Athlon 64 3200+ 90nm (S939) - The next P4 2.4 "C" ?
 

Combining affordable pricings, good overclockability, top level performance and the ability to keep things cool is what we enthusiasts pursue in our quests to build high performance computers. Does the new S939 3200+ 0.09µ have these attributes? Find out as we compare it to an expensive FX53, a cheaper S754 3400+ and a mid-range S939 3500+ CPU.

http://www.madshrimps.be/gotoartik.php?articID=230

When available in BE? :D

Very well articulated article;)

thanks :)
was by far the most demanding article done to date for me. The data collection was about exhausting lol. In the end though it was well worth it and I hope those who had been waiting for it to come out are happy. I know I was recieving alot of PM's regarding this article as I had wetted a few appetite with teaser benches etc....
Thanks Jmke :)!

hi, when u were ocking at 250mhz, the ram were 1:1 ?

excellent mem results :) i have the same 250mhz fsb setup and can only get a mem benchmar of 66xx

any tricks? :)

thanks

Hi there,

ram was at 1:1 in all tests except for the S754 platform where a divider was used. The PC3700 can clock up to 265Mhz without problem, check the review of that OCZ memory here: http://www.madshrimps.be/gotoartik.php?articID=229

yes memory was at 1:1
cas 2.5 3:3:8 I believe. Depending on your mhz will help/hurt your memory bandwidth. If your at say 2.2ghz and 250mhz you will lose about 200-300mb/sec.
Also make sure 1T is enabled or 2T in bios is Disabled which should force 1T then depending on your board etc...

what the OC performance gain between 1T <> 2T in your testing? I suppose you can clock higher with 2T ?

2T is horrible to performance. So much so I don't use it at all costs. That has been my experience anyway.
I would rather use on my 754 setup ram at 333 divider and 250HTT at final ram speed of DDR416/208mhz at tight timings and 1T vs same timings(if it was possible) at 1:1 250 and 2T.

seems this was a let down in comparison of the s754 system

would love to see s754 system with 250mhz 1:1 :)

look at the timings of the S754 vs S939 though... :)

yeah eva2000 I could have probably run 1 512mb stick at same timings as the 939 systems. I chose what I though was the best compromise. The 754 setups seem to gain more running at a lower overall ram speed and tight timings vs higher ram speed with looser timings. Either way, it was what I had to work with at the time

a small error

2nd page, superpi calc 1m @ 2.5ghz there is no fx there but 2* 3500+

will be fixed tommorow :)
don't have source files at this location.

Good article, just a nitpick but it may confuse quick/lazy readers.

There's a lot of people (myself included) who just go straight to benchmarks. I was amazed that the 3200 was performing better than the FX-53 in some tests and the 3500 in every test. I was just about ready to cancel my 3500 order and get a 3200! Then I actually read the article a noticed you were running everything at equal speeds. That's cool, but any chance you'll upload some comparison graphics for tests at factory speeds? They're useful for doing a price/performance comparison.

By the by, anyone know if AMD will be making a 90nm 3500+?

hello all. i just joined your forum for this article. i was wondering how easy it is to overclock the 3200+. i'm a total noob when it comes to overclocking. but i know what i'm going to buy for the system. what do you think about these specs.
asus a8v rev 2.0
3200+
ocz pc3200 plat rev 2.0 512mb(another 512 next month)
some case
either enemax or thermaltake 420maybe 480w psu.

i'm wondering if you can tell me if i would be able to hit 2.5ghz first attempt with those parts. also can you mention how to overclock step by step. if it's not to much trouble. atealst some links for a64 ocing.

too complex to go into right here I'm afraid. I was thinking of writing a comprehensive A64 oc'ing article. Maybe I'll pursue that a little more.

personally, i'd buy a dual channel kit instead of buying 1stick now and 1stick a month later. They'll probably both work in dual channel, but you might end up with one stick that clocks better then the other (although you can have this with dual channel kits aswell).

I'd also go for a different PSU, just because of prejudice :D tagan is very good and not expensive brand I believe they also produce the OCZ PSU's, not sure of this though).

I don't have any experience with A64 so here's a guide: http://www.dugu9tweaks.net/guides/a64oc/

should get you started: http://www.dugu9tweaks.net/guides/a64oc/index.html

edit: duh Kristos already posted this :p

the [M] version on step by step OCing A64 is a good idea.:)

fuk yoo all, my pci/agp busses are not locked on that a8v, so fuk yoo all :)

que?

yeah sure, blame it on the lack of locks...

pussy


just kiddin ;)


Is it with the A8V rev 2 that the locks work or was it another model for socket 754?

Got rid of the Lian-Li PC65, P4 2.4C system ....
Next rig will be 939 with PCI/PCIe with lock? I have the patience.:)

The overclocking results are good news if you've got DDR 500 but if you've only got DDR 400 that won't go much over that, what kind of results would you get running the RAM at 80% of FSB speed, ie ram at 400 (or slightly higher), FSB at 250 (or slightly higher). If I've got to ditch my RAM to upgrade and buy even more expensive RAM, it seems to kind of defeat the purpose of overclocking...

? not at all; the memory @ 250mzh is not running at CL2 2-2-5 1T, but at a slower CL2.5 3-3-7

the difference between 200 @ tights timings and 250 @ loose timings, it less then you think =)

Yes, by running a ram divider using the 333mhz setting and then scaling htt to 250 your getting just a bit above DDR400 at that speed, ie like the 754 setup. If you have tight timings on that DDR400 it will perform very very well. You would potentially lose a bit of performance, but within several percentage points, nothing I would consider horrible. The benefit over your current setup would be very noticable in benches.

The A64 since the mem controller is on cpu doesn't suffer a penelty persay to have ram running non 1:1. In fact, all ram is running asynch to cpu. Sure, higher ram speed will benefit cpu, but only if decent timings used. Read the OCZ article and see what those results showed. As timings had to be loosened, there was a performance loss until HTT could be scaled enough to recover. I run 333 divider in my 754 setup and it's a beatiful system. Given your situation I would not hesitate to run the 333 divider until better ram could be purchased.

I really appreciate the great information in your review of the 3200+ 90u. I have been planning a system rebuild for several weeks. Your article came out just as my existing system began to go crazy on me. I was able to get delivery of the Asus A8V deluxe board 2GB of DDR3200 and a Athlon 64 3500+ 90u CPU yesterday. I will spend the weekend in a rebuild.

Thanks again for the hard work on the review.

Can anybody say when the s939 cpu's will be available in Belgium (the lower clocked cpu's 3000+ and 3200+, the 3500+ and 3800+ are already available)