Winchester 90nm review + OC

[Ref]

In Czech language, but I will tell you the most important things.

www.svethardware.cz (http://www.svethardware.cz/)

Winchester 90nm review 3000+ (http://www.svethardware.cz/art_doc-26F4878E99D78D0BC1256F1B006A1928.html)

The 3000+ is working @ 1.8 GHz. Die size is 84 mm2 (NewCastle = 150mm2).
Their voltage is 1.4 V and TDP REMAINS at 89W. Cooling requirements remain the same. 90nm 3500+ replaces the old 3500+. They say that there won't be any Athlon 64 desktop chips at 90 nm and no chip faster than 3700+. Semprons and mobiles will be made for 754 at 90 nm.

Winchester in CPU-Z (http://www.svethardware.cz/art.jsp?doc=452DBECDED5116F7C1256F1B006FF49F)

They say that the transfer to 90 nm was done great by AMD. ;)
They were surprised that the Winchester is nicely cold (their own words :) ).

They tested that after 10 minutes of BurnK7 the temperature was 46 °C.
The processor liked the voltage of 1.2 V and did 50 minutes in Prime95 without error. They didn't test longer because of shortage of time. The processors temperature was 40°C. They used a normal aircooler with 4000 rpm and copper core. NewCastle did 61°C at full load and WinCheater 46°C.
Their piece of processor was cold. Let's hope that all of them are :toast: .
They can compare it only to Thourghbred 1700+ B0 DUT3C (by °C).

OVERCLOCKING
They say it was a bit problematic because of RAM they had. They ran it @ 2350 MHz with 1.525 V. It was a 500 MHz OC. Nearly a 30 % OC.
Temperature at full load was 57°C. Prime stable for an hour. They didn't test longer.

It was a bit slower than a 3000+ for 754, because of lower clocks. They would give it a rating of 2900+.

Mods, feel free to correct errors of my bad english :toast:

[Lithan]

Hopefully the 3200+ overclock better... or else a second revision comes out soon. 2350mhz is pretty sad. NC .13's were hitting as high as 2500mhz or more on stock cooling. If we don't start seeing at least 2600mhz overclocks on air soon, I'll probably wind up with a dfi754/3200+DTR setup.

[Ref]

Hopefully the 3200+ overclock better... or else a second revision comes out soon. 2350mhz is pretty sad. NC .13's were hitting as high as 2500mhz or more on stock cooling. If we don't start seeing at least 2600mhz overclocks on air soon, I'll probably wind up with a dfi754/3200+DTR setup.

We will see soon ... I would like to see some in hands of our Xtreme overclockers :D

[p4z1f1st]

man, y u all are such naively?

y u all think, 90nm-CPUs will bring an extreme oc-boost?

that are just the first revisions of 90nm by AMD....so, don't wonder, when they just do, what they are aproofed to do :rolleyes:

when they will oc high, it's good :)

@ ocforums.com has one a winnie at 2500MHz with 1.65V on air....i think, that's ok for the first revision (my waterblock is waiting for them :D)

[pik-ard v1.1]

they(the czech site) only hit 1.525v, not like what *we* wouldnt mind setting it at. ;)

also, they're starting point was 1.8ghz, not 2ghz like the 754 NC 3000+'s, so they're already at a 200mhz deficit. percentage wise, that's like a 2ghz chip hitting 2.6ghz. ;)

[Lithan]

man, y u all are such naively?

y u all think, 90nm-CPUs will bring an extreme oc-boost?

that are just the first revisions of 90nm by AMD....so, don't wonder, when they just do, what they are aproofed to do :rolleyes:

when they will oc high, it's good :)

@ ocforums.com has one a winnie at 2500MHz with 1.65V on air....i think, that's ok for the first revision (my waterblock is waiting for them :D)


Expecting them to overclock as well as .13's isn't naive. I didn't say anything about wanting exteme overclocks out of them. I said I wanted them to overclock almost as well as the chips I've been able to buy for six months.

[p4z1f1st]

how i told u: they are the first revision/stepping of 90nm-AMD-CPUs ;)

[pik-ard v1.1]

how i told u: they are the first revision/stepping of 90nm-AMD-CPUs ;)
which ofcourse, is generally not even close to the best... and if there's already been a 500mhz OC on both the 1.8ghz and 2ghz chips... outlook looks good to me. ;)

[BluD]

if they do start 200mhz lower, they are overclocking ok as far as i see. Does anyone know what speed the 3700+ will be? 2400mhz i assume.

[pik-ard v1.1]

it will most likely be a 3800+ at 2.4ghz, 512kb L2 cache. ;)

[khmerballa]

Heres a amdzone review on the 3000+ showing only a 220mhz overclock on stock voltage/cooling
http://www.amdzone.com/modules.php?op=modload&name=Sections&file=index&req=viewarticle&artid=63&page=1

[Ref]

New review ! Newer chip (week 35, czech had 31).

AMDZone review of Winchester (http://www.amdzone.com/modules.php?op=modload&name=Sections&file=index&req=viewarticle&artid=63&page=1)

:toast:

They OCed it to 2027 MHz @ 1.4 V, not sure if stable though. :confused:

[BluD]

ew that aint gud, might just need a few more batches to get the machines warmed up =)

[Lithan]

how i told u: they are the first revision/stepping of 90nm-AMD-CPUs ;)

But even the first revision should clock as well as the .13
Tbred A's clocked at least as high as Palimino's when they were released.

[Gogeta]

... outlook looks good to me. ;)

I'd have to agree. Don't forget their (Czech review) OC was limited by the RAM they were using.

[esoteradactyl]

i have a 3500+ 90nm on order. i will throw it under the LS sometime this weekend and post some results :D

[Gogeta]

i have a 3500+ 90nm on order. i will throw it under the LS sometime this weekend and post some results :D

Sweet :toast:

[saaya]

tdp doesnt mean anyhting, amd doesnt give their a64 cpus and opterons individual ratings. 89W is the rating all heatsinks and motherboards have to be able to resist, it doesn mean the cpu will run with 89W, it rather means future higher clocked 90nm cpus might reach 89W.

very nice :)

thx for the link :toast:
those numbers remind me of the winchester preview we had here on xs, the numbers sounds just like what the guys (boss50 etc) achieved back then. they reached 2.6ghz on air iirc with just a slight vcore bump.

[***Deimos***]

true about tdcp..

I find it suspicious that they didn't change the TDCP for the 90nm chips. Intel, right off the bat raised the bar from ~75W to 103W (if memory serves). I expected AMD to raise it too, to accomodate higher frequency chips... since they didn't I dont expect faster models to come anytime soon.

Regarding current OC outlook, and future - as someone else said... give time for machines to warm up. Although D0 is already new revision, meaning another one wont come right away. Either a new revision or just some tuning of the equipment will get it. But, "it" ofcourse certainly isn't P4 clocks... but considering how good AMD is against Intel, even 25% higher Mhz (2.4Ghz->3.0Ghz) would be fantastic. But realistically it doesn't matter. 90nm is all about much needed volume... and volume will bring a lot of smiles to many believers.

[Noktar]

Looks good.
Great Job by AMD !
Hope there would be Revisions like TBred B :D

@Lithan: They clocked as "good" as the Palolminos. But, look how they oced, at least 2Ghz was good for a TBred A.
The new A64 0.13 are good overclockers, the Palominos weren't !
So, wait and see.
Hope they get the better Rev. in a few weeks/month.

And see, the 0.09 AMD CHips are better than the 0.09 Chips from Intel (imho)

[Snowman89]

true about tdcp..
Regarding current OC outlook, and future - as someone else said... give time for machines to warm up. Although D0 is already new revision, meaning another one wont come right away. Either a new revision or just some tuning of the equipment will get it. But, "it" ofcourse certainly isn't P4 clocks... but considering how good AMD is against Intel, even 25% higher Mhz (2.4Ghz->3.0Ghz) would be fantastic. But realistically it doesn't matter. 90nm is all about much needed volume... and volume will bring a lot of smiles to many believers.
i tested my a64 3200+ NC and even at stock speeds i got 40s in superpi and around 250 in cpumark and i beatend a p4 2800@3400 1:1, and then when i ran 250x10 i ran pcmark02 and i got lower memory score because i run 4:3 but all the other test i kicked his ass

[spaceman]

Well, that 90n 3000+ is basically the same as my mobile 3000+ CH with CG core, 1800 MHZ. stock, OCs to 2400 MHz. I think once they run awhile, and fine tune the new process, we'll see better stuff. :)

[saaya]

the a64s have such a high termal rating because they get hot under 64bit! if you stress them in windows xp-64 they get quite a bit hotter than under 32bit as all parts of the cpu are running with a 64bit app.

even the 3700+ ch is probably at 80W in 32bit or even less, and the fx55 ,even though its rated at 100W, is probably around 90W in 32bit

[marius7]

Why is Winchester so bad in memory bandwidth, only 4.7 GB/s :confused: , and 3800+ 939 had 6.2 GB/s. I'm disappointed :( WTF ?:mad:
http://www.amdzone.com/pics/cpus/athlon64/3000+90nm/sandraint.gif
http://www.amdzone.com/pics/cpus/athlon64/3000+90nm/sandrafloat.gif

[Ref]

Why is Winchester so bad in memory bandwidth, only 4.7 GB/s :confused: , and 3800+ 939 had 6.2 GB/s. I'm disappointed :( WTF ?:mad:
http://www.amdzone.com/pics/cpus/athlon64/3000+90nm/sandraint.gif
http://www.amdzone.com/pics/cpus/athlon64/3000+90nm/sandrafloat.gif

Maybe they tested it with 2T, or some different timings. Who knows. There is NO reason for it to be lower ... except for lower clocks. :confused:

[p4z1f1st]

lower multi?

A64 @ DualChannel benefits from a high mutliplicator....