AMD Bulldozer server info

[jmm5351]

I don't think it's impossible to do, just not practical for AMD.

I know what you mean. I wish though AMD would make Zambezi dual socket. I mean think about if Thuban was dual socket. I bet almost everyone who was interested in a Thuban would end up buying two Thubans CPUS and a dual socket compatible motherboard. Just imagine running two 1090T's or two 1100T's or any Thuban processor for that matter. Two 1090T processors would only be $440.00. Half that of a 980X and it would have to be faster than an Intel Core i-7 980X. Especially after you overclocked them

[Sn0wm@n]

the dream of a dual socket oc'ing mobo is just that ... a dream for now ... amd did it in the past .. and since then the market has shrunk .. and not only that .. the platform didnt do so well .. so add 1 + 1 and we get no dual socket client platform

[Tomasis]

so much complains.. build two computers and use cluster?

[jmm5351]

so much complains.. build two computers and use cluster?

excuse me, but I was not complaining there. Merely just hoping if you will. But definitely not complaining. Can't a man dream these days anymore ?? And no I thought about doing that and it will cost too much money, maybe that is why I don't want to do that. For the cluster to work reliably on a daily basis, all the hardware should match, this also ensures getting the most efficiency out of your cluster. You ever wonder why clusters of super computers are all composed of identical computers with identical hardware in each. And another thing, you need special software to run the cluster. More money. And a cluster doesn't scale well at times because the software has to be written properly in order to scale the cluster when doing different tasks on the systems. Another reason I don't want a cluster is because I don't have room to put another computer. So please don't give me your stupid opinions. ouch...burned

Oh and if they were to make a dual socket Zambezi, the Mobo's better support crossfire

[Tomasis]

excuse me, but I was not complaining there. Merely just hoping if you will. But definitely not complaining. Can't a man dream these days anymore ?? Oh and if they were to make a dual socket Zambezi, the Mobo's better support crossfire

i understand how you are thinking.. JF AMD has stated its reason.. I dont know why it has to be expensive to build dual socket for AM3+. Couldn't mobo manufacturers take hand of the design? or it is more complicated than like this when chip design is involved?

However, if we look to our real needs, it is cheaper to build two computers like crossfire system. Take what it is available.

to build a bling car like from MTV program or a real pure breed racing car. This last alternative is more fun but more expensive as to have a server with Gxx socket or show off with consumer dual socket

[Tomasis]

I thought about doing that and it will cost too much money, maybe that is why I don't want to do that. For the cluster to work reliably on a daily basis, all the hardware should match, this also ensures getting the most efficiency out of your cluster. You ever wonder why clusters of super computers are all composed of identical computers with identical hardware in each. And another thing, you need special software to run the cluster. More money.

you have answered for yourself.. if you dont even know what is cluster

so FORGET dual socket. it is not for you

it is exactly JFamd talking about.

[Fatfool]

I have known about these boards since they first existed, and like you said they are Socket C32. You may want to consider re-reading my post. I did not say Bulldozer, more specifically I said Zambezi dual socket systems. There is no way that a Zambezi AM3+ CPU is going to be compatible with C32.

Zambezi pretty much the same as Valencia. That is if AMD decides to do what it has been doing with Thuban and Lisbon (okay, lisbon isn't exactly the same as thuban as it lacks the turbo core feature but you get the idea.)

I'm using the term interchangably as essentially, they are the same, only with a difference in packaging. Even pricing is similar. Hence, if you want dual Zambezi, just get 2 Valencia opterons, which will hopefully be priced similarly to Zambezi Phenoms and use a dual socket C32 board.

You can use the same memory, PSU, case, blah blah blah but just with a second socket and processor.

What's the point of maintaining two SKUs with no difference (except the socket)? C32 and G34 is a different case because G34 really has some features which C32 doesn't. (quad sockets, dual dies etc)

[RaV666]

Zambezi pretty much the same as Valencia. That is if AMD decides to do what it has been doing with Thuban and Lisbon (okay, lisbon isn't exactly the same as thuban as it lacks the turbo core feature but you get the idea.)

I'm using the term interchangably as essentially, they are the same, only with a difference in packaging. Even pricing is similar. Hence, if you want dual Zambezi, just get 2 Valencia opterons, which will hopefully be priced similarly to Zambezi Phenoms and use a dual socket C32 board.

You can use the same memory, PSU, case, blah blah blah but just with a second socket and processor.

What's the point of maintaining two SKUs with no difference (except the socket)? C32 and G34 is a different case because G34 really has some features which C32 doesn't. (quad sockets, dual dies etc)

Well, lack of unlocked multiplier and overclocking functions come to mind.

[Fatfool]

Well, lack of unlocked multiplier and overclocking functions come to mind.

That really doesn't require AM3+ to accomplish. C32 can do that as well, just that the current boards are all server ones without such functionality and the processors themselves of course aren't unlocked in either case. AMD could put unlocked opterons out, yet those settings will never be touched by any of the current boards.

[RaV666]

Thats what im saying for some time now.AMD has its all ready, all it has to do is to give green light to mb manufacturers, and release a opteron with unlocked multi for them.Call it zambezi fx or whatever and its done.

[JkS]

I know what you mean. I wish though AMD would make Zambezi dual socket. I mean think about if Thuban was dual socket. I bet almost everyone who was interested in a Thuban would end up buying two Thubans CPUS and a dual socket compatible motherboard. Just imagine running two 1090T's or two 1100T's or any Thuban processor for that matter. Two 1090T processors would only be $440.00. Half that of a 980X and it would have to be faster than an Intel Core i-7 980X. Especially after you overclocked them

Totally Agree, I would definitely love to buy a dual socket, especially in the case of Bulldozer with turbo core 2.0, I know I am not alone in this.
While JF-AMD disagrees time and time again. Quite honestly though 2x6cores at half the price of a single 980x is quite delicious

[AliG]

Totally Agree, I would definitely love to buy a dual socket, especially in the case of Bulldozer with turbo core 2.0, I know I am not alone in this.
While JF-AMD disagrees time and time again. Quite honestly though 2x6cores at half the price of a single 980x is quite delicious

That's because like any logical company, AMD has simulated the statistics and projected the net sales out of making such a design addition. If he says that they would lose money on it, then I trust him. I don't see why anyone would lie about something like that and turn down more potential profits

[JF-AMD]

Correct. There are definitely a lot of people who would like to see that product, I don't deny that. There just aren't enough people to make it anywhere near being profitable.

[Manicdan]

Correct. There are definitely a lot of people who would like to see that product, I don't deny that. There just aren't enough people to make it anywhere near being profitable.

so if we can get a petition (and preorders) for 1 million units of client side overclockable 2P BD systems, you would have no problem building it (and buying your own island)

[RaV666]

That's because like any logical company, AMD has simulated the statistics and projected the net sales out of making such a design addition. If he says that they would lose money on it, then I trust him. I don't see why anyone would lie about something like that and turn down more potential profits

That logic assumes, corporations are never wrong in their assumptions ;-).
I believe that they choose to ignore that small market just because they are afraid about opterons sales and brand image.
Or they have nothing to show.Thats a possibility also.

[zalbard]

so if we can get a petition (and preorders) for 1 million units of client side overclockable 2P BD systems, you would have no problem building it (and buying your own island)

LOL, man, be realistic, even 1000 is hard to get in this case, 1M is just plain impossible...
XS has only 100K members, and doubt that more than 10% of them are active, and it's unlikely that more than 0.5% will vote.

[Manicdan]

LOL, man, be realistic, even 1000 is hard to get in this case, 1M is just plain impossible...
XS has only 100K members, and doubt that more than 10% of them are active, and it's unlikely that more than 0.5% will vote.

i wouldnt even count their vote until i got a check in the mail, lol

i think we just need to see some very awesome board maker throw in some overclocking tools into a bios first and then see what happens. although i know almost nothing about writing a bios and have no clue if you can "add" base clock adjustments to a board that was never built to run at anything other than 200mhz

[Sn0wm@n]

That logic assumes, corporations are never wrong in their assumptions ;-).
I believe that they choose to ignore that small market just because they are afraid about opterons sales and brand image.
Or they have nothing to show.Thats a possibility also.

and your assumption is that you are also right about the issue at hand ... so who's right??? you or a big corporation who did thousands of simulations???

i wouldnt even count their vote until i got a check in the mail, lol

i think we just need to see some very awesome board maker throw in some overclocking tools into a bios first and then see what happens. although i know almost nothing about writing a bios and have no clue if you can "add" base clock adjustments to a board that was never built to run at anything other than 200mhz

then take a 2p mobo ... mod the voltage dealing part of the mobo and bios yourself and use locked multi opterons .... its doable .. but its a pain in the you know what

[RaV666]

and your assumption is that you are also right about the issue at hand ... so who's right??? you or a big corporation who did thousands of simulations???

Youre assuming they did any.We dont know that.
I personnaly think they didnt and wont.Too small market, and theyre changing theyre course (starting with changing the CEO)
I dont know if im right, for that, we would have to know exact performance numbers.
What i do know is that intel is making cash with high performance single socket, and evga is making money with their two socket board.
The rest go figure yourself.
You seem to defend AMD restlessly with very emotional attachement , you know they dont care about you me or anyone else ?

[Sn0wm@n]

Youre assuming they did any.We dont know that.
I personnaly think they didnt and wont.Too small market, and theyre changing theyre course (starting with changing the CEO)
I dont know if im right, for that, we would have to know exact performance numbers.
What i do know is that intel is making cash with high performance single socket, and evga is making money with their two socket board.
The rest go figure yourself.
You seem to defend AMD restlessly with very emotional attachement , you know they dont care about you me or anyone else ?

i am assuming they did and your always assuming they didnt ... why do i have the feeling that you will never understand amd's pov in this situation???



it isnt worth it to invest in a tiny market ...

and when all else fail ... call someone else a fanboy ...

[RaV666]

Its not like something failed me ;-) .
My arguments are:
AMD did high performance markets before.
Intel does it all the time.
All sugests it wouldnt be hard for amd to do enthusiast 2P c32 board or even single G34 system.Its all ready basically.

Your arguments are:
Amd doesnt want to :-).
I continue with my line of thinking because i havent seen proof to think otherwise.JF didnt and probably even cant (for the sake of keeping his job) show us some hard data why yes or why not.From his point of view such systems would be even bad for his work.
Everyone is entitled to his opinion if there isnt any hard proof to think otherwise.Btw, AMD doesnt have a point of view, its not a person :-).They have statistics, accountants, different groups fighting for different things ,board ,engineers.Its not like they all share the same viewpoint.

[Sn0wm@n]

amd didndt do well with their last 2p client mobo .. remember quad-father??? well he wasnt so much hot .. and considering the market is shrinking i bet it adds up to the reason to not build such a platform again

they all want the same thing ...


the amd pov is more money ... duh ....

[spicypixel]

Want dual socket? buy a server board. Want overclockable dual socket? buy skulltrail. AMD isn't going to get upset if you buy products they can't offer, it's not a financially sound plan. Intel can afford to have unprofitable lines of products countered by their other products in basically making skulltrail a marketing stunt.
AMD's stratergy of aiming to make every division and every product line profitable is a sound one.

edit: just to make it clear I'd love a dual phenom board for 3d rendering on the cheap, but there in lies a problem, cutting into the opteron business.

[MrMojoZ]

Its not like something failed me ;-) .
My arguments are:
AMD did high performance markets before.

So by your own argument, you acknowledge that AMD knows more about the subject than you do. Interesting.

[RaV666]

amd didndt do well with their last 2p client mobo .. remember quad-father??? well he wasnt so much hot .. and considering the market is shrinking i bet it adds up to the reason to not build such a platform again they all want the same thing ...
the amd pov is more money ... duh ....

Yup Quad FX failed, but the reason was Q6600, not the platform itself.
If it would be king of the hill, it would not flop like that.
BTW skulltrail was for core 2 quads ;-)

.EDIT.
Of course AMD as a company has vast amounts of information more than i do.
It does not necesarilly mean it as a company is making the right choices.
Intel went with P4 having vast amounts of information and years of simulations.They build itanium having all the information in the world and putting in it billions of dollars.
Nobody did tablets seriously before apple came and showed it can be done and it will sell.But i digress.
IF AMD has comparable architecture, i see no harm in going for performance oriented market.I said all i could.
Lets wait for some performance numbers first.