Socket 939 Athlon 64 3400+

[c627627]

It was very late when I posted the stuff above and now that I think about it,
I don't know which was more ridiculous, my statement that 3400+ is a rebadged 3500+ (that they're the same chips,) or yours that newegg was practicing Photoshop art .

So now,
it's either a 2 or 2.1 GHz part. It can't be 2.1 because of multipliers and if it's a 2, then AMD has a 3400+ and a 3200+ 90nm at 2.2.

Could it be at 2 GHz + have 1 MB cache? Like a renamed FX?
No it can't! because it's a 4AZ, 4 means 512 KB cache.

So it's a 939 for sure + it has 512 Kb cache for sure + it's a 130 nm, right?

Since it's officially out, can AMD be asked about this?

 

[Shade00]

I would be tempted to agree, but the model number indicates only 512k cache.

 

[c627627]

Since it's officially out, can AMD be asked about this?

 

[NeoGeo]

Since it's officially out, can AMD be asked about this?

Isn't it officially out when AMD says it's out? So far no info from AMD's webpage or news archive suggests this chip exists.

 

[c627627]

AMD's web site is notorious for that. I mean, I remember them not updating the Duron page to include the 1.4, 1.6 and 1.8 Applebred Durons more than 6 months after their release! Did they even update it to this day?

AMD engineers are just fine, it's the marketing department that needs a really good flush imho.

 

[c627627]

Hm.

First off:

• it's a 130nm chip for sure.
• it has 512 KB cache for sure.
• it's a Socket 939 Newcastle for sure.
• it's not a Sempron for sure.
• newegg did not doctor the picture for sure.

So now:

1. It's a rebadged 3500+ at 2.2 GHz.
2. It's a 2.1 GHz chip (Can't be, 1/2 multipliers are not officially supported + they degrade performance of Athlon 64s.)
3. It's a 2.0 GHz chip, making it a 939 2.0 GHz 3400+ with 939 2.0 GHz 3200+ 90nm existing along side it.


Option 1 is the least ridiculous option (when compared to others).

 

[billy_cat]

I guess I'll join the speculation party.

Maybe AMD is repositioning the NC 3500 as 3400 and have the 90nm winchester 3500 be the only 3500?

 

[The Mad Duke]

This is really bizarre! Try following this link on NewEgg:
http://www.newegg.com/app/Viewprodu...talog=343&manufactory=1028&DEPA=1&srchFor=939
Now scroll down the page and you will see, right next to each other, a 3200+, a 3400+, and a 3500+ AMD CPU. Notice they are all described as Socket 939 2.2 Ghz 512 Kb L2 cache chips. Notice also there is a $100 spread in price! What on earth is AMD up to here ???
-The Mad Duke

 

[The Mad Duke]

Welcome to the Twilight Zone

Does NewEgg read this forum ??? The 3400+ has disappeared!
-The Mad Duke

 

[c627627]

Well:

I emailed newegg about it on Friday, it should be corrected by Monday.

but thanks to Internet Explorer > Tools > Internet Options > Settings... button (in the middle of the page) > Change the settings to "Never"

I got this Exhibit A:



Internet Explorer > Tools > Internet Options > Settings... button (in the middle of the page) > Change the settings back to "Automatically"

 

[c627627]

...and this Exhibit B:

 

[OC Detective]

The fact it has "disappeared" lends weight to my theory that Newegg should not have been selling them now (Assuming they have NOT been photoshopping). I suspect this is part of AMD's rationalisation program (stop laughing!) whereby the pr ratings are realigned after the socket 754 desktop A64's are sold through and no longer available.

 

[c627627]

newegg sells a huge amount of CPUs, they built the company on being reputable, I'd say most people at the forums here buy their CPUs there. The chips are real and the photos are real.

AMD is going to continue to make and sell Socket A chips, let alone Socket 754 chips so newegg selling these 130nm chips "too early" is highly improbable.

It may be that they weren't supposed to sell them at all but the chips are real and they're probably identical to 130nm 3500+s.

 

[OC Detective]

You kind of contradicted yourself - if they were not supposed to be selling them "at all" then why would they have shown them as in stock? That means that they received them from AMD however EITHER AMD may have stipulated an embargo on them until a later date OR Newegg realised that if they sold these at $293 then they were losing out on revenue to those identical 3500+'s being sold at $319 (I didnt say who decided they were selling them too early - could be Newegg or AMD).
Time will tell if I am right. If they have them in stock then when are they going to sell them - presumably later meaning they were not supposed to be selling them now which is precisely what I said!

 

[Vector]

If c627627's roadmap is accurate, the reason for this is pretty obvious. The roadmap shows the Winchester core as having a ~5% increased IPC over equivalently clocked S939 Newcastle cores (probably more efficient memory controller, ALU, or FPU performance...but I'd go with the mem controller). A ~5% increase for a 2.0GHZ processor would correspond to the PR rating of a ~2.1GHZ processor or a PR of 3600+ or so, which is very close to the 3700+ or 3800+, so they shouldn't exactly do that...and they also shouldn't have 2 equivalently clocked processors on the same platform (S939) and having the same cache specs (Newcastle vs Winchester) with one beating the other clock for clock and no differentiation between the two.

So the easiest thing for them to do is to slowly change all the Newcastle 3500+'s to 3400+'s so when the 2.2GHz Winchesters come in volume, there will be a clear distinction between the two that reflects the better performance ~100PR points. They'll probably change the current 3800+ to 3700+ when the 2.4GHz Winchesters come in volume.

Just my 2 cents

 

[c627627]

No, no, this mystery is finally solved. The word is:

The 3400+ Socket 939 really exists as a 2.2 GHz chip with 512KB cache but it has 800 MHz HyperTransport whereas 3500+ has the standard 1000 MHz.

I take it this is the first and only Socket 939 chip with 800 MHz hypertransport?

 

[Vector]

I doubt 200MHz of HT speed corresponds to 100 less PR points. In fact, most people wouldn't see a performance difference between 600 and 1000MHz HT.

I think someone did a comparison a while back and actually got better benchmark scores using a lower HT multiplier than the standard (but I think it was within 1%, so it falls within an exceptable margin of error to call the results the same).

Think I'll stick with my enhanced memory controller theory

 

[OC Detective]

No, no, this mystery is finally solved. The word is:

The 3400+ Socket 939 really exists as a 2.2 GHz chip with 512KB cache but it has 800 MHz HyperTransport while 3500+ has the standard 1000 MHz.

I take it this is the first and only Socket 939 chip with 800 MHz hypertransport?

Your source? (perhaps from Eastern Europe?)

 

[confuzzedintel]

surely they do exist i don't think new egg is into photo chop also dose any know why the 3000-3200 winchester have disappeared

 

[AEsnowboarding]

Maybe they sold out. I know mine shipped and should be here tomorrow