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

[terrace215]

That last one should be a great overclocker, at a great price:

4000+ 939 1MB L2 2.4 GHz "BN" (Rev E) $637
3700+ 939 1MB L2 2.2 GHz "BN" (Rev E) $327
3500+ 939 1MB L2 2.0 GHz "BN" (Rev E) $270

http://epscontest.com/prices/rm_2005_04_11.htm

(This list somes from someone who has access to 2 AMD distributors, and shares the data weekly.)

[saiamne]

Interesting chart...however I just can't buy into AMD releasing 2 s939 3500+ CPU's for the SAME PRICE, one with 1mb L2 cache on the San Diego core...the other with 512k L2 cache on the Venice core...why even bother releasing a Venice 3500+ at all then?

I just don't know what to think about this whole new core mess anymore...except that it's a mess

[S0nic]

the san diego version is also 200mhz slower at stock

[koei]

Wonder when are these due for release.

[dimasdw]

is the OC performance beter than Venice ?

[terrace215]

Interesting chart...however I just can't buy into AMD releasing 2 s939 3500+ CPU's for the SAME PRICE, one with 1mb L2 cache on the San Diego core...the other with 512k L2 cache on the Venice core...why even bother releasing a Venice 3500+ at all then?

I just don't know what to think about this whole new core mess anymore...except that it's a mess

What?

You can choose: 512K L2 and 2.2GHz, or 1MB L2 and 2.0GHz. Both "on average" perform about the same at stock clocks.

Presumably you are also traumatized by Intel having a 5xx and a 6xx series that differ by L2 cache size?

These parts should start shipping this Friday, 4/15, per AMD's original distributor notice.

[terrace215]

I don't see why the o/c would be different from venice. Same core. Only the L2 cache is different.

But, what it gets you is a slower core / larger cache for a given performance rating ( = price). And slower clocked cores have more headroom to overclock.

For example, if you buy Venice 3500+, you are already at 2.2GHz.
San Diego 3500+ starts at 2.0GHz.

They'll probably, on average, o/c to the same speed. But you'll have more cache if you overclock the San Diego, thus higher performance.

[matt9669]

It does seem odd that AMD would be selling 3500+ and 3800+ chips in both 512K and 1MB considering that the Venice cores are cheaper to produce due to the smaller die size, and AMD certainly can't sell one 3500+ for more than the other . . .

Doesn't bother me particularly, but I don't see the business sense in it - maybe Venice yields at 2.2 and 2.4GHz are not good? (I'm hoping that's not the case . . .)

[TSFroggy]

Interesting - another mention of the 90nm FX-55. I wonder what the BN and Tray stand for?

[Faceless Rebel]

Seems to me that there is a sweet spot at the 3700+ San Diego (1MB L2, 2.2ghz). You get the 11x multi, and the bigger cache.

[CTKP]

Interesting - another mention of the 90nm FX-55. I wonder what the BN and Tray stand for?

BN is San Diego, BP is Venice. Tray usually means OEM, cpu only.

[TSFroggy]

BN is San Diego, BP is Venice. Tray usually means OEM, cpu only.

Thanks :) Now I wonder if this is truly what we're going to see, or if this was again speculation. Though as it's been said before, distributor information is rarely incorrect.

[rick_fx]

I just hope the 3700 san diego likes sub zero temps

[Stuperman]

another thing with the San Diego's with larger cache, you will get more performance gain per Mhz OC'd.

[RTI]

I believe the 3700+ San diego will be 2.2, 1Mb cache, while the 3500+ will also be at 2.2, but with 512 cache, similar to a venice. I recall seeing ADA3500DAA4BN. The 4 means 512 cache. I think the 3700+ will be ADA3700DAA5BN. The ADA3500DAA4BN probably failed with the 1Mb cache, similar to the 3500+ clawhammers that came out recently. Those seem to overclock well, so i wonder how these 3500 san diegos will do. Perhaps better than venice?

[matt9669]

The ADA3500DAA4BN probably failed with the 1Mb cache, similar to the 3500+ clawhammers that came out recently.

Ahh, this would make more sense

[terrace215]

It does seem odd that AMD would be selling 3500+ and 3800+ chips in both 512K and 1MB considering that the Venice cores are cheaper to produce due to the smaller die size, and AMD certainly can't sell one 3500+ for more than the other . . .

Doesn't bother me particularly, but I don't see the business sense in it - maybe Venice yields at 2.2 and 2.4GHz are not good? (I'm hoping that's not the case . . .)

It's the exact parallel of Intel having the 6xx and 5xx lines.

[terrace215]

I believe the 3700+ San diego will be 2.2, 1Mb cache, while the 3500+ will also be at 2.2, but with 512 cache, similar to a venice. I recall seeing ADA3500DAA4BN. The 4 means 512 cache. I think the 3700+ will be ADA3700DAA5BN. The ADA3500DAA4BN probably failed with the 1Mb cache, similar to the 3500+ clawhammers that came out recently. Those seem to overclock well, so i wonder how these 3500 san diegos will do. Perhaps better than venice?

Where did you see those OPNs?

I doubt this. AMD has no interest in having distributors distinguish between 512K L2 on-die, all functional, and 1MB L2 on-die, but 1/2-disabled.

That's another way of saying they should label that part "4BP" not "4BN".

[saiamne]

The ADA3500DAA4BN probably failed with the 1Mb cache, similar to the 3500+ clawhammers that came out recently. Those seem to overclock well, so i wonder how these 3500 san diegos will do. Perhaps better than venice?

Seems like this would be a reversal of previous AMD procedure since they apparantly stockpiled the extra Clawhammers last time then crippled them and remarketed them as 3500+ to clear stock...not sure if the Clawhammer 3500's were fails or just overstock of the Hammer cores, however since the Clawhammer 3500's seem to OC well it doesn't seem like they would be failed speed-bin'd cores...

I'd be a little leery of getting one of these if in fact they are San Diegos that failed to speed-bin at 2.4/2.6ghz...wish we knew more about what it would take for these cores to be considered "failed" (maybe some cache problems in the fab?) I don't think there's any chance that they have so many EXTRA San Diego's at this point that they need to dump them...If they were all good cores they'd just slow San Diego production, hold these, and ship them as the intended part as the market required instead of remarketing them under a lower product line...plus yields have always been low on their premium chips, so again, I can't see them having extras at this point...

Too much about Venice/San Diego isn't making sense, and it seems to get more and more muddled as we get closer to commercial availability...it's giving the logical side of my brain a headache

who knows, maybe the world's gone crazy and AMD is giving us overclockers a reward for being good little fanboys by putting a premium product on the market that only WE would recognize at an equivalent price...

edit - anyone got an Uncle on the Board of Directors over at AMD that could get us some GD answers already?

[robberbaron]

It does seem odd that AMD would be selling 3500+ and 3800+ chips in both 512K and 1MB considering that the Venice cores are cheaper to produce due to the smaller die size, and AMD certainly can't sell one 3500+ for more than the other . . .

Doesn't bother me particularly, but I don't see the business sense in it - maybe Venice yields at 2.2 and 2.4GHz are not good? (I'm hoping that's not the case . . .)

That doesnt make a lot of sense. How could they have San Diegos if the Venice yields are down? Venice is essentially a toned down San Diego with about 2/3 the transistors.

The Clawhammer 3400+ sold for more than the Newcastle one. Why? Because it had more transistors, thus more cost.

It's the exact parallel of Intel having the 6xx and 5xx lines.

This isn't anything like the 5xx vs. 6xx. San Diego won't replace Venice like 6xx replaced 5xx. It's really the exact parallel of the Clawhammer and Newcastle.

[terrace215]

Seems like this would be a reversal of previous AMD procedure since they apparantly stockpiled the extra Clawhammers last time then crippled them and remarketed them as 3500+ to clear stock...not sure if the Clawhammer 3500's were fails or just overstock of the Hammer cores, however since the Clawhammer 3500's seem to OC well it doesn't seem like they would be failed speed-bin'd cores...

I'd be a little leery of getting one of these if in fact they are San Diegos that failed to speed-bin at 2.4/2.6ghz...wish we knew more about what it would take for these cores to be considered "failed" (maybe some cache problems in the fab?) I don't think there's any chance that they have so many EXTRA San Diego's at this point that they need to dump them...If they were all good cores they'd just slow San Diego production, hold these, and ship them as the intended part as the market required instead of remarketing them under a lower product line...plus yields have always been low on their premium chips, so again, I can't see them having extras at this point...

Too much about Venice/San Diego isn't making sense, and it seems to get more and more muddled as we get closer to commercial availability...it's giving the logical side of my brain a headache

who knows, maybe the world's gone crazy and AMD is giving us overclockers a reward for being good little fanboys by putting a premium product on the market that only WE would recognize at an equivalent price...

Besides the speed bin distribution, there are marketing issues at play. I'm sure AMD can make tons of 2.4 and 2.6GHz San Diego parts, but not many people can afford them. So, many get clocked down.

[Geforce4ti4200]

I believe the 3700+ San diego will be 2.2, 1Mb cache, while the 3500+ will also be at 2.2, but with 512 cache, similar to a venice. I recall seeing ADA3500DAA4BN. The 4 means 512 cache. I think the 3700+ will be ADA3700DAA5BN. The ADA3500DAA4BN probably failed with the 1Mb cache, similar to the 3500+ clawhammers that came out recently. Those seem to overclock well, so i wonder how these 3500 san diegos will do. Perhaps better than venice?

Thats good speculation and makes more sense than amd undercutting the 3500+ when they make more with 4000+ so it could either be 512k cache or 1mb but speed binned cause it didnt make 4000+ clocks. I guess we will find out in time. I doubt if I can afford a retail 3700+ diego unless it magically costs around $300. The 3500+ diego looks ok if its below $300 and overclocks above 2.7GHz. We all speculate, after all it gives something to pounder

[ozzimark]

Where did you see those OPNs?

I doubt this. AMD has no interest in having distributors distinguish between 512K L2 on-die, all functional, and 1MB L2 on-die, but 1/2-disabled.

That's another way of saying they should label that part "4BP" not "4BN".

then why do the 3500+ clawhammers exist?

[RTI]

Where did you see those OPNs?

I doubt this. AMD has no interest in having distributors distinguish between 512K L2 on-die, all functional, and 1MB L2 on-die, but 1/2-disabled.

That's another way of saying they should label that part "4BP" not "4BN".

Try froogle search for ADA3500DAA4BN. I currently see 8 resellers with this processor. Don't you think its unlikely they all made the same mistake and wrote BN instead of BP?

[Ferry82]

AMD Athlon 64 3000+ S939 90nm Venice @1.8 Ghz 512k
AMD Athlon 64 3200+ S939 90nm Venice @2.0 Ghz 512k
AMD Athlon 64 3500+ S939 90nm Venice @2.2 Ghz 512k
AMD Athlon 64 3800+ S939 90nm Venice @2.4 Ghz 512k

AMD Athlon 64 3500+ S939 90nm San Diego @2.0 Ghz 1M
AMD Athlon 64 3700+ S939 90nm San Diego @2.2 Ghz 1M
AMD Athlon 64 4000+ S939 90nm San Diego @2.4 Ghz 1M
AMD Athlon 64 4200+ S939 90nm San Diego @2.6 Ghz 1M
AMD Athlon 64 FX55 S939 90nm San Diego @2.6 Ghz 1M?????
AMD Athlon 64 FX57 S939 90nm San Diego @2.8 Ghz 1M

So this are all the nieuw CPU's???????
Correct me if i am wrong!!!