Bulldozer coming 19/09/2011

lets hope that they are good performance, anything close to the sb chips would be good
tongue.gif
, as an 8 core cpu for a bit more than an i7 2600k would be amazing value
biggrin.gif


hopefully the quad core chips will be cheap and good performance, as I may just have to upgrade my i3
tongue.gif
 
^ and also

programming does not work like that!!!

it is not down to the threads that are available, BUT, more down to the amount of IRQs pushed on the CPU's registry stack.

this is something that modern programmers fail to realise (bad teachings) - only those that have an understanding for assembly language programming could understand this.

...and also why modern apps crash and BSOD, due to memory bleeding and CPU bottle-necking.

IMHO, modern .NET languages should be banned. every programmer should be taught ANSI-C as standard (no OOP - just pure realtime coding)

biggrin.gif

[I have the worst understanding of software on this whole forum for the most part]

Could they make code or whatever for it in the future though? Or not a chance? (Recognising two or more cores as one) if ways were found to tackle the error rates and power demand?
 
lets hope that they are good performance, anything close to the sb chips would be good
tongue.gif
, as an 8 core cpu for a bit more than an i7 2600k would be amazing value
biggrin.gif


hopefully the quad core chips will be cheap and good performance, as I may just have to upgrade my i3
tongue.gif

I think this could be the start or generation before the start of dual cores being a thing of the past for desktops.
 
[I have the worst understanding of software on this whole forum for the most part]

Could they make code or whatever for it in the future though? Or not a chance? (Recognising two or more cores as one) if ways were found to tackle the error rates and power demand?

in a word.... a very short one..... NO

a program do not run differently on a machines with different of cores

IE:

32-bit code does not run faster on a 64bit machine... in fact, it actually wastes memory.

cores in coder terms are known as affinities (core0; core1; core2 ... coreN)

to write to all cores simutaneously requires OOP code (object orientation), or acces to a low-level language (purebasic can do it)

a part of a program (class) is assigned an affinity, but always reports to core0 :: core0 holds a class that encases all global variables

core0 will always wait until all other affinities (cores) have done their processing before triggering the said affinities for processing again.

so.... it does not matter how many cores you have... they will only work at the speed of the slowest core (normally core0, as it has to manage the others)

are you lost yet?

as for bugs and problems occuring - it is down to two things:

the coder

or

the compiler (m$ sucks big time .... borland used to rule:))
 
in a word.... a very short one..... NO

a program do not run differently on a machines with different of cores

IE:

32-bit code does not run faster on a 64bit machine... in fact, it actually wastes memory.

cores in coder terms are known as affinities (core0; core1; core2 ... coreN)

to write to all cores simutaneously requires OOP code (object orientation), or acces to a low-level language (purebasic can do it)

a part of a program (class) is assigned an affinity, but always reports to core0 :: core0 holds a class that encases all global variables

core0 will always wait until all other affinities (cores) have done their processing before triggering the said affinities for processing again.

so.... it does not matter how many cores you have... they will only work at the speed of the slowest core (normally core0, as it has to manage the others)

are you lost yet?

as for bugs and problems occuring - it is down to two things:

the coder

or

the compiler (m$ sucks big time .... borland used to rule:))

So my understanding: you can't alter how many cores make up core0 because that would be like pulling the gate of it's hinges/ changing the size of a plughole and expecting the plumbing further down the pipeline to still cope - which it won't?

But with hyperthreading and adding cores, core0 still remains in the same place?

I think I'm on the starting block of understanding this but will need some considerable shouting at
laugh.gif
 
So my understanding: you can't alter how many cores make up core0 because that would be like pulling the gate of it's hinges/ changing the size of a plughole and expecting the plumbing further down the pipeline to still cope - which it won't?

But with hyperthreading and adding cores, core0 still remains in the same place?

I think I'm on the starting block of understanding this but will need some considerable shouting at
laugh.gif

intel cleverly thought of utilising more core power by creating HyperThreading

most apps today are still 32bit, but most CPU are 64bit core(s).... meaning that a 64bit core could be set-up to run 2x 32-bit threads "virtually".

even thought affinity0 (core0) would have extra overheads by the workload of virtualising itself and the other cores, it would still gain work power from the said HyperThreading
wink.gif
 
intel cleverly thought of utilising more core power by creating HyperThreading

most apps today are still 32bit, but most CPU are 64bit core(s).... meaning that a 64bit core could be set-up to run 2x 32-bit threads "virtually".

even thought affinity0 (core0) would have extra overheads by the workload of virtualising itself and the other cores, it would still gain work power from the said HyperThreading
wink.gif

Now I'm with you
smile.gif
 
Just been reading somewhere else the fx-8150p is to be sold at 300$
63.gif


Eight native cores........Talk about value for money
tongue.gif

Or more of a worry. Doesn't that speak volumes?

If they're able to knock out a chip with 8 cores for $300 it's a concern to me. AMD don't give things away. Price to performance.. It's always price to performance. I'm sure it'll be $300 worth, but it ain't gonna be no Sandy beater at that price. If it was then trust me, they would want blood for it. I clearly remember AMD bashing out the original FX series which were rebadged Opterons. How much? about $650 much. And the reason? they laughed at anything Intel had going.

Intel's only response, due to their terrible CPUs at the time was to rebadge a Xeon with loads of cache to a "Pentium 4 extreme" and sell that... For $850.

If the 8 cored Bulldozer is really as fast as a 4 cored Sandy then it would be $600 or more. There's simply no way to manufacture 8 cores going that fast for that.

If I'm wrong of course then I have no problem dining on my headwear, but I just can't see it happening. Especially when AMD issued a statement around year ago that basically said they were no longer even going to try competing at the top end as they just couldn't afford to. On those rebadged Opterons they were taking losses, just to have the fastest CPU in the world. If you consider that at that time the equivalent Opteron was $1000+ and the only real difference was ECC support then yeah, they were making losses.

Usually a flagship model such as that will lose a company money. Asus do it often... Asus Mars, Asus Ares. The R&D that goes into them and the 'special' manufacturing of the unique components (coolers, shroud, PCB, even the schemtatics) coupled with the low sale amount (I mean they're a bit too exclusive for Joseph Public) usually sees them making a loss. But, they do it for the bragging rights and to attract people to their lesser cards. Kinda like, earns them a cool factor.

Edit. Here we go.. Take this into account....

http://www.lambda-tek.com/componentshop/index.pl?origin=gbase30.5&prodID=B279781

^ 8 cores. The kicker, of course, is you then need something like this.

http://www.amazon.co.uk/Asus-KGPE-D...1_1?s=computers&ie=UTF8&qid=1312027529&sr=1-1

And then horribly expensive ECC ram. Now I did my research before buying my I7 950 and it seems that Supermicro do a regular ATX sized single CPU board for around £200 or so (PCIE slot too). Couple that with the CPU and it wasn't that much more than what I got. Thing is, I read a review and they're not all that special. They're good for folding..

Any way yes, what I was about to get at. IMO all Bulldozer will be is the Magny Cours with higher clocks and an onboard GPU (if they're going to use them on Bulldozer, can't remember now).

Ed again, sorry. Now think about it like this.. that board is cheaper than the SR2 EVGA whatever it's called, and the CPUs are a quarter of the price.. Value? most certainly. Faster? lmao, no chance.
 
Or more of a worry. Doesn't that speak volumes?

If they're able to knock out a chip with 8 cores for $300 it's a concern to me. AMD don't give things away. Price to performance.. It's always price to performance. I'm sure it'll be $300 worth, but it ain't gonna be no Sandy beater at that price. If it was then trust me, they would want blood for it. I clearly remember AMD bashing out the original FX series which were rebadged Opterons. How much? about $650 much. And the reason? they laughed at anything Intel had going.

Intel's only response, due to their terrible CPUs at the time was to rebadge a Xeon with loads of cache to a "Pentium 4 extreme" and sell that... For $850.

If the 8 cored Bulldozer is really as fast as a 4 cored Sandy then it would be $600 or more. There's simply no way to manufacture 8 cores going that fast for that.

If I'm wrong of course then I have no problem dining on my headwear, but I just can't see it happening. Especially when AMD issued a statement around year ago that basically said they were no longer even going to try competing at the top end as they just couldn't afford to. On those rebadged Opterons they were taking losses, just to have the fastest CPU in the world. If you consider that at that time the equivalent Opteron was $1000+ and the only real difference was ECC support then yeah, they were making losses.

Usually a flagship model such as that will lose a company money. Asus do it often... Asus Mars, Asus Ares. The R&D that goes into them and the 'special' manufacturing of the unique components (coolers, shroud, PCB, even the schemtatics) coupled with the low sale amount (I mean they're a bit too exclusive for Joseph Public) usually sees them making a loss. But, they do it for the bragging rights and to attract people to their lesser cards. Kinda like, earns them a cool factor.

300$ is too low i think.with 1mb L2 per core and 8mb shared L3,they must be pretty expensive!?
 
Indeed. Well, mind you, maybe not.. I mean, here is your 8 cored system.

http://www.supermicro.com/Aplus/motherboard/Opteron6100/SR56x0/H8SGL-F.cfm

£212 from Lambdatek.

Then

http://www.amazon.co.uk/dp/B003BYRH...de=asn&creative=22206&creativeASIN=B003BYRHOE

£220. Which when I got my 950 was about £50 more all in. I soooooooo nearly did it, but the ECC ram was a kick in the cods. Thing is, in all honesty, the 950 wiped the floor with it due to all of those cores not being used. They fold amazingly well though, and there is a 12 cored CPU too.. Only downside? £900+

Now I'm certain that by now AMD can make those chips cheaper. And, that they will be clocked higher due to the skimming of the ECC etc. But, I just can't see them being that much better than those due to what it would cost. I said before that they won't put out a desktop CPU, costing half of the workstation equivalent, that's faster.

AMDs main business now is anything but the desktop market. When you consider that you can build a system of theirs with four 8 cored cpus for less than that of the dual xeon EVGA? *that's value*. Their servers are tremendous value and can carry up to 48 cores. Sick, insane.. call it whatever you like I guess
laugh.gif


Apparently they're absolutely incredible for folding... 48 cores I'm not surprised
laugh.gif
 
Hmm... looks like I willn't be switching to the bleeding edge of AM3+ Bulldozer. Why?... three characters, FM2. I still hope AMD move onto LGA like their Opterons; ZIF, heatsink/waterblocks and thermal paste is dodgy business.
 
Yup seems like it's another new socket.. Could be wrong of course...

I suppose we can just wait and see, but for gawd's sake, don't rush out and buy one until you have seen them reviewed properly.
 
Ah its inevitable, same as some people rushed to buy Sandy asap
smile.gif


FM2 is the new socket coming out with the decent Bulldozer chips (think the 10-core "Komodo"
smile.gif
), I guess they need a fresh slate anyway.
 
My wife needed a new board and CPU recently and I got her the A8-3850 and an ASRock F1 board. No, it's not breaking any speed records, but it's plenty fast for what she does, and it has USB3.0 and SATA-III on it too. It's much quicker than the P-4 it replaced, and she's happy with it. (all that I really cared about)

I could have gone with an i3 setup and been a little quicker for near the price, but I like to support the underdog AMD when I can. Without them providing some competition into the mix, prices would go crazy. (I think)

So yeah, wait for the reviews, then buy AMD if it's close enough to what you need, just to keep them viable.
 
That's like buying a Skoda so they can compete with Ferrari....

Ha-Ha! That would apply if those were the only two Car makers out there.

I'm using a 2600K for myself, but when I can, I buy from AMD just to keep them going because as I said, they help to keep prices in check.
 
It's such a shame that the news of another new socket has essentially ruled out the Bulldozer/Zambezi for my Christmas build. Been stung like this by AMD before, with the first batch of Phenoms/AM2+.

Still looking forward to seeing how they bench though.
 
Back
Top