AMD Zambezi news, info, fans !

[2good4you]

Much of the speculation is embarrassing in this thread.

First, it has some serious decoding power (4 complex-ops or "cops" and in some cases 4+1 decode) and in total a massive front-end to keep up feeding two integer execution units, which more guarantees it's throughput to a single ALU. With that said, singelthreaded performance will not be bad. How it will do in 4+ threads, it will still be better then SMT.

The FPU, it's since long time very clear that it is very powerful and flexible. It will be as good or better than the one in Sandy Bridge.

Also the cacheprefetchers should be very aggressive and some singelthreaded games should see some great speed.

Don't forget the additional performance improvement in total from the Turbomode in 1-4 threaded load, which will give ~15% plus.

Be patient, soon it's here, and also the numbers.

And OBR... I have no word for so much time spent on nothing. Don't write about him.

[chew*]

chew* Thuban had the same TDP as many Phenoms and they were on the same process just older Stepping so I don't see any reason why they can't make an 8 core Thuban with the same TDP on a smaller node.

Thuban had the same tdp but lost effeciency core per core to deneb if you guys all forgot about that, so you shrink it maybe increase ipc which will increase tdp somewhat.......you can do one of two things, scale clocks and be right back at same tdp or add two more cores, sacrifice scaling, that's really not an option.

An 8 core thuban would have been damn close to gulfy in multithreaded however.

[Manicdan]

i put mine back together in an online spreadsheet (was able to copy teskatlipoka's as a starting point, thanks!)
https://docs.google.com/spreadsheet/...UE&output=html

in summary the numbers i have show near identical scaling to SB up to 4 threads, then quickly takes a lead up to 8 threads

in reality i bet that 1-4 threads BD will be behind SB by 5% on average no worse than 10%, and in 8 threads be ahead by 20%+ in most cases.

[TESKATLIPOKA]

chew* My reply was just a reaction to this

Your forgetting X factor.

TDP.

Doesn't matter if it would be faster if it can't fit into an acceptable TDP range.

Thuban 8 core would be able to fit in an acceptable TDP range even if it didn't have the same clock as a 6 core on the same 32nm process but thats the case even with BD or other cpu, a 2-3 modul can be always clocked higher than a 4 module and have the same TDP.

[nex_73]

dump the thuban compare, recalculate it to sandy, then compare both to thuban.

If both 8 thread SB and 6 thread Thuban = 6
Then 6 thread BD = 5,8*1,1842 = 6,8683 ???
(This is for wprime @ same clock + turbo off)

[TESKATLIPOKA]

Manicdan Interesting, I didn't know google had an online document creator(viewer).

In a 4 module BD 1-4 threads are processed in different modules, only when you need 5-8 threads the remaining integer clusters are activated.
At least thats how I remember It from some leaked slides.

[Manicdan]

Manicdan Interesting, I didn't know google had an online document creator(viewer).

In a 4 module BD 1-4 threads are processed in different modules, only when you need 5-8 threads the remaining integer clusters are activated.
At least thats how I remember It from some leaked slides.

i dont think this is true, a slide by AMD said that the next level of turbo only happens when half the modules are power gated
Attachment 119457

[chew*]

If both 8 thread SB and 6 thread Thuban = 6
Then 6 thread BD = 5,8*1,1842 = 6,8683 ???
(This is for wprime @ same clock + turbo off)

Stop getting hung up on "CORES"

Do a price/performance compare of 8 thread sandy to 8 threads BD.

So we know 2600K = $279, we have suggested prices of 8150 at $300

How much performance does that come out to, it most likely close.

Then compare to 6 thread thuban performance that we have figures for.

[TESKATLIPOKA]

Manicdan but that's just max turbo, doesn't mean there won't be a lower turbo.
I would like to see how max turbo works on 3 module BD, half of it is 1.5
Why should they use solo modules in situations were SB uses only HT? They should use them in situations where SB uses cores that means 1-4 threads.
It's from OBR but on an official hardware site he works for, just translate it. It's up to you if you believe it or not, but this is reasonable if BD is aimed against SB 4 core+HT.
http://pctuning.tyden.cz/index.php?o...id=1&Itemid=57

[TESKATLIPOKA]

chew* I don't remember AMD commenting if the price tag 300 dollars was for the highest BD model although I think so to.

better by 7.5% on average would be very good. Actually my first chart was showing 10.5% on average.

It doesn't really matter, only 2-3 weeks remain and we will see how good BD really is. I want a Trinity so I want BD to perform really well.

[chew*]

I come up with this.

Based on price compare with sandy the difference equates to 1.075%.

Sandy at 6000 in wprime about 4.77

Thuban at 6000 in wprime about 4.78

BD projected performance based on price to sandy at 6000 4.72.

I personally don't think the gap is that small, but at least i'm being a realist, peeccimists are never disapointed.

Optimism is the path to a very dissapointing life.

[Manicdan]

Manicdan but that's just max turbo, doesn't mean there won't be a lower turbo.
I would like to see how max turbo works on 3 module BD, half of it is 1.5
Why should they use solo modules in situations were SB uses only HT? They should use them in situations where SB uses cores that means 1-4 threads.
It's from OBR but on an official hardware site he works for, just translate it. It's up to you if you believe it or not, but this is reasonable if BD is aimed against SB 4 core+HT.
http://pctuning.tyden.cz/index.php?o...id=1&Itemid=57

it has to do with how much TPD can be removed from inactive threads. if a module is still drawing power then the chip cannot turbo nearly as much.
thats why they have 2 turbos, one for when less than half of the module are active (max turbo boost, according to the slide), and one for when more than half are active (all core boost)
the amount of power needed to run a second thread on a module is probably very low (im thinking like 20-40% more power)

[TESKATLIPOKA]

chew* true about pessimists but I like being an optimist more

Manicdan i know about that and I share your opinion about the amount of power needed for the second integer in module.

[El Gappo]

Speculating with the aid of graphs now, srs business.

[Nintendork]

Manicdan
It's from...

Didn't it was established that no info from * is allowed? No matter if it comes with a picture of the pope.

[undone]

Some guys at amdzone find a result that a mysterious AMD Eng Sample(2*12 cores) rank first in sisoftware bench, surpass 4.21Ghz X5690

http://www.sisoftware.co.uk/rank2011...dcbb885b5&l=en

[cesariuth]

Some guys at amdzone find a result that a mysterious AMD Eng Sample(2*12 cores) rank first in sisoftware bench, surpass 4.21Ghz X5690

http://www.sisoftware.co.uk/rank2011...dcbb885b5&l=en

Interesting...but how true can this?

[BeepBeep2]

Interesting...but how true can this?

I think it can be very true...

[EniGmA1987]

^^^^Amazing for an Interlagos if true

[BeepBeep2]

^^^^Amazing for an Istanbul if true

Istanbul?

EDIT:
...oh lol

[SnipingWaste]

Interesting...but how true can this?

Its a ES and has unlocked multi. I have one too and with air cooling I can get all 12 cores stable at 3.3gig. Most G34 socket boards don't have multi setting in BIOS but Phenom tweak in OS can change the multi and voltage on the fly.

[chew*]

Some guys at amdzone find a result that a mysterious AMD Eng Sample(2*12 cores) rank first in sisoftware bench, surpass 4.21Ghz X5690

http://www.sisoftware.co.uk/rank2011...dcbb885b5&l=en

I believe it its 2 real cores glued together x 2 sockets which will always beat 12 threads X 2 sockets (12 real cores) 24 threads.

We would have a beast on our hands if AMD sold a desktop variant of glued together thuban.

60K cpu score on ln2 easy.......

[thematrixhazune]

Is it 12 or 16 core? Dresdonboy's blog states 16 but the site has it as 12

[Glow9]

I believe it its 2 real cores glued together x 2 sockets which will always beat 12 threads X 2 sockets (12 real cores) 24 threads.

We would have a beast on our hands if AMD sold a desktop variant of glued together thuban.

I always thought that was pretty much what AMD did with Deneb, couldn't beat 4 Intel cores so they just stuck 2 more cores on more or less

Are they releasing a 4 core bulldozer, or are we stuck with Denebs and such for quite some time? This interim rig is seeming like it is going to become my main rig with all these delays.

[SkullCracka]

its 2 - 12 cores...