Prepare for San Diego 4000+, 3700+ and 3500+ (!)

[saiamne]

Is it just me or do those numbers seem a bit low compared with the OC's that have been released on the Venice core? Wasn't Venice hitting over 2.72 @1.4v stable on AIR?

However if they just hooked it up to Phase to start with maybe those numbers are achievable regardless the temps?

Also looks like they topped out at 3.1ghz on phase...wasn't Venice hitting past that on phase? Or is this just showing a difference in the benches that were run? Don't remember how high the Venice went stable while Priming...

[Faceless Rebel]

These results seem to be very preliminary, plus I don't know what got lost in the Google translation, but it seems that San Diego may not be quite as good a sheer clocker as Venice. Maybe, since these numbers are so very preliminary.

However the 1MB L2 cache really will make up for a lot of clocking deficiencies. For an aircooler person like myself will like San Diego more, since insane clocks are impossible with aircooling anyways. I'd rather have 3700+ with ~2.8ghz than 3500+ with 3ghz because the 2.8ghz will be cooler for me. But I'm using Zalman CNPS-7000Cu which is probably laughable to most of you guys...

[s7e9h3n]

Is it just me or do those numbers seem a bit low compared with the OC's that have been released on the Venice core? Wasn't Venice hitting over 2.72 @1.4v stable on AIR?

However if they just hooked it up to Phase to start with maybe those numbers are achievable regardless the temps?

Also looks like they topped out at 3.1ghz on phase...wasn't Venice hitting past that on phase? Or is this just showing a difference in the benches that were run? Don't remember how high the Venice went stable while Priming...

Depends on which Venice you're talkin about. My 3800+ easily ran Spi16m on the stock hsf at 2.8Ghz @ 1.4vcore. I only attempted Prime95 blend @ 3.2G on single phase and it failed after an hour and a half. It did, however, run Spi1M @ 3.3G. Of course Pc-Ice was able to take this cpu to another level in his cascade - I don't recall if he ran p95, but he was able to bench it @ 3.6+Ghz. Then there's my 3500+ Venice, which I never tested, but sent to PC-Ice to write his review. I don't think there was any stability testing done on that cpu, but it could do 3.1G @ 1.4Vcore on single phase. You'll find all the info here:
My screenies from the 3800+ Venice
Pc-Ice's results with the same 3800+
PC-Ice's results with the 3500+ Venice

[terrace215]

These results seem to be very preliminary, plus I don't know what got lost in the Google translation, but it seems that San Diego may not be quite as good a sheer clocker as Venice. Maybe, since these numbers are so very preliminary.

However the 1MB L2 cache really will make up for a lot of clocking deficiencies. For an aircooler person like myself will like San Diego more, since insane clocks are impossible with aircooling anyways. I'd rather have 3700+ with ~2.8ghz than 3500+ with 3ghz because the 2.8ghz will be cooler for me. But I'm using Zalman CNPS-7000Cu which is probably laughable to most of you guys...

Random variation, probably. San Diego and Venice have the exact same core, and exact same processing, so unless the problem arises from a critical path related to a larger L2 cache, they should have virtually identical o/c potential, overall. Of course, part to part, it will vary, as always.

[saiamne]

Depends on which Venice you're talkin about. My 3800+ easily ran Spi16m on the stock hsf at 2.8Ghz @ 1.4vcore. I only attempted Prime95 blend @ 3.2G on single phase and it failed after an hour and a half. It did, however, run Spi1M @ 3.3G. Of course Pc-Ice was able to take this cpu to another level in his cascade - I don't recall if he ran p95, but he was able to bench it @ 3.6+Ghz. Then there's my 3500+ Venice, which I never tested, but sent to PC-Ice to write his review. I don't think there was any stability testing done on that cpu, but it could do 3.1G @ 1.4Vcore on single phase. You'll find all the info here:
My screenies from the 3800+ Venice
Pc-Ice's results with the same 3800+
PC-Ice's results with the 3500+ Venice

Yeah...been following all the OC's on the venices...I just couldn't remember off the top of my head which benches ran stable at what speeds, probably has comparable OC potential to this CPU...just a little twitchy I guess since I've got the 3700+ ordered...been putting off this system build for a couple months since these weren't too far out and the CBBIDs were pissing everyone off...I've considered just buying a current 4000+ more than once to just be done with it, get built, and stop trying to play games on my damn XP 2000+

Researched every part to death and have everything picked out, just need this damn chip to ship...just want to be reassured I'll be able to hit my target envelope of 260ish x 11 w/ PC4000 VX and a SLI-DR on this CPU

[Faceless Rebel]

Researched every part to death and have everything picked out, just need this damn chip to ship

I have an entire computer sitting here except for a CPU. The EXACT INSTANT my 3700+ shows up at my door I'm going to rip it out of the box, cram it onto the mobo, strap the Zalman CNPS-7000Cu, and cram the mobo into the case. Seriously. The mobo already has the RAM installed, and the case already has the drives and fans installed. URGH! I NEED CHIP NOW!!!!!!

[[XC] leviathan18]

I have an entire computer sitting here except for a CPU. The EXACT INSTANT my 3700+ shows up at my door I'm going to rip it out of the box, cram it onto the mobo, strap the Zalman CNPS-7000Cu, and cram the mobo into the case. Seriously. The mobo already has the RAM installed, and the case already has the drives and fans installed. URGH! I NEED CHIP NOW!!!!!!

you missed the AS5 LOL.... just kidding... i think a lot of ppl is like you

[EMC2]

FYI... peeps need to quit trying to use AMD's model numbers as an indicator of the actual multi and default core speed, especially with the new Venice and San Diego cores

Just look at the previous core "models" and actual speeds/cache size:



You'll notice there are four different CPU's that have a 2.4GHz core speed, a 4000+ w/1MB, a 3800+ w/512KB, a 3700+ w/1MB, and a 3400+ w/512KB. There's a pattern there... 939s have higher ratings than 754s, 1MB cache have higher than 512KB cache, even though all 4 have the same core speeds.

[ozzimark]

there's another one.

3700+ newark. same specs as the s754 3700+, but with a revE core.. and it's a mobile.

[LoKi2k]

as you can see, our review is online. the second part will be online in a few days.

and there are some more news. next week we´ll get a venice and make a review, too. btw, our ocz pc4800 arrived. this review will be available next week, too.

btw. the fx is ordered

if here are germans... we´re searching people who can translate our reviews in english. when u are interested, please pm me or go in our forum

[saiamne]

Saw that, thanks for putting it up!

Seem like it scales about the same as the Venice CPUs we've already seen?

[reject]

that was on air? seems low for mach2

[Hombre]

How many chips based on San-Diego there are?

FX57, 3700+, 4000+, and 4200+? What about 3500+ and others?
Someone around here said that the 4000+ is the highest speed-binned version, but there a sticky with a preview of the 4200+... So I guess that the 4200+ is the highest.

[s7e9h3n]

How many chips based on San-Diego there are?

FX57, 3700+, 4000+, and 4200+? What about 3500+ and others?
Someone around here said that the 4000+ is the highest speed-binned version, but there a sticky with a preview of the 4200+... So I guess that the 4200+ is the highest.

Someone was wrong

[saiamne]

Right now there is only 3700+, 4000+, FX-55

4200+ / FX-57 are future chips

[Hombre]

Let me get this straight.

3700+, 4000+, 4200+ and FX57 are the only CPU's which are based on the 90nm San-Diego core? Because AFAIK lately it was revealed that there will be no SD/90nm FX55.

[onethreehill]

Let me get this straight.

3700+, 4000+, 4200+ and FX57 are the only CPU's which are based on the 90nm San-Diego core? Because AFAIK lately it was revealed that there will be no SD/90nm FX55.

FX-55 will be a SD/90nm

[saiamne]

you can order a San Diego/90nm FX-55 right now

can't really count the 4200+ and FX-57 since they won't be out until (projected) 3rd quarter of this year...4200+ will be similar to the current 4000+ in that it will be a multi locked chip based on the previous FX iteration (4000+ is a multi locked FX-53) so in the case of the 4200+ it will be a multi locked FX-55 when the FX-57's come out since there will only ever be one "FX" chip in production at any given time

I don't follow all the threads here (not enough time in a day) but as far as I know there's only one person still denying the existance of a 90nm FX-55...

This shows adafx55BNbox and adafx55daa5BN (oem), ETA 4/18...BN is San Diego...you can also order adafx55asbox off the same site which is a ClawHammer FX-55

[Hombre]

Cool, thanks for the explanation.

Think i'll get the 3700+ SD.

[sandybeach]

Todays the 15th all over the world!!

Lets see from today the ETA to "shipping" chips.

[Geforce4ti4200]

I dont think we should be jumping to conclusions that Diego is a worse clocker than Venice. One thing is the top Venice for now is only 2.4GHz while the top Diego is 2.6GHz with 2.8GHz fx57 comming. 1mb cache may run a little hotter, just try using better than stock cooling

[ingentingmendeg]

EXACTLY.

I don't know why people are getting so upset. Would you rather not know about new CPUs until you buy one yourself? We should be thankful for all this pre-release info, not mad at it because we can't get a CPU that isn't even released to the public yet.

i totally agree with you buddy. cheers for you

[Unrealcpu]

you can order a San Diego/90nm FX-55 right now

can't really count the 4200+ and FX-57 since they won't be out until (projected) 3rd quarter of this year...4200+ will be similar to the current 4000+ in that it will be a multi locked chip based on the previous FX iteration (4000+ is a multi locked FX-53) so in the case of the 4200+ it will be a multi locked FX-55 when the FX-57's come out since there will only ever be one "FX" chip in production at any given time

I don't follow all the threads here (not enough time in a day) but as far as I know there's only one person still denying the existance of a 90nm FX-55...

This shows adafx55BNbox and adafx55daa5BN (oem), ETA 4/18...BN is San Diego...you can also order adafx55asbox off the same site which is a ClawHammer FX-55

Ingram Micro an AMD major distributor has adafx55BNbox arriving the 18th apr

Anyone who denies its existence is plain stupid

[ben805]

I'll make it easier for you guys and did the online translation of that 3700 Sandy review, the wording is not dead on accurate but you should get the idea what they were trying to say



Page1:
We are happy to be able to present you the Overclockingbericht first worldwide of a new AMD Athlon 64 with San Diego kernel.

PART 1: since a couple of days, it gives complete computer of the firm Fujitsu Siemens that are equipped with the new AMD Athlon 64 processors with San Diego kernel. Presently this processor not yet is listed however and in no Hardwareshop available. Excluding OEM complete computer manufacturers, this new Highlight of AMD is withheld currently.

In the next days, 64 sockets 939 are introduced two new CPU generations of the AMD Athlon, the San Diego and the Venice. Both processors differ only and alone in the L2 cache. The Venice possesses comes 512 KB L2 caches, that San Diego with double quantities therefore, therefore 1024KB L2 cache. Both new processor generations offer the SSE well known by Intel 3 supports. In how far the SSE 3 noticeably makes itself, will see one in the later Benchmarks in the second part of the Reviews. The Venice and the SanDiego are made both in that 0.09µ procedure and have a standing type-V-Core of 1.4 volts.

The manufacture week of our processor reads 10 weeks 2005. that 3700 + San Diego offers a tact of 11x200MHz = 2200MHz and 1024KB L2 cache, the reason architecture corpulent in comparison with the prior sockets 939 processors identically.

* AMD Athlon 64 * ADA3700DAA5BN * CABHE 0510FPAW * 1225008C50445

The values on the CPUS must yet confessed occur are some of us, exact the same values, how in the at that time Flagschiff the FX 51. To Be Sure necessary of the FX 51 yet 1.55 volts and was produced in that 0.13µ procedure. First once couple Screenshots of CPU-Z.


Page2:

We dedicate now enough with the Vorgeplänkel the actual thing, the ascertainment of the maximal tact of the CPUS thanks to E. C. T to ourselves. 2 GT make.

Test system:

* Mainboard DFI-SLI D Biosversion 310P * ram 2 xs 512MB Gskill PC4800 GDR LA Dualpack * graphic card 2 xs Geforce 6800
Ultra Point of View * 80 gig SOWED/ DVD LW/ Floppy * make 2 GT, graphic cards cooled with Innovatek and looked Ice Radi/
Eheimpumpe * and naturally my new San Diego 3700).

The most important ahead: the tested San Diego has „no Sub-Zero-Bug.“ that is, that starting and the business with minus degrees has to 100% problem-free with the components named above and the 310 P Bios functioned.

Whether that in all new processors now just as simply functioned one will be able to see not until the future. This fact agreed me naturally equally very positively into the OC. In the Bios, the optimal Settings were loaded, that set ram attitudes and set the LDT/FSB Ratio to 3. All other attitudes were left on car and/or standard. First I determined then the maximal tact with 1,4V Vcore and 11er Multi with Clockgen.

On the basis of that, 2 GT cool units make was able to reach I now following stable results. 1.4 volts Vcore, real tact 2720 MHz Primestabel. This was nevertheless already an increase of 520MHz. After the AMD Rating for San Diego processors, that would come a 4500 + processor immediately. In the 1 M SuperPI test I could good 31,093s with 2722MHz reached.

I am the expanse already good 120MHz faster, than the previous top product of AMD, the FX55. at the pictures see we that according to CPUZ 1.36 volts was selected. After that I raised the V-Core in the Bios directly to 1.55 volts, what yielded selected 1,504 V according to CPUZ. 11 xs 263MHZ = 2897MHz were therefore possible. SuperPI 1M 29,219s.

Now I come also unfortunately already to the maximal tact of the processor: Primestabel were reached 274MHz once 11 = 3014 MHz in 1,712 volts. To adjust a higher V-Core no improvement of the maximal tact brought more. Unfortunately there is this point in each CPUS. The result really can see let itself. Because I had last week another new FX 55 to the test here, which well 750 - 800 costs€ and got out in 3150MHz Primestabel, the San Diego makes here with a price of 350€ a figure outstanding. SuperPI 32M yielded 25 minutes and 12.9 seconds. The SuperPI 1M could be tested even yet with 3105MHz, what yielded a best value of 27,297s. The timings were set in all carried out Benchmarks to CL 2,5/3/3/5 and the remainder on car. With something Feintuning and a couple of BH very good 5 would have to be possible around the 26,5s in the 1M and 24,30M in the 32 M.

Aquamark 01,03 and 05 was tested without OC of the graphic cards. In the driver, all was set to achievement.



Page3:

Result: all in everything is that 3700 + San Diego a very good CPU with very good Overclockingeigenschaften. 3000MHz Primestabel corresponds a 5000 + after AMD Rating. The coming Athlon FX 57 will be with 2800MHz getaktet, I was able to surpass this, in this first little test, around 200MHz. Further statements strong Benchmarks are published in the next days.

Anzumerken remains yet that the CPUS very stably ran together with the Gskill-ram, my test system. The San Diego reacts the expanse very well to V-Core-increase how one is accustomed that also in the FX 55. In the part 2 of the Reviews, I will represent a comparison of that 3500 + Winchester FX 55 and San Diego 3700 + in absolutely identical clock speeds. Therein are contain all well known 3D Benchmarks, a couple confessed Spielebenchmarks, different SuperPI´s and CPU Benchmarks. Reasons temporal end, came now the most important, the presentation and the Übertaktbarkeit of the processor, first. The remainder follows in the next days.

On parts 2 now already can please your you.

I hope had you a little joke at the reading of this little beforehand Reviews. (end of review)



To sum it up, I think what the reviewer trying to say is

"Clock for clock the 3700 Sandy is going to wipe the floor of the Winchester and Venice with the 1MB L2 cache, on top of that these sexy beasts runs cooler than the good ol' FX-53 and 55 due to the 0.09 die and 1.4v so higher clock is obtainable and 3.0Ghz air-cooled is not a dream anymore. Imagine prime95 and superpi 32M stable at around 3.0Ghz. The 3700 Sandy is looking mighty good and it will be the hot ticket of this year, they will be selling like hot cakes soon, and mark my words...expect the price jump in a month or two, I'll provide more details and screen-shots in the next few days so stay tune!!"

[ben805]

I spoke with another sales rep at Sparco today just to make sure that they'd canceled my Venice and have the 3700 Sandy on pre-ordered instead, the rep is very pleasant to deal with, he told me they WILL have the 3700 San Diego on Monday 4/18/05, as of now at 1:08pm EST they still have them available for pre-order...didn't ask how many left but he said they only going to have 60 units of the 3700 Sandy. I'm not going to wait forever so I had them ship me via 2-days, was told that I should receive it on next Wednesday or Thursday....rock on baby!!