Impressive Athlon 64 benchmarks

[AAlcHemY]

nVENTIV Mach II: The Athlon 64 FX at 2.8 GHz

*Drewl*

 

[horvendile]

Somewhat in connection to the former post, I'd like to recommend the review over at Ace's (http://www.aceshardware.com/read.jsp?id=60000253). A very good and thorough read, IMO.

I'm very interested in what's going to happen when games start using 64-bit code. IIRC, both Valve and Epic have said that it's a comparatively easy do-it-once fix that gives significant performance benefits.

 

[CorwinB]

I'm very interested in what's going to happen when games start using 64-bit code. IIRC, both Valve and Epic have said that it's a comparatively easy do-it-once fix that gives significant performance benefits.

Well, Ace mentions in the 2.8GHz review that HyperThreading is another way to gain a huge performance boost in games. IMHO, future "high profile" games will ship with both HT and x86-64 optimizations, in addition to a generic DX9 code and a specific NV30 path.

BTW, this 2.8GHz review with a CPU cooled at -38°C (eek) makes me wonder about AMD's ability to ramp up clock speed on the Athlon64... Ace's sample was only able to gain 7% with air cooling, which is really, really low, and makes one wonder about the stability of the chip in hot environments...

 

[davefb]

aces review seems to show the 64's in the best light.. it's also the review with the most games and different apps reviewed. the normal 64 looks very-very good value.. well unless all you play is q3 .

-dave-

anand also has a test of recompiling the q3 source code... the 64 wins in that too,,, woot ! so i could justify buying one because it recompiles code faster !!!!!!

 

[horvendile]

Aaah... there's nothin' like a freshly compiled game of Q3!

 

[nonamer]

I am pretty unimpressed with the Athlon FX right now. It pretty much ties with the P4 3.2C (a bit faster in some, a bit slower in other benchmarks) performance and pricewise. The thing that really suprised me most about this launch however, is how good Intels P865/870 Northbridges seem to hold up with the FXs integrated memory controller in terms of latency (if Anands benchmarks are accurate). If this trend holds true, it remains to be seen whether the integrated solution will not be seen as a rather unfortunate solution two years from now, when its inflexibilities in regard to newer memory standards become apparent (i have no doubts that integrated memory controllers will be the way of the future, i just question this particualler implemention at this point with newer mem solutions like ddr2 on the immediate horizon).

We must not be reading the same benchmarks (if you're reading THG's benchmarks, look elsewhere, fast) but the FX easily beats the P43.2C in most applications, especially games. Only place where the P43.2C still stays ahead is in streaming apps which are the strength of the P4. The only real complaint is that it's a rebadged Opteron, which requires registered RAM. The 939 socket one (which doesn't require registered RAM) should be faster since non-registered RAM has somewhat lower latency.

An integrated memory controller has much lower latency then a non-intergrated mem controller. I think some sites have simply tossed in a latency test without really knowing what they're doing, which may be giving you the impression that the Northwood memory controller is almost as good. Here's a site you can trust more.

PS: DDR2 has worse latency than DDR, and will probably be pretty lousy until DDR2-667 at least, and won't really shine I don't think until DDR2-800. This is a situation where I think an intergrated memory controller for DDR1 only is a good thing.

 

[Prophet]

corwin thats ur problem rite there kt-266 every damned kt266 system i have seen has problems also the scsi thing effects p3-p4 via sets as well but most makers of p4 sets bios "fixed" the issue b4 they put there boards out

via has admited this issue but still hasnt adressed it

if you want to fix it u need the pci lanancy patch made by a user

via's problem is that there pi set cuts transfers off at 24 insted of 96 like most hence the soundblaster live,scsi,raid card issue

thats why i never use via anymore and avoid nvidia(cache issues) sis chipsets are non problmatic
i have used the sis 735,745,746 and none have given me any reasion to complain infact out of the 40+ systems i have put there chipsets into only 1 ever came back with a problem and it was a bad fdd controler

the boards i use are mainly ecs k7s5a or l7s7a2 i have used the pc chips 746 based board as well as the asrock(asus) sis boards there ok but the ECS: l7s7a2 shines and are well worth the 53$ they sell for on newegg

the l7s7a2 is a verry compact board with less then 5% performance difrance over the nforce2 in dual chanil mode(memory) as well as the fast that there ide controler is acctuly the fastest out there now

dont blame the cpu for the issues of the chipset

oh as to drivers and flashing the only drivers u need for this motherboard are the ones on the cd they work without a hich via is the one who requiers u to update ur chipset drivers constantly to patch over bugs
nvidia is even better in most cases :/

oh and so u know i have seen as menny via//intel issues by volume as amd the latancy patch fixes most of them

 

[Natoma]

nVENTIV Mach II: The Athlon 64 FX at 2.8 GHz

*Drewl*

Holy Jeebus!

 

[Star_Hunter]

I am pretty unimpressed with the Athlon FX right now. It pretty much ties with the P4 3.2C (a bit faster in some, a bit slower in other benchmarks) performance and pricewise. The thing that really suprised me most about this launch however, is how good Intels P865/870 Northbridges seem to hold up with the FXs integrated memory controller in terms of latency (if Anands benchmarks are accurate). If this trend holds true, it remains to be seen whether the integrated solution will not be seen as a rather unfortunate solution two years from now, when its inflexibilities in regard to newer memory standards become apparent (i have no doubts that integrated memory controllers will be the way of the future, i just question this particualler implemention at this point with newer mem solutions like ddr2 on the immediate horizon).

You need to keep in mind as when you go higher in clock on the Athlon 64 as does the Memoryt controller and because of that the latency goes down even more thats partly why it scales so well.

 

[Saem]

Um.... the latency from the controller and the CPU would scale linearly.