AMD Zambezi news, info, fans !

[Manicdan]

amd has their definition of core, some do not agree with that definition
but the product is not going to perform any differently based on the name

[freeloader]

I'm not disagreeing with you but some people you just can't please. It has shared resources, big deal. I read somewhere that the module is 80-90% of the efficiency of a true dual core. Close enough for me to call a module two cores.

[informal]

informal so BD can do 2*4 and K10 just 3 macroops?

BD integer scheduler receives the 4 macro ops but the very execution units work on micro ops sent by that same scheduler. Execution units can only work with micro ops. Back end's job is to retire the macro ops that schedulers received. Per clock and per AMD (per AMD"s Mike Bulter) each core can retire 4 macro ops.

[TESKATLIPOKA]

freeloader yet you are still missing 10-20% of performance, thats a lot not to mention what they added was just a second cluster which means only 20% bigger core size without L2 cache, true 2 core without sharing would mean 100% bigger core size on a die.
something like this(without L2 cache)
module without one integer(no sharing, everything can be used by one integer cluster) 15.58mm2
module design 19.4mm2
2 core design 2*15.58mm2=31.16mm2
The performance gain per area increase is worth the performance penalty but comparing it to a regular 8 core is not right in my opinion thanks to the penalty hit.

[halofanman]

less than a month before release and people are still debating how many cores it has?!

from AMD.com "Experience responsive game play and mega-tasking performance with the AMD FX 8-Core Processor Black Edition.

World's first native 8-core desktop processor."

so if a core doesn't do two threads then it isn't a core according to intel fanboys
I find it somewhat ridiculous that they are questioning its amount of cores.

[TESKATLIPOKA]

informal thanks for the info.

[informal]

This is from dresdenboy @ AT:

It can do both In recent papers the Bulldozer designers call those ops "Cops" (complex ops, equivalent of a ALU/FP micro-op + a mem op [load/store/load+store]).

Decode: 4 Cops/cycle/module or up to 5 in case of branch fusion (IIRC branch op has to be in last place then)
Issue: 2 ALU ops + 2 AGLU ops per cycle per core plus 4 FP/SIMD ops per cycle per module in the FPU (belonging to both threads)

Bulldozer’s decode unit extracts and
decodes up to four x86 instructions per
cycle from raw instruction bytes. The decode
pipeline converts x86 instructions into Cops
that can directly execute on the functional
units.

The scheduler picks and
schedules four Cops per cycle to the execution
units out of order.

on FPU:

AMD designed the Bulldozer FPU to
deliver industry-leading performance on
HPC, multimedia, and gaming applications.
The primary means of achieving such
performance is a four-wide, two-way, multithreaded,
fully out-of-order FPU, combined
with two 128-bit FMAC units supported by
a 128-bit high-bandwidth load/store subsystem.

Source: Michael Butler, Leslie Barnes, Debjit Das Sarma, Bob Gelinas, "Bulldozer: An Approach to Multithreaded Compute Performance," IEEE Micro, pp. 6-15, March/April, 2011

Here are some links related to Chuck Moore's comments on Financial Analyst Day 2010, where he mentioned the 4 "instructions" per cycle issue per core and as also the same bandwidth of decode:
http://citavia.blog.de/2010/04/22/pr...143/#c12914412

David Kanter's article on BD gives more details if the software optimization manual is too cryptic.
http://realworldtech.com/page.cfm?Ar...WT082610181333

[undone]

I read somewhere that the module is 80-90% of the efficiency of a true dual core. Close enough for me to call a module two cores.

I wonder where did you read the '80-90% of the efficiency of a true dual core' statement, if it's true the Zambezi 8 core will only act like 8 * 80% = 6.4 cores in worst scenario?

[Dumo]

It will be a few weeks before Newegg put it on sale, lets just stop arguing.......unless peep really work at AMD's BD design team. Lol

[SkullCracka]

i think nda is up the 9/9 read that somewhere.

[informal]

I wonder where did you read the '80-90% of the efficiency of a true dual core' statement, if it's true the Zambezi 8 core will only act like 8 * 80% = 6.4 cores in worst scenario?

It's in official slides from FAD 2010. This is an average number,there are probably some cases where sharing doesn't cost any performance and somewhere where it does cost more. BTW Bulldozer has all (integer) cores turbo so this alone will negate some of the penalties that *may* occur.
Also note that there is no multicore chip design out there that scales perfectly with more threads and at the same time has none of the MT technologies(fine grained,coarse grained,SMT). Usually conventional CMP designs (a la Opteron,Conroe) scale ~90% with second thread,on well multithreaded workloads.

[freeloader]

freeloader yet you are still missing 10-20% of performance, thats a lot not to mention what they added was just a second cluster which means only 20% bigger core size without L2 cache, true 2 core without sharing would mean 100% bigger core size on a die.
something like this(without L2 cache)
module without one integer(no sharing, everything can be used by one integer cluster) 15.58mm2
module design 19.4mm2
2 core design 2*15.58mm2=31.16mm2
The performance gain per area increase is worth the performance penalty but comparing it to a regular 8 core is not right in my opinion thanks to the penalty hit.

Too me it's no different than Intel classifying it's parts as having 4 cores/8 threads. Bulldozer's turbo function will also make up for some of the 10 to 20% deficiencies in a multi threaded work load, however no processor (that I'm aware of) with a shared cache system can obtain a 100% unified work load over two cores on the same die. (cache latency, cache miss, operating system overhead, etc, etc...)

[informal]

Too me it's no different than Intel classifying it's parts as having 4 cores/8 threads.

The problem with comparing BD and SB with SMT is that in SB's case you have 3 ALU units(and 2 AGU units) in each integer core that are shared by 2 execution threads. The main idea behind this concept is to use pipeline as much as possible and not let those 3 ALU units sit idle (note that those ALUs are both for int and fp!). This happens usually if MT code is not written well enough or the nature of the problem that code is written for is such that it is not parallel-friendly((not able to be broken into smaller amounts of threads and efficiently executed due to many dependencies among the threads).
Now in BD's case,in each module you have a big shared front end and behind it you have 2 execution cores each of which has 2 integer ALUs and 2(+2) floating point execution units. As you can see there is a BIG difference between AMD's "cores" and intel's "threads/cores". Former are real hardware units that are partly dedicated(int) and partly shared(FP) while in later case we have fixed amount of units that are shared equally between 2 threads.

[EniGmA1987]

Per clock and per AMD (per AMD"s Mike Bulter) each core can retire 4 macro ops.

So just to confirm, that means a single module can retire 8 macro ops per clock cycle then? Because a module is supposed to be two cores. Which means top level bulldozer can retire 32 macro ops per clock cycle?

[chew*]

less than a month before release and people are still debating how many cores it has?!

from AMD.com "Experience responsive game play and mega-tasking performance with the AMD FX 8-Core Processor Black Edition.

World's first native 8-core desktop processor."

There is no debate about how many it has. AMD has designated 8 so 8 it is........

Regardless of all those specs, docs and whitepapers you can dredge up I am telling people how it works.

If they wish to not accept that fact.....well thats their problem.

[crazydiamond]

i think at this point we all understand how it works..

a 4 module bulldozer has 8 cores. yes some shared resources but regardless there is still 8 cores in the cpu. intel cpu has 4 cores, 8 threads. bd has 8 cores 8 threads. if you still see it as a 4 core cpu that is perfectly fine with me, no matter how we see/classify it we will all have the same thing powering our pc

anyways this is a 100% pointless "argument" so i will return to how i have been the last n months, quietly waiting patiently

[chew*]

i think at this point we all understand how it works..

a 4 module bulldozer has 8 cores. yes some shared resources but regardless there is still 8 cores in the cpu. intel cpu has 4 cores, 8 threads. bd has 8 cores 8 threads. if you still see it as a 4 core cpu that is perfectly fine with me, no matter how we see/classify it we will all have the same thing powering our pc

anyways this is a 100% pointless "argument" so i will return to how i have been the last n months, quietly waiting patiently

I see it as a 4 core part that can execute up to 8 threads, An engineer at AMD while I was there tended to agree with me, but what does he know.........

We obviously have members here that know more about BD than AMD.

[JumpingJack]

So just to confirm, that means a single module can retire 8 macro ops per clock cycle then? Because a module is supposed to be two cores. Which means top level bulldozer can retire 32 macro ops per clock cycle?

Yes and no. A module can potentially retire up to 8 macro ops per cycle when executing two threads and tracking two contexts. If only executing one thread it can potentially retire only 4.

This is the strongest argument that it is an 8 core CPU because when running 8 threads the total potential retirement will be 4 issue x 8

[crazydiamond]

I see it as a 4 core part that can execute up to 8 threads, An engineer at AMD while I was there tended to agree with me, but what does he know.........

We obviously have members here that know more about BD than AMD.

no need for the smart ass reply

considering amd calls it an 8 core on amd.com but you call it a 4 core, i dont see how you can call anyone out who says its 8core as "knowing more about bd than amd"

but anyways like i said, i honestly could not care any less what people call it because regardless its performance wont change depending on what the owner calls it.

edit: i even said in my last post i am perfectly fine if you see it as a 4 core so idk why you try to put me/anyone who thinks it 8core on blast like that

[duron]

I think a lot of the guys here should work for AMD,
maybe they could come up with a better CPU.?

[BeepBeep2]

no need for the smart ass reply

considering amd calls it an 8 core on amd.com but you call it a 4 core, i dont see how you can call anyone out who says its 8core as "knowing more about bd than amd"

but anyways like i said, i honestly could not care any less what people call it because regardless its performance wont change depending on what the owner calls it.

edit: i even said in my last post i am perfectly fine if you see it as a 4 core so idk why you try to put me/anyone who thinks it 8core on blast like that

crazydiamond I'm pretty sure chew has Bulldozer in possession or has seen real internal numbers and you haven't so why argue back? Just because marketing team thinks its cool to make sure everyone knows it's an 8 core doesnt mean .

Very strong 4 core + 4 half cores = great 4 threads and moderately crippled 8 threads. I expect IPC to be up well, but not when all 8 threads are used. Multi-core scaling will be nothing like people will want it to be, but in four threads I believe AMD will be very close to Sandy Bridge.

His smartass reply is in reply to your smartass reply about his facts...engineer behind Bulldozer that Brian talked to is not advertising/marketing person. Each module is literally a core and a half. Dumbed down it's Hyper-Threading in a physical form, with extra parts in the core not only to help some single thread scenarios but help execute 8 threads...imo I guess AMD should have called each module a core and made it execute 8 threads but then again each module is not a normal core.

I bet the CPU is around Thuban performance in multi-thread, if AMD wanted a good high core CPU the stars core is still a great arch. It's very competetive with Sandy Bridge from a performance per mm2 standpoint core only. Unfortunately IPC must increase which means larger cores and AMD doesn't have the cache density intel has so something has to give in the multi-threading dept for that to happen...

I think a lot of the guys here should work for AMD,
maybe they could come up with a better CPU.?

+1

/end sarcasm

[crazydiamond]

thx for the lulz...

[BeepBeep2]

thx for the lulz...

I'm so glad you've tested the CPU yourself and know exactly how wrong chew* is. Why the hell do you think chew* is leaving? He gives real information and advice and people like you act like he has no idea what he's talking about. Why do you think Zambezi scores so low in cinebench? I have a hunch that it makes a great quad core...

[dirtyd]

I'm so glad you've tested the CPU yourself and know exactly how wrong chew* is. Why the hell do you think chew* is leaving? He gives real information and advice and people like you act like he has no idea what he's talking about. Why do you think Zambezi scores so low in cinebench? I have a hunch that it makes a great quad core...

Are you chew*'s bodyguard or something? If he is not here giving advice, the world will go on Also it's worth noting crazydiamond never actually accuses chew* of being "wrong", and he's actually stating quite calmly a laughably obvious observation - that it is pointless arguing over whether you define this is a quad core or an octo core. The nameless engineers at AMD can call it a purple-headed people eater for whatever difference it makes (ie, none). AMD is officially calling it an octo core, end of story.

I can't believe I got sucked into posting in this thread, I wish they'd hurry up and bloody release it

[chew*]

no need for the smart ass reply

considering amd calls it an 8 core on amd.com but you call it a 4 core, i dont see how you can call anyone out who says its 8core as "knowing more about bd than amd"

but anyways like i said, i honestly could not care any less what people call it because regardless its performance wont change depending on what the owner calls it.

edit: i even said in my last post i am perfectly fine if you see it as a 4 core so idk why you try to put me/anyone who thinks it 8core on blast like that

It's not a smart ass remark, it's fact, merely stating what was said factually in a discussion.

Thank you for reinforcing one of the reasons why I am just about done with this.

Try to help people and get called an ASS, especially after helping that person only days ago in a PM.