CPU/Mobo Opteron 1xx vs Desktop parts [a inside look]

[KiD0M4N]

and btw, thanks for the rep Switch..... same to you zhop. appreciate it.....hyeah:

EDIT: Btw, you might wanna see this:

http://www.techenclave.com/forums/374502-post51.html and,

http://www.techenclave.com/forums/374481-post49.html

 

[Darklord]

Opty 1xx series are "NOT" special in any form. They are just left over San Diego cores which AMD wanted to get rid off. So they branded it differently and sold it off.

Huh ???:huh: I thot it was the other way round,thanks for clearing my wrong concepts.:S

The A64 cores run cooler at the same speed compared to a Opteron because of less cache (translates less TDP.)

Not necessarily my dear.The Opteron 144 consumes 67W,so does the 3000+

So according to your logic,Opty 144 shud consume more power and hence more heat,the situation is exactly opposite here karan,how come ? :huh:

It could be a bad batch of FX-57 Sandiegos which actually didnt meet the TDP requirement and thus were branded and sold off as different lower frequency chips.

Karan,FX-57 has 1MB L2 Cache whereas Venice's had 512Kb Cache.The die size is different for god sake.:S
The die size isnt same for all.FX-57 failures = 3700/3800 is ok but not 3000/3200/3500/3800 man.:hap5:

See it didnt cost AMD anything because a FX-57 wafer takes almost as much to manufacture as a 3000+ (its all dependent on the die size.)

Huh ????? :huh:
FX-57 and 3000+ cost the same for producing ????
Man....i was so wrong till today with my assumptions.

You can urself check out that the TDP of all three processors is the same. So anish, your argument of 146 being a "better" silicon, cooler is totally unresearched

How ? well th 146 is having same TDP @ 2.0GHz as the 144 @ 1.8 GHz
Doesnt that translate to 146 having a superior,efficient core ?:huh:

Hell, 285 are cooled by 1U hsf and they do just fine. (2 cores, 2.6 Ghz each, 2 MB L2 Cache)

Well i partially agree but i guess you havent seen a rack mount chassis for this Dual CPU setup.I know they have comparitively,smaller Heatsinks,although full copper BUT i think you havent seen the fans that blow over these heatsinks.Check that out.Now if this setup is built in a normal tower chassis,then the cooler,even though smaller,is full copper,extremely high quality.

The extra cache is not something you will sacrifise 400 Mhz of CPU speed upon.

Extra 512Kb cache translates to around 200 Mhz speed bump in the case of A64 or Opteron . So while overclocking...the chip with higher cache would always have advantage at same clock speeds ,isnt it karan ?

A 11x multiplier helps a lot in that.

Not always
I get better results when i drop the multiplier to 9x,same thing happened with my Venice.dropping to 8x gave slightly better results.

 

[Shripad]

The speed binning is a confusing case. We will never know how AMD does it. But i assure you the manufacturing cost of 1mb cache chips and 512k chips is different.

The extra cache means more real estate on the chip core and that costs more.

Specially L2 cache specially takes up huge real estate on the chips.

The ocability depends on lots of thing, majorly if the chip is manuffactured using strained SOI or not. At the moment AMD is using both normal SOI and strained SOI manuacturing to fulfill demands and thats why there is this hit and miss luck with the chips.

Also the chips that come from centre of the weafer usually tend to be better cores, they can be clocked higher.

And karan about TDP, dont forget that these CPUs clocked at different speeds get their speeds from multipliers. When you overclok you are increasing FSB speed, not by changing multipliers. Increasing FSB increases the speed at which the processor runs, your cache speed and basically everyting is pushed from its stock speeds. That might well be the reason why it gives out more heat when cpu is overclocked even at the same vcore.

 

[KiD0M4N]

Ok sarcastic comments bashing time

How ? well th 146 is having same TDP @ 2.0GHz as the 144 @ 1.8 GHz

Doesnt that translate to 146 having a superior,efficient core ?

I am reprimanding him for giving general views regarding 1xx and 3500+ not between 144, 146. Even a child would be able to tell that 146 has a superior core compared to 144 because the 146 clocks higher than 144 at same TDP.

Take your argument forward, the 3500+ runs @ 400 Mhz more speed with same TDP. Wait theres more, a Opetron 165 has a TDP of 110W @ 1.8 Ghz..... where as a lowly (inefficient as per ur argument) 4200+ has a TDP of 89W..... that 2 cores, with 400 Mhz extra each over a Opteron 165. Care explain it?

Not necessarily my dear.The Opteron 144 consumes 67W,so does the 3000+

So according to your logic,Opty 144 shud consume more power and hence more heat,the situation is exactly opposite here karan,how come ?

Hmm, have you heard of a tool called TCaseMax Darky? Each E rev core has different exact specifics of the TDP and the max case temparature. That tools helps in finding that out. Check between a Opteron 144 and a 3000+. A 3000+ E6 will have a TDP of around 50W - 55W (on an average) whereas a Opteron 144 will have it around 55W-60W (again average.) Again, I have a point to counter your possible counter arguments. Many times AMD passes off higher end cores as lower parts just to fill stocks. Supply and demand.

Karan,FX-57 has 1MB L2 Cache whereas Venice's had 512Kb Cache.The die size is different for god sake.

The die size isnt same for all.FX-57 failures = 3700/3800 is ok but not 3000/3200/3500/3800 man.

Hmm, have you heard of CCBWE 06xx XXXX 3500 single cores Darky? Those are DUAL core Manchesters being passed along as lowly 3500+ Venice? Hows that happening? Surely, now we are talking difference between the die size of 3500 Single core venice vs Dual core manchesters. Well, AMD is comparitively small in its manufacturing capability area and thus it has to resort to these antics to get by supply demands. Similarly, my argument (which I said was my own theory) of bad higher end chips being passed off as lower end parts because they didnt meet TDP requirements.

Karan,FX-57 has 1MB L2 Cache whereas Venice's had 512Kb Cache.The die size is different for god sake.

The die size isnt same for all.FX-57 failures = 3700/3800 is ok but not 3000/3200/3500/3800 man.

U missed out the word "almost" FX-57 has a die size of 115mm2 and a 3000+ venice has a die size of 84mm2. If it takes $30 (approx figure) to manufacture and produce a FX-57 SanDiego core, how much do u reckon will it cost them to manufacture a 84mm2 die? Compare that with having to throw away that core not meeting some spec (speed or TDP) or rather selling it off as which part it best fits in and/or demand of market.

Well i partially agree but i guess you havent seen a rack mount chassis for this Dual CPU setup.I know they have comparitively,smaller Heatsinks,although full copper BUT i think you havent seen the fans that blow over these heatsinks.Check that out.Now if this setup is built in a normal tower chassis,then the cooler,even though smaller,is full copper,extremely high quality.

Hmm, a 1U rack is as small as it gets. Imagine having a 80mm DELTA fan sitting atop a FULL COPPER hsf (measuring all of 8 mm height, again approx) trying to cool a overclocked 170 @ 2600 Mhz. The thing will rise to higher temps within seconds. Even a XP-90C is not sufficient to cool OCed dual core monsters just because its size is too small and it is unable to radiate enough heat quickly (>>> 110W) And you dont have to take my word for it, Darky: Opteron 180 vs Opteron 280 Now does that convice you at all? Or you still think I am speaking out of my ass.

Extra 512Kb cache translates to around 200 Mhz speed bump in the case of A64 or Opteron . So while overclocking...the chip with higher cache would always have advantage at same clock speeds ,isnt it karan ?

Or the AMD marketting people would have you believe.

I had a 3000+ Venice before my Opteron 165 (the irony ) It had a measly 512KB cache. It gave me SPi1MB runs of 33.20 seconds at best (oced to 2475 Mhz). Now I have Opteron 165 oced to 2.675 Mhz with even faster ram. So thats a 200Mhz of RAM CPU power and another 200Mhz for extra 512KB cache (according to ur point) So thats a cool 400Mhz extra over the 3000+ venice. Why am I still getting best SP1MB times of 31.2 seconds? I should easily break 30s isnt it Darky? 512KB extra cache shows "selective" performance improvements because the A64 architechture is really really independent of "the need for higher memory bandwidth." I repeat, for A64, Mhz is the king....

Not always

I get better results when i drop the multiplier to 9x,same thing happened with my Venice.dropping to 8x gave slightly better results.

Hmm I am taking about the flexibility that offers you, not the efficiency. Do you know current world records in 3DMark05 are held by FX-60 running multiplies like 17 and 18 (to combat the low HTT speeds rewarded with sub 0 cooling, bad coldbugs)

But you are totally misinterpreting the point. The improved performance is "NOT" due to "lower" multiplier (which lol, is still a possibility with the 3500+) but due to the higher HTT speed you end up using. AFAIK, higher HTT yeld points in SP1MB as well as various 3Dmarks (most prominently 03.)

With a 144 you would be stuck with a max multi of 9x whereas with a 3500+ you would have choices upto 11x. And that you eliminate the need to get a high priced MOBO because a lower multiplier demands a MOBO which is able to do higher HTTs. Isnt it?

I hope I have cleared some of your "comments".... Atleast I have tried to (note, only 3 smileys used )

 

[KiD0M4N]

The speed binning is a confusing case. We will never know how AMD does it. But i assure you the manufacturing cost of 1mb cache chips and 512k chips is different.
The extra cache means more real estate on the chip core and that costs more.
Specially L2 cache specially takes up huge real estate on the chips.
The ocability depends on lots of thing, majorly if the chip is manuffactured using strained SOI or not. At the moment AMD is using both normal SOI and strained SOI manuacturing to fulfill demands and thats why there is this hit and miss luck with the chips.
Also the chips that come from centre of the weafer usually tend to be better cores, they can be clocked higher.

And karan about TDP, dont forget that these CPUs clocked at different speeds get their speeds from multipliers. When you overclok you are increasing FSB speed, not by changing multipliers. Increasing FSB increases the speed at which the processor runs, your cache speed and basically everyting is pushed from its stock speeds. That might well be the reason why it gives out more heat when cpu is overclocked even at the same vcore.

Exhacctly, thats what I am saying Funky. But it doesnt mean the CPU (which comprises of the Execution unit, Inst fetch/decode, various caches and register banks, etc, etc) as a whole is not giving out higher heat.

 

[Shripad]

As far as my knowledge goes AMD wont waste their time binning the processors by TDP. They test perticular core under lab conditions for TDP and thats it. Thats done. All binning is done is speed binning and that too does not mean you wont get some cherry silicon in mainstream processors or opterons.

It simply does not make sense to test silicon for TDP again and again. It will be too expensive for AMD to do that.

FX is guaranteed to be strained SOI and unlocked multipliers.

ANd karan 4200 and 165 are different cores.

AMD TDP ratings are for perticular core. Perticular core is having same TDP irrespective of their clock speeds.

As far as battle between the processors goes, honestly i will advice to get a 3000+ or opteron.

Buying 3500+ for higher multiplier is just not good enough reason.

The boards like A8N-E which are cheap are easily capable of reaching 300HTT and more. You always have ablity to use memory dividers. And the best VFM you will get is out of 3000+ and opterons, on an average opteron gives more consistant overclocking result. Crappiest of opties are not having much trouble reaching past FX speeds. Except few cases of people which contradicts this theory but that will always happen.

If you can have a look at the processor stepping before buying then its great. 0601/02/03 week 3000+ are doing exceptionally well.

3500+ i have seen here is really old stock, none of them were from this yeat. mostly from week 40s of last year. And those were not known for their overclocking potentials. i dont know if there is new batch that has shipped here in last week or so but previous to that most of them were old stock.

 

[BIKeINSTEIN]

Hey Funky,
I thought 3000+ production discontinued.:S
Or are these the last ones??

Hey Karan,
About Wafers- Better are always more expensive to produce!!
Hence the premium on their retail prices!!
eg. Pringles vs Lays = FX vs Rest!!:rofl: :cool2:

And me got a pal at AMD!!
Seriously!!
He said he could arrange for sample pieces pre-release, Dunno how!!
Else if he is bluffing, he's gonna get it from me!!
Will ask how their Binning is done- Speed, TDP or Both!!:cool2:

 

[Shripad]

Well there was a news floating around that 3000+ production has stopped and yes there was time when it was almost impossible to find.

But there are plenty of new stock rolling out from AMD, and this is new stock year 06, that means AMD might have thought otherwise. And it does make senser. This is the processor that was selling like hot cakes for them anyway, specially in asia.

 

[Darthcoder]

^^ They are really selling like hotcakes here,I went to a shop a week ago and saw the 1st week 2006 and within a week there was this new batch of 3rd week 2006 at the same shop.

 

[Chaos]

^^I got a 0603 DPMW a week back. Overclocked well beyond my expectation. Its doing 2.6GHz stable at stock vcore . Haven't tried pushing it higher but I'm sure with 1.5-1.6V, it'll do a lot better.

 

[Darthcoder]

Wow!! hyeah: See if you can beat this

http://www.amdgeeks.net/overclock/156

Which am sure you can do .

 

[Darklord]

You can urself check out that the TDP of all three processors is the same. So anish, your argument of 146 being a "better" silicon, cooler is totally unresearched..

You say this and now you say,

I am reprimanding him for giving general views regarding 1xx and 3500+ not between 144, 146. Even a child would be able to tell that 146 has a superior core compared to 144 because the 146 clocks higher than 144 at same TDP.

Once you ask to compare the 3 which includes 3500+ then you say you are not.Please,you may call me a retard,but i am confused here.COuld you please make your point more clearer ?:huh:

Take your argument forward, the 3500+ runs @ 400 Mhz more speed with same TDP. Wait theres more, a Opetron 165 has a TDP of 110W @ 1.8 Ghz..... where as a lowly (inefficient as per ur argument) 4200+ has a TDP of 89W..... that 2 cores, with 400 Mhz extra each over a Opteron 165. Care explain it?

Well the 4200+ is based on Manchester Core having 512 x 2 Kb L2 cache.The Opteron 165 is based on Denmark core which is 1MB x 2 L2 Cache. 1MB of L2 cache difference between the 2.Both are based on entirely different cores.
And how come 400MHz each ? doesnt the Opty run @ 1.8GHz and the 4200+ @ 2.2 GHz ?
So shouldnt that be 200MHz for each core ?
Correct me if ia m wrong.
Also i didnt get your point in the above comment.Please explain.

Again, I have a point to counter your possible counter arguments. Many times AMD passes off higher end cores as lower parts just to fill stocks. Supply and demand.

When did i ever disagree on that ?

Hmm, have you heard of a tool called TCaseMax Darky? Each E rev core has different exact specifics of the TDP and the max case temparature. That tools helps in finding that out. Check between a Opteron 144 and a 3000+. A 3000+ E6 will have a TDP of around 50W - 55W (on an average) whereas a Opteron 144 will have it around 55W-60W (again average.)

So ?
Also i would love to learn more about this tool.Please help me.

Hmm, have you heard of CCBWE 06xx XXXX 3500 single cores Darky? Those are DUAL core Manchesters being passed along as lowly 3500+ Venice?

How ? this means AMD is giving one core free...lol
j/k
But i am not convinced that the 3500+ had such a huge demand and AMD had so many spare Machester cores as to castrate them and sell as 3500+
Infact AMD is finding it difficult to meet the demand for their Dual Core CPUs.If we consider your logic then this means AMD is having a lot of spare cores.Care to explain ? :huh:

Hmm, have you heard of CCBWE 06xx XXXX 3500 single cores Darky? Those are DUAL core Manchesters being passed along as lowly 3500+ Venice? Hows that happening? Surely, now we are talking difference between the die size of 3500 Single core venice vs Dual core manchesters. Well, AMD is comparitively small in its manufacturing capability area and thus it has to resort to these antics to get by supply demands. Similarly, my argument (which I said was my own theory) of bad higher end chips being passed off as lower end parts because they didnt meet TDP requirements.

Dont get me wrong dude.The point i am trying to make is
Yes i agree AMD dumps bad cores [Meant for FX series] into lower models.I am just saying they will use it in CPU models that feature 1MB L2 cache and not 512Kb cache.Anyways this is all speculation,we REALLY dont know what is REALLY happening,do we ?

Hmm, a 1U rack is as small as it gets. Imagine having a 80mm DELTA fan sitting atop a FULL COPPER hsf (measuring all of 8 mm height, again approx) trying to cool a overclocked 170 @ 2600 Mhz.

Well karan,not a single Opteron [Used in Servers] is overclocked.So why does this overclocked thing come in between ?

Even a XP-90C is not sufficient to cool OCed dual core monsters just because its size is too small and it is unable to radiate enough heat quickly (>>> 110W)

I dont think XP90-C was designed for Dual Core CPUs,was it ?

Darky: Opteron 180 vs Opteron 280 Now does that convice you at all? Or you still think I am speaking out of my ass.

What are you trying to say karan ? i dont get your point.:huh:
Also between these 2,just see the current rating,theres a difference of 10A.And believe me 10A is a lot of current,taking into consideration that we are talking about CPUs here.

Or the AMD marketting people would have you believe.

AMD PR people never market that :S

I had a 3000+ Venice before my Opteron 165 (the irony ) It had a measly 512KB cache. It gave me SPi1MB runs of 33.20 seconds at best (oced to 2475 Mhz). Now I have Opteron 165 oced to 2.675 Mhz with even faster ram. So thats a 200Mhz of RAM CPU power and another 200Mhz for extra 512KB cache (according to ur point) So thats a cool 400Mhz extra over the 3000+ venice. Why am I still getting best SP1MB times of 31.2 seconds? I should easily break 30s isnt it Darky? 512KB extra cache shows "selective" performance improvements because the A64 architechture is really really independent of "the need for higher memory bandwidth." I repeat, for A64, Mhz is the king....

Well Karan,SuperPi is not optimised for dual cores.Perfect way to compare would be to test 3000+ Vs 3700.
Comparing 3000+ [Single Core] Vs Opteron [Dual Core] is baseless,keeping into mind we are talking about SUperPi here.
You think i dont know that ?
I have myself experienced this,Superpi scores of my 3000+ are unbeatable by the Opty.

Dont tell me you were unaware of the SuperPi thingy.

For A64,MHZ is the king ------> Really? i didnt know that.Thats a revealation to me.:tongue:

But you are totally misinterpreting the point. The improved performance is "NOT" due to "lower" multiplier (which lol, is still a possibility with the 3500+) but due to the higher HTT speed you end up using. AFAIK, higher HTT yeld points in SP1MB as well as various 3Dmarks (most prominently 03.)

Dude i was not talking about performance man,i was talking about PURE OC capability.
For example,my Opty refuses to go beyond 2.4GHz @ 10X multiplier.Whereas when i drop it to 9X i achieve 2.8GHz.THat was my ONLY point.
My point was,yes higher Multiplier helps but not always.sometimes you get better yields with lowering the multi.THis takes out the higher multi thing outta the picture.
Thats my point nothing else.
I repeat,i AM NOT talking about performance here,sheer OC capability.

With a 144 you would be stuck with a max multi of 9x whereas with a 3500+ you would have choices upto 11x. And that you eliminate the need to get a high priced MOBO because a lower multiplier demands a MOBO which is able to do higher HTTs. Isnt it?

Completely agreed but i would personally give more weightage to cache rather than multipliers.

Just my personal opinion Karan.

 

[Crazy_Eddy]

Opty 1xx series are "NOT" special in any form. They are just left over San Diego cores which AMD wanted to get rid off. So they branded it differently and sold it off. But, because of the "chance" higher OC ability of the initial batches, it generated a lot of demand and AMD had to specially continue manufacturing the "San Diego" cores which was against their plan. Hence they are trying to curb it with a price hike.

Absolutely true. However, I'm not sure if they were left over cores that AMD wanted to get rid off. The Socket 939 Opterons were designed mainly to encourage very small businesses to set up Cheaper Opteron-based servers, using cheap/easily available Socket 939 boards, and standard DDR RAM instead of ECC RAM. The San-Diego core was the obvious choice for these Socket 939 Opterons. The initial CABNE stepping batches happened to be FX-57 cores, thats why they overclocked tremendously.

Of course, the pricing being very close to the Desktop A64 parts as also the overclock potential led to AMD finding the parts were not reaching the intended parties/and their desktop parts were not selling, hence they tried to put a block on availability of these parts to the retail customers.

Take the case of the magical stepping of 0517 DPMW/EPAW....(Almost) All these chips would do 2.6 Ghz with default Vcore. It could be a bad batch of FX-57 Sandiegos which actually didnt meet the TDP requirement and thus were branded and sold off as different lower frequency chips.

This i'm not too sure, cause the 0517's were typical Venice cores with 512KB of L2 cache, not the San Diego ones with 1MB of cache.

You can urself check out that the TDP of all three processors is the same. So anish, your argument of 146 being a "better" silicon, cooler is totally unresearched....

Agree again. I have no idea why everyone keeps claiming the Opty's should run cooler :huh:. Its only obvious that the more L2 cache you add, the warmer it is bound to run. In fact, the increased L2 cache on the Prescotts was one of the reasons why it runs so hot.

 

[Shripad]

BTW adding to Tcasemax tool. That tool is no longer an indicative of overclocking ability of all E revision CPUs.

The newer processors comming like opterons have fix tcasemax of 49C, these processors do not support tcasemax feature.

And about CPU multipliers i have experienced it myself. My 3700 oced more on multiplier 9X than stock 11. all overclocking you will see in that post your oc thread of my 3700 are at 9X multiplier. At 11x it simply refused to go past 2.66Ghz.

 

[BIKeINSTEIN]

Is that 1.8 and 2.2 resp per core??:S
Then isn't 2200-1800= 400??:huh:
Now i was trying my best to make sense of this discussion, but i am giving up!!:ashamed:
i am dazed!!:S
Exit- Stage Left!! ( in same voice as Pink Panther):rofl:

Mods!!

Request!!:cool2:

Please merge relevant replies of optty vs 3500 threade and this one!!!

That guy who needs to decide is most probably bulldozed by the discussion there!!:rofl:

At the end of it, he will have no doubts whatsoever over his purchase!!:hap2:

Or he just might give up the idea of a new rig!!:rofl:

Mod Edit: Thanks for the suggestions! Implemented!

 

[Shripad]

The only thing is you just cannot do generalised comparison between the A64 and opteron.

You need to understand that AMD is using both strained SOI and regular SOI to meet the demands at the moment. So it boils down to the stepping and feedback when you are to determine which processor is better.

As far as 0517 being left over FX 57, there is no chance of that happening.

they are both different revisions. 0517 was E3 revision venice. And FX 57 was E4 revision SanDeigo.

It might well be the case that 0517 was strained SOI chips.

the server grade processors usually undergo from better Quality check than desktop processors, thats about it.

You cannot generalise which is better, if you choose 4400 over 0550UPMW DC opteron when you plan to overclock then you can be called a fool. We all know how 50th week DC opterons overclock. So it fianally boild down to the price and the stepping and user feedbacks. Which has better silicon depends on stepping and fab. Earlier opterons did had better silicons compared to later year 05 venice. 054x and 055x E6 venice were mostly trash.

EDI: Umm wth did thread was just edited???

 

[hunt3r]

OK right now im confused which thread this actually is

Now one thing guys, i thought AMD's TDP rating was just a maximum value, andiIt isnt necessary that all chips run at the value specified? And Intel's TDP rating was actually a average value? :huh:

 

[Harshal]

I Need to know and clear "Demand & Supply" thingy man, this thread is kinda eye-opener (hehe "opener", Beer comes to mind ). As I always thought AMD isnt foolish enough to sell higher-range/rated parts as lower rated ones unless its not properly working at desired speed(Clocks) or Heat-specification(TDP Envelop). As there was a time when all kinda high-end parts were selling, Opteron which came along is a HOT CAKE for them. AMD does this to maximise the profit.. as throwing the semi-working silicon which costed it something (whatever lil) will be Suicidal for them.

 

[KiD0M4N]

mods..... i had started this thread with my post, and suddenly i come back, a brilliant idea "had" been implemented. ah well........

OT:

i told around 2 months back (again rough guess) that this thread:

http://www.techenclave.com/forums/opteron-939-oc-database-thread-64762.html

be renamed to a Opteron Overclock Discussion thread because by no stretch of imagination could be called a "Database" however it got no replies. but this was done uber quick. any reasons?

BACK TO TOPIC:

like i said, in the world of overclocking nothing is guarenteed. a guy buys 10 0550 UPMW chips and overclocks them to 2.9Ghz+, but he cannot give a guarentee that his next 0550 UPMW will do the same or anywhere near.

You cannot generalise which is better, if you choose 4400 over 0550UPMW DC opteron when you plan to overclock then you can be called a fool. We all know how 50th week DC opterons overclock. So it fianally boild down to the price and the stepping and user feedbacks. Which has better silicon depends on stepping and fab. Earlier opterons did had better silicons compared to later year 05 venice. 054x and 055x E6 venice were mostly trash.

hmm, yes buying a 4400+ over a 0550 UPMW would be retarded. but buying a select CDBHE over a 0550 UPMW would be a even better buy because then instead of being limited to 3.0-3.1Ghz you would be able to hits in the range og 3.4 Ghz-3.6Ghz because its common accepted fact that the "clock generators" in those chips are not cold bugged as badly as the BW/B2/B3/B1 etc.

Regarding "trash" silicon:

Do you realize that neither AMD nor INTEL produce chips so that they can be overclocked, rather their target is "probably":

1) get the stock speeds

2) minimum TDP (hence Vcore)

3) maximum "yeild"

they rate their produce depending on the type of yeild they get (meaning the number of working chips in a wafer) because at any point of time, a particular model is in high demand (which explains the sudden revival of production of 3000+) and thus they need to take higher bin to lower part workchains so that they are able to satisfy the market demand.

also: AMD has a process of self adjusting manufacturing. the manufacturing process keeps on adjusting to improve yeilds. thats why sometimes, the process becomes on "AMD"s viewpoint whereas the "OC"bility suffers badly on the overclockers viewpoint. A trend had been noticed on the new chips coming out now. they are doing wonderfully high speeds at very low Vcore increases in air but are getting badly coldbugged and HTT limited when going subzero. Some FX-60 are so badly coldbugged that when temparture are dropped below +14, it is becomign mighty dificult to take HTT any higher than 220Mhz. So it is quite possible that the manufacturing process is being adjusted to produce HIGH "air" clocking chips due to which the "sub" temp ocing is suffering. evidently, AMD is more concerned about producing chips which run stock @ 3Ghz+ on air instead of pertaining to a niche market of overclockers who can afford to free the core below 0C. I hope i hv not called "anyone" retarded in this and made my point clear.

Dont get me wrong dude.The point i am trying to make is

Yes i agree AMD dumps bad cores [Meant for FX series] into lower models.I am just saying they will use it in CPU models that feature 1MB L2 cache and not 512Kb cache.Anyways this is all speculation,we REALLY dont know what is REALLY happening,do we ?

I Couldnt agree more with you. What I am rambling on about is, in the "wild", many weird chips are noticed:

1) Cores which are labelled 3000+ but are the size of a regular 2MB L2 dual core size.

2) Single core Venices, whose cores are actually those of Dual core manchesters.

3) Single core Sandys (no reference to hamara Sandy ), which are actually dual cores where one core didnt cut it.

Hence I was just pointing out a possiblity and GIVING a wild example of FX57 being dumped as 3000+. You cant prove me wrong, I cant prove myself right. Just a example to communicate my point accross. Hence I suggest not to bring that point up again.

Well Karan,SuperPi is not optimised for dual cores.Perfect way to compare would be to test 3000+ Vs 3700.

Comparing 3000+ [Single Core] Vs Opteron [Dual Core] is baseless,keeping into mind we are talking about SUperPi here.

You think i dont know that ?

I have myself experienced this,Superpi scores of my 3000+ are unbeatable by the Opty.

Dont tell me you were unaware of the SuperPi thingy.

For A64,MHZ is the king ------> Really? i didnt know that.Thats a revealation to me.

Hmm, lets jog you memory into the history. You said extra cache of "512KB" translates to a extra "200Mhz" gain. I was coutering that argument. When did I ever compare Single core vs dual core? You must be reading me wrong or something.

I said the 512KB cache does not straight translate to a 200Mhz speed bump. Infact it cannot do that. The cache helps in reducing the latency of the A64 core to the memory somewhat and thus its use it totally dependent on the application at hand.

What are you trying to say karan ? i dont get your point.

Also between these 2,just see the current rating,theres a difference of 10A.And believe me 10A is a lot of current,taking into consideration that we are talking about CPUs here.

You have a knack of missing my "point" dont you? Look at the TDP ratings. And Rofl man, you of all people, comparing currents? You do know that that 10A means "nadda" unless that exact voltage at which the current is being drawn is known? You tell me (test of you standard 10 studies):

20 A @ 5V

100A @ 1V

Which is more?

Agree again. I have no idea why everyone keeps claiming the Opty's should run cooler . Its only obvious that the more L2 cache you add, the warmer it is bound to run. In fact, the increased L2 cache on the Prescotts was one of the reasons why it runs so hot.

Thanks for the clearance bro. Exactly my point. Some more examples:

4400+ @ 2.2 Ghz with 1MBL2 x 2 has TDP of 110W whereas 4200+ @ 2.2Ghz with 512KBL2 x 2 has TDP of 87W. Go figure.

How ? this means AMD is giving one core free...lol

j/k

But i am not convinced that the 3500+ had such a huge demand and AMD had so many spare Machester cores as to castrate them and sell as 3500+

Infact AMD is finding it difficult to meet the demand for their Dual Core CPUs.If we consider your logic then this means AMD is having a lot of spare cores.Care to explain ?

I guess you should frequent XS as much as you frequent VR-Zone then

Yes its true, the CCBWE 3500+ are dual cores where the second core didnt cut it. Didnt perform at stock Vcore, hence the processor was shelved. Does it mean that if you havent seen something, it "must" not exist? Also, no one has found a way "still" to enable the second core in a full proff manner. They just are sure because of the stepping and the huge core size which is much larget than usual Venice cores.

and finally:

Once you ask to compare the 3 which includes 3500+ then you say you are not.Please,you may call me a retard,but i am confused here.COuld you please make your point more clearer ?

ROFL man. I indeed asked told him that AMD 146 is not superior silicon when compared to 144 and 3500+ because you have the 3500+ which clocks 200 Mhz lower and runs cooler compared to the 146. BUT, the 146 is superior when just put in context of a 144 and a 146. End of story? Itna parishan kyun he mamu?

So ?

Also i would love to learn more about this tool.Please help me.

Take NIRVANA.

Our very own TE thread regarding that: http://www.techenclave.com/forums/tcasemax-tdp-and-overclockability-60036.html

Google and TE rocks:



Nice to have this debate with you guys....