90 nm Athlon 64 je tu

Kot lahko preberete tu so 90 nm Athloni 64 končno pricurljali v štacune. Tu je bablefish prevod une strani, da se ne boste zezali z kitajščino:

AMD Athlon64 3500+ Winchester processor

Now in the market condition AMD tabletop processor all uses 0.13 micro metric system regulation, but will be able Q3-Q4 to change to 90 ? metric systems regulations according to the AMD Roadmap expression in 2004, after but many people worried changed to 90 ? rice to be able otherwise step of Intel ?? to turn Proscott to be able to send out the high fever! ? After the HKEPC editorial department continues the whole world to test 90 ? rice P4 for the first time, delivers global first 90 ? rice K8 once again the test.

On the chart that this is first uses 90 ? metric systems regulations Athlon64 3500+, under in the contrast chart 0.13 micron Athlon64 3500+ semblance nothing greatly distinguishes, all because Athlon 64 has the IHS protection, we cannot see on its Die Size the distinction, but we also have been able separately to come out from the processor writing. On the CPU first line of writing, ADA on behalf of AMD Desktop Athlon, 3,500 on behalf of the model, D on behalf of Socket 939, I on behalf of 1.4V, K represented the highest operating temperature is 65C, 4 on behalf of 512K L2 Cache, finally two English writing "BI" represents it already to enter Rev D, but the old edition finally two English were "AW", or used Rev. C.

What is Rev D! ?

Any is Rev D! ? The material which according to AMD official provides demonstrated that, it is first uses on 90 ? rice IBM the SOI system regulation core, the core code number is "Winchester". Winchester and not merely stops can appear in Athlon in 64, moreover can use in Sempron, Opteron, Athlon64 FX and Mobile Athlon on 64, AMD expressed first batch produces goods the Winchester processor already produced goods on September 15, including Opteron and Mobile Athlon64, but tabletop Athlon 64 FX and Athlon 64 will be able in October only then to be able to arrive step.Winchester transistor number and previous generation NewCastle equally for approximately 6,850 ten thousand transistors, but the core size reduces to 90 ? rice by 0.13 micro metric system regulation 144 square millimeters 84 square millimeters, the voltage aspect reduces to 1.4V by 0.13 micron 1.5V.

Winchester ?? the Socket 939 platforms further popularize

Will penetrate Rev D from the document 90 ? rice Socket 939 distributions situations, Winchester core Athlon 64 has been able to promote 3000+, 3200+ and 3500+, formerly Socket 939 Athlon64 will be belongs compares Gao Tuan, only some 3500+ above model, the selling price continuously inched higher, but 90 ? rice editions Socket 939 will be able to promote compares the low price 3000+, 3200+, will be able the command more people extension to throw the Socket 939 platforms, indirectly sped up the multi- Socket 939 popularizations. Although the Winchester core 3000+ and 3200+ promotes, but the AMD expression also can promote Socket 754 Athlon64 3000+ and 3200+, but Socket 754 certainly cannot change to 90 ? Mi Er only to hold for use 0.13 micron, but Sempron Socket 754 can have REV D, because Sempron must make the fair price, 90 ? metric systems regulations are helpful the lower the cost price. Now the AMD already quoted price, 90 ? rice Athlon64 Socket 939 3000+, 3200+ and 3500+ fixed price are 189, 227 and 346 Dollar levels.

Early some websites pointed out the new generation of 90 ? rice Athlon 64 will be able to join the SSE 3 instructions collections the support, but after we will test discover new generation of 90 ? rice Athlon 64 in a CPUZ Instructions fence, or will only support MMX, 3DNow! SSE, SSE2 and x86-64, but has not supported SSE3. Our this Athlon 64 Winchester core 3500+ and the old edition same time arteries are 2.2Ghz, in constructs 512K L2 Cache, in a Rev item obviously may see it is D0 Stepping.

Many foreign website formerly formerly reported that, 64 except supports SSE3 besides new generation of 90 ? rice edition Athlon, has imposed the memory body controller potency, to this matter us inquired the excessively many main engines board merchant all expressed the Rev D 90 ? rice originally certainly were not the final editions, but is changes in the same old way the original 0.13 micron 90 ? rice, therefore designs on simply does not have the significant distinction. But Rev D also only can maintain at 04 year Q4, 05 year Q1, Q2, because final edition 90 ? rice Athlon 64 will be able the Q2 tail to promote Rev E, it will be is the new core again which 90 ? Mi Er will design, this will be the genuine lead, it will be able officially to join the SSE 3 instructions collections and the optimized memory body controller, but we formerly saw Dual the Core 90 ? rice projects sample also was Rev E, therefore the author estimated Dual Core Athlon 64 also not necessarily will be able to promote in this half year.

Also remembered when same year Intel promoted 90 ? rice Prescott, the major part main engine board all could the direct support, after its one be diverts 90 ? rice the Prescott core power loss sharp increase, made many main engines board the power supply ?? extremely to shoulder, its two was diverts the Prescott nucleus with to calculate the power supply ?? was allowed to support, also needed to renew BIOS to be able normally to start. That 90 ? rice Athlon 64! ? According to the AMD project personnel indicated that, AMD must face after 0.13 micron diverts 90 ? rice, appears leakage situation, but because AMD obtains IBM SOI Low the K technology, makes the question too to be by no means serious. BIOS aspect, because new generation of Winchester CPUID changed, therefore the main engine board possibly does not recognize the processor true body, but or may normally operate all does not have the question, greatly may feel relieved with the family.

But new generation of REV D 90 ? rice Winchester power losses or enhanced, formerly Athlon 64 3800+ TDP highest only had 89W, but new generation of REV D 90 ? rice Winchester Athlon 64 3500+ also could maintain 89W TDP, but above 3500+ could promote to highest 104W TDP, therefore AMD only then could not promote in the initial period 90 ? rice 3500+ above the model, because certainly was not all K8 main engine board all can support above 89W the processor.

Saw Intel changes over to 90 ? rice after 0.13 micron the thermal energy to compel the person, makes many people also is AMD worried, Winchester can otherwise become Prescott ??! ? Therefore we specially contrast 0.13 micron with 90 ? rice tests ? Wichester 3500+ VS NewCastle 3500+.

Whether tests us to be able to use Gigabyte the K8NSNXP-939 main engine board and Gigabyte 3D Rocket Cooler, transfers to the rotational speed lowest 2000rpm, the temperature detects uses CoolerMaster AeroGate 3, in the use CPU Burn4 for K8 enters the series the complete load ?

Winchester 3500+
New Castle 3500+
Idel
41.6C
37.2C
3DMark03 @20Mins
62.2C
57.8C
CPUBurn 4 @ 20mins
67.4C
62.7C

May see the Winchester core obviously compared to the NewCastle old core hot 1., but differs certainly is not very big, probably only some 3-4 distinction, looked like changes over to 90 ? rice although obviously can make the K8 processor temperature rise, but actually has not liked Intel Prescott to be fearful, believed is when Athlon 64 itself the arteries are not certainly high, after therefore changes over to 90 ? rice certainly not to appear the too big leakage like Intel.

Tests the platform ?

AMD Athlon 64 3500+ Socket 939 (Winchester)
AMD Athlon 64 3500+ Socket 939 (Newcastle)
Gigabyte GA-K8SNXP-939
Magic-Pro Geforce 5,600 128MB 128Bit
Magic-Pro DDR 400 CL 2.5-3-3-7 512MB x 2
Gigabyte 3D Rocket Pro Cooler
2themax 3XIO 420W PowerSupply
Seagate ATA100 80G 7200rpm 8MB


Winchester 3500+
NewCastle 3500+
PC Mark 04
Total
4,173
4,195
CPU
4,168
4,184
Memory
4,672
4,633
Graphic
2,153
2,145
3DMark 2,001 SE
1,024 x 768 @ Default
10,106
10,121
3DMark 03 Built 340
Total
2,057 (VERJETNO NAPAKA)
2,740
CPU Test Total
776
778
CPU Test 1
94.8fps
95.0fps
CPU Test 2
12.5fps
12.5fps
Aquamark 3
GFX
2,312
2,306
CPU
8,346
8,243
Super PI
1C$m Test
40s
40s
4M Test
3m18s
3m18s
Sandra 2,004 SP1
ALU
9,582
9,583
FPU
3,521
3,486
SSE2
4,562
4,541
Integer
21,065
21,049
Floating
22,963
22,630

AMD smoothly transits 90 ? rice -- soft landings

Basic Winchester 3500+ design and NewCastle 3500+ approximately same, only is in the system regulation has distinguishes, therefore tests the result reflected two potency are same are completely do not make outside the people's desires, therefore AMD also does not have the meaning separately comes out on the packing 90 ? rice and 0.13 ? rice, therefore AMD may say is the soft landing. Moreover the author also felt AMD in market slightly intelligent, knew can surpass 89W in 3500+ above TDP, therefore retains Athlon which 0.13 ? Mi Tsokao the step product if will be able to promote 64 4000+ and Athlon 64 FX 55.Although 90 ? rice REV D certainly does not have in technical the progress, but makes AMD further to reduce in the manufacture cost, can deliver more cores under the identical crystal circle, may have AMD the bigger competitive ability and the match battle, looked like match Intel really must thought the method prevents AMD ??! ?

Clear sky

Verjetno bo podobno kot je bilo ob splavitvi Tbredov B, ko so se najprej pojavili v Aziji pol pa dost hitro še drugje po svetu. Glede na tole so prispeli tudi že v ZDA, a za zdaj samo preko spletne strani Monarch Computers (če ni to kakšna nategovalna akcija) in tako sedaj folk čaka, da jih dobijo po pošti. Dostavili naj bi jih 28.9.04, torej prihajajoči torek -- poslani pa naj bi že bili . Takrat se lahko nadejamo poplave informacij, za zdaj pa je vsaj na tem forumi edini z 90 nm čipom en Čeh ali Poljak Petr, ki dela za eno revijo in je čip dobil na test. To verjetno pomeni, da jih imajo že tudi številne spletne strani in sedaj že laufajo benchmarke, ki se bojo upam, da kmalu pojavili tudi na spletu.

Ampak, pri vsem skupaj je najbolj žalosto, da smo verjetno čakali zaman. Če je tist kitajski test legitimen, pol je 90 nm tudi na SOI-ju bolj sranje. Ni sicer tako hudo kot pri Preshott-ih a vseeno naj bi imel 90 nm čip povprečno 5 C višjo temp kot 130 nm, pa še navijajo se bolj za en kurc. Tam 2300 -2400 MHz s povišano voltažo.

Dobra (?) novica, ki izvira od tukaj je, da bojo baje zelo kmalu (nov. 2004) prišli tudi 90 nm Semproni. Po sedaj videnem bi osebno raje imel 130 nm jedro, a ker AMD očitno ni ravno najbolj navdušen nad svojim 130 nm procesom ( 50 - 100% preveliki čipi) bo skoraj celo leto prezgodaj na 90 nm prešaltal tudi Semprone (samo S754). Komaj čakam, da rubam tole into Viditor's face (en tip na Anadtech-u, ki oh in sploh hvali (po komentarji) kako ge AMD-ju ful dobro in vse po načrtih točno tako kot so si zamislili -- v AMD delnice ima vloženo celo svojo penzijo).

Skratka zdaj se začenja novo obdobje za AMD. AMD je prišel do točke ko lahko dovolj poceni in v zadostnih količinah dela K8-ke in upajmo, da to končno pomeni, da bojo cene padle na neke normalne vrednosti, vsaj za A64 2800+ tudi pri nas.

Škoda, khm, kaj pa vem. Tukaj bi se pogovarjali o navijanju 90 nm K8 čipov, medtem ko je tam splošna tema.

To misliš default ali pol pri navijanju? Baje je default 1.40 V, kar je malo veliko za 90 nm, pri navijanju pa okoli 1,55V za cca. 2400 MHz.

Hm.... mogoče pa bo držalo...

Govorilo se je o 1.2V, nekaj takega.
Nekdo, ki ima Winchesterja je rekel, da 3500+ deluje na 1.2V stabilno Prime95.

Fuck off, js vedno usekam tako plato da ne morm undervoltat. Mami, js ne več hitr, js tih komp bi mel.....

Dr_M: Ni bil point pokazat konkretno število, saj bi za to moral pogledati tabele in zadevo preučiti. ;))

Saj ne omenjam mobilnih?

Poleg tega, bi bilo bolje zakon postaviti tako;
"Na vsakih 18 mesecev se napetost na jedru procesorja zmanjša za n %." Bolj uporabno ;)

Ta kitajski test je po mojem mnenju "izdelan"!

Predvsem me moti voltaža 1.472V. Tudi če je procesor pravi 90nm 3500+ je to verjetno glavni razlog za 5degC višjo temperaturo.
Moti me še "preveč gladek" in neodseven 90nm heatspreader v primerjavi s 130nm, malce čudne številke v testih ipd.

Bom še jaz dal en test gor za primerjavo
Bom primerjal FX-55 in P4 EE 3.77GHz

X74

http://setiweb.ssl.berkeley.edu/team_display.php?teamid=30319

to sploh ni relevantno.

itak je pri intelu ena osnovna napetost potem pa +-0.025V, razen v izjemnih primerih.

Test je čuden!

V Cpu-Z piše voltaža 1.472V, kar pomeni da je procesor overvoltan, ker po defaultu naj bi 90nm laufal pri voltaži 1.4V. To je razvidno iz OPN kode ADA3500DIK4BI (D=SOCKET 939, I=1.4V, K=case temp, 4=512kB L2, BI=revision) Če bi bil 1.5V bi tam pisalo E.

Heatspreader je pomoje fejkan v photoshopu.Očitno ni mogel drugače potvoriti kode in je zablural strukturo. Prava struktura heatspreaderja 90nm socket 939 je vidna Tule!

X74

http://setiweb.ssl.berkeley.edu/team_display.php?teamid=30319

V Petrovem testu so popolnoma druge cifre!

po 10min 100% burn testa:
A64 3500+, 130nm CG Newcastle (2.2Ghz) = 60 stopinj C
A64 3000+, 90nm s939 D0, navit na 2,3Ghz = 56 stopinj C
A64 FX53, 130nm CG = 56 stopinj C
Pentium 4, 130nm, 3GHz Northwood = 55 Stopinj C
A64 3000+, 90nm s939 D0, (1.8GHz) = 46 stopinj C

X74

Hm tole je res zanimivo!
kaj to pomeni? ali je 3000+ navit na 2.4GHz manj požrešen, kot 3500+ navit na 2.4GHz?

Saj bo Anand naredil teste, upam da takšne, da so popolnoma primerljivi, ker performanse A64 3400+ približno že poznam, zanima me samo test A64 s939 512KB L2 2GHz 90nm PROTI A64 s939 512KB L2 2GHz 130nm. Seveda na popolnoma enakih sistemih, zraven pa še poraba procesorja.

Tukaj je en tipo s S939 3500+ 90 nm dosegel 3,0 GHz z Vapochill LS-om.

Še vedno ni poplave informacij. Ful čudno. Mogoče zato AMD ni preveč publiciral izida 90 nm, ker jih še ne morejo zanesljivo dobavljat. V ZDA jih je kao zmanjkal in to za več kot 2 tedna pravijo na newegg-u.

AT se še vedno obira. Teden je že naokoli od obljube, članka pa še vedno ni. Samo da ne bo tako kot pri Sempronu 3100+, k jim je trajal skor 2 meseca da so vsaj omenili kako se navija, čeprav so ob izidu 3100+ obljubili še članek o navijanju.

Kar nekaj skratka.