only manifest in the form of benchmark scores, etc.

Just for perspective, it would have been interesting to add in an ATi card to the games mix just to see what sort of differences, if any, there'd be.

Nice article guys, dugg.

Geez man, it's a bloodbath. X2 carcasses lying on the ground everywhere.

FWIW, a significant sub-story here is that while there is a measurable performance gap between the 5000-65 and the 5000-90, the difference is very minimal in almost all cases - so minimal as to basically be negligible unless you spend a lot of time with the rare particular app that the 65-nanos take a hit on. Now this would hardly be something worth mentioning to the enthusiast since the enthusiast would be better served looking at a c2d rather than any of the Athlons tested here today - but then there's the x2-3600 for only $80 which seems, on average, to clock up just as high as you can get any of the other Athlons up to. In other words, the significance is that if you're so strapped for cash that you need to be looking at something below the c2d4300 (raises hand), then don't bother looking anywhere else besides the 65nano x2-3600.

Just a short little mentioning. Chipsets dont have much to say. But why not use a 680i chipset to kinda equal things out abit more. 975x is by all means an old fellow from 2005.

Pretty cool review, I wish you guys would have benchmarked the 3600+ just so we could get a feel for how much of a performance hit one takes by going with an uber cheap X2.

I wish you'd use oggenc/lancer instead of LAME MT. From their own homepage LAME MT gives about 32% speedup over the normal LAME, while from my experience, oggenc/lancer gives at least 80% (on two cores). Would love to see some lancer results on a quad.

You can't win one race out of 10 and call yourself champion Kitty.
On balance ,overall,taking everything into consideration....Conroe and Kenstfield are AMD's owners right now.....Read the article...continue to wait for K8L. :)

Awesome article. Good work Scott!

That was a good article....nice work guys.
If April's pricecuts materialize then an E6850 2.93GHz and quad core Q6600 trample the $250-350 bracket.
We are living in interesting tech times.

Well done. Good article. Pleased to read that one.
<quote>
I must admit, though, that I have an almost irrational fondness for the Core 2 Quad Q6600, probably because it's the most energy efficient processor in our Cinebench power test. The thing is by no means a great deal—two E6600s will set you back over $200 less than a single Q6600 ....
</quote>
Come Q3-4, my E6300 will become a Q6600. :D

I like seeing CFD results. If I were building myself a workstation, that would be the most valid benchmark. It's too bad you can't get a hold of an industry CFD program.

No 3600? :P Hopefully next time.