Zambezi ES performance weirdness

[XRL8]

As its a new angle on the subject and you guys dont mind poking me ;-) .But boy i do feel stupid about the thread not being on topic anymore...

The bios needs to be tweaked for zambezi, so does the microcode. So i would not call it final.
Think they are working on AGESA rev0.0.7.6 right now.
The boards are released as "BD ready"... they will work with a BD right on, but for full support/functionality a bios update is needed...

My understanding was, as mainboard makers had samples for months now, chipsets as well (the 9 series) and that launch was supposed to be around now.That the bios is ready,should be some time ago.
When they advertise support for Am3+ CPU`s, i dont really think they can do it if its gonna work horribly...
What AGESA code is in current AM3+ MB`s ?

If you dont want to post in this thread, fine, but I will tell you one thing...
Latest AGESA code for Bulldozer does not equal bios's on released boards tuned for Thuban.
Considering windows vista/7 are specifically shipped with AMD drivers from K7 to K10.5, that makes sense

I partially didnt get what youre saying. Youre saying AGESA for BD isnt on current MB`s ?
That could happen, but it would mean there was "something" that derailed mainboards/Bios optimization as they should be ready some time ago.
Thats also a problem for people, i mean, yes, we can just flash the bios and voila.However most people arent that tech savy, they bought something that should just work.
Theres some marketing problems with that theory, if there is bios/agesa problem that needs addressing with a new agesa and bios update they could just postpone 9xx chipsets Mb`s liek they did with cpus and deliver fully functional product from the start.

The driver need seems somewhat more plausible.Phenom I comes to mind with its problems with CNQ ,that drivers fixed somewhat.And before that Dual core optimizer.
Thats weak tho, it would mean BD is going to work like it should only on driver supported systems.From what i know Intels cpus have not had this problem with either turbo boost, normal operation or idle operation.They just work.I was hoping for fully hardware turbo on BD.Turbo core on X6 was barely working , seems to be the same with Llano.

[nex_73]

What AGESA code is in current AM3+ MB`s ?

The leaked tests we have seen are probably done with AGESA rev0.0.7.4
http://www.xtremesystems.org/forums/...=1#post4845936
I know Gigabyte released a bios with rev0.0.7.5 just a few days ago
http://www.xtremesystems.org/forums/...=1#post4879777

And it's probably so that B2/C0 works far more better with the shipped mb AGESA rev0.0.7.4 then a B0. This the mb manufacturers & AMD know for sure, therefore it's fine to sell the boards now.

[BeepBeep2]

As its a new angle on the subject and you guys dont mind poking me ;-) .But boy i do feel stupid about the thread not being on topic anymore...



My understanding was, as mainboard makers had samples for months now, chipsets as well (the 9 series) and that launch was supposed to be around now.That the bios is ready,should be some time ago.
When they advertise support for Am3+ CPU`s, i dont really think they can do it if its gonna work horribly...
What AGESA code is in current AM3+ MB`s ?



I partially didnt get what youre saying. Youre saying AGESA for BD isnt on current MB`s ?
That could happen, but it would mean there was "something" that derailed mainboards/Bios optimization as they should be ready some time ago.
Thats also a problem for people, i mean, yes, we can just flash the bios and voila.However most people arent that tech savy, they bought something that should just work.
Theres some marketing problems with that theory, if there is bios/agesa problem that needs addressing with a new agesa and bios update they could just postpone 9xx chipsets Mb`s liek they did with cpus and deliver fully functional product from the start.

Latest AGESA rev (ie. not 0.0.7.5) is not ideal for Thuban/Deneb by itself...

[SkullCracka]

XRL8

wtf are you talking about turbo dont work on x6..... works fine and down clocks fine. do you read reviews at newb sites or something or suffer greatly from PEBKAC?????

you keep saying your gona stop the newbness. but 5 minutes later you post another long rant about your grassy knoll theories.

[XRL8]

Im not saying it doesnt work at all! I didnt say anything about downclocking also!.
Turbo core has some problems ,it seems like a bios/cpu/os combo and not pure cpu feature.
When i turn it on, performance drops for every kind of workload because foxconn didnt care to rewrite bios properly, they just added agesa.So thats one problem.
On windows xp it suffers too, because xp has different sheduler.And from what i read (cause i cant verify it myself) it doesnt provide the same performance as manually clocking to turbo value even if load is single core only, so its efficiency is not exactly superb.If you have a link to educate me more on the matter , please do.As i cant find any valuable myself now.
Anyhow there must be a reason AMD decided to make new version of it ,dont you think ? Its called Turbo core 2.0 for a reason.
EDIT.
Found something:
http://www.anandtech.com/show/3674/a...55t-reviewed/9
If turbo would work as advertised this shouldnt look like this.It works, but not fully efficient.

@Nex_73 , so there is new agesa available.Maybe someone with B0 sample will redo a test.

some pron:
http://www.youtube.com/watch?v=SvqD4282hgU

EDIT:

OBR says on his blog, hes going to do a comparison on few different mainboards with few bioses soon.So we should know about the agesa

[SkullCracka]

so your running a outdated am2+ board with a legacy OS with a am3 cpu and you dont see the problem there?

oh and if you trust anandtech for numbers. i have some swamp land in florida you can buy.

[XRL8]

so your running a outdated am2+ board with a legacy OS with a am3 cpu and you dont see the problem there?
oh and if you trust anandtech for numbers. i have some swamp land in florida you can buy.

Well, shouldnt it work always on every mainboard and every os ?I believe thats the case with intels solution.And windows XP 64 is hardly "legacy solution" ist based on windows server 2003 which is very widely used.
As for anands numbers, yes i do believe his numbers, i dont believe he is stupid enough to put false numbers.He just picks benchmarks that favor intel and his comments are thrash.
And as of another note, i dont know if you know this, but there was a time when he was very AMD sympathethic, it changed rather drastically around sandy bridge preview when he got new server farm from intel probably for free.
That is not however topic of this discussion.Core of the problem is, is this weird behaviour driver related, if yes, than in theory this impacts linux solaris etc.

[Manicdan]

intels solution is not to offer backwards compatible cpus.

be happy they put a ddr2 controller in thuban.

[XRL8]

intels solution is not to offer backwards compatible cpus.

be happy they put a ddr2 controller in thuban.

I dont get your reply, what it has to do with either BD B0 or AMD`s not perfect implementation of turbo core ?
Just because Intel does something else wrong, we shouldnt take notice of AMD`s shortcomings ?We should bash both of them.

And in this regard i think youre even wrong ;-) .Thuban doesnt have dedicated DDR2 controller, its the same one with two modes.
As for intel, core2quads worked with ddr2 and ddr3 also.And when they moved to nehalem ddr3 wasnt a problem as IMC integration forced changing of mainboards so lack of ddr2 support isnt a problem.
When i think about its pretty much only time when intel has NOT screwed its customers with changing the socket, 1366 was a necessity.The ones that followed were not.Dont you think ?

[BeepBeep2]

Well, shouldnt it work always on every mainboard and every os ?I believe thats the case with intels solution.

That doesn't make sense...
Intels solution is a socket switch, 1100T in AM2+ (I have a board it will run on AM2 as well) is like running Sandy Bridge in LGA775...

And in this regard i think youre even wrong ;-) .Thuban doesnt have dedicated DDR2 controller, its the same one with two modes.
As for intel, core2quads worked with ddr2 and ddr3 also.And when they moved to nehalem ddr3 wasnt a problem as IMC integration forced changing of mainboards so lack of ddr2 support isnt a problem.
When i think about its pretty much only time when intel has NOT screwed its customers with changing the socket, 1366 was a necessity.The ones that followed were not.Dont you think ?

First of all, if it has a DDR2 mode then it is obviously a DDR2 controller as well.

As for Core 2 Quad running DDR2 and DDR3, that was a mainboard thing. Those chips didnt have an IMC. AMD has had an IMC since A64 began...and you act like its a bad thing that new AMD CPU's work with DDR2 now?

[XRL8]

Christ, it seems AMD doesnt make any mistakes, and anything they put out is pure gold :-/ .Im a skeptic at heart, i find things that dont go well.Thats my thing :P.
And again you read too much into what i wrote.I didnt write anywhere its a bad thing that X6 works in Am2+ (it should!Thats good!).
I wasnt even whining about the fact that some mainboards work like crap with turbo.
I was only explaining why i think AMD`s turbocore 1.0 wasnt exactly pure hardware (cpu) solution.And Am2+ thing is only part of it.
Yes i know Core`s DDr2/DDr3 was due to memory controller being in the northbridge of the chipset, i was replying to manicdans assumption that intel always changes sockets for cpus.LGA 775 was the same for ddr2 and ddr3 ,and that was the point.Sheesh.

I played a little bit with intel chips sporting turbo boost ,and it seems to perform regardles of OS or mainboard used always the same way.Thats my point.And more efficient at that.If im wrong ,i apologise for the inconvenience .Its just that when driver related B0 problem was mentioned, i remembered my problems with X2 without dual core optimizer, which my friend with core2duo hasnt had.

ps.I dont even care about turbo, i mentioned it only because it may be one of the reasons why B0 performs poorly :-[

[SkullCracka]

allright so now lets end it here and get back on track.... except one last thing xp64 has less support then win98ME and that isnt going to change ever. so ya xp64 is a legacy OS. 790fx legacy chipset.

hope NDA is over soon and we got some OMGWTFBBQ reviews and benches. if not it doesnt matter my 1100T is working awsome. passmark im right up there with a 2600k area of points.

[XRL8]

For me legacy os means something like dos ,win98 or win2000.OS that isnt used anymore beside some niche places.
XP is probably still most widespread OS in the world, and Xp64 is 64bit evolution of that with added stability and features.
You probably hadnt tried it in a while.In the begininng there was lack of driver support.However now its very good.
Best microsoft OS ever, no Legacy 16bit support ;-) .No bsods, stable, fast.
I would gladly pay MS to just include trim and dx11 cause thats only two things it misses.But i have windows 7 installed too and problem was the same, on my mainboard turbo just doesnt work properly.
As for 790FX, at least NB is VERY similar to 890FX and 990FX.Hardly ancient technology.Remember that AM3 started off with 790FX.

[SkullCracka]

so you agree there is no weirdness. ok cool. no 790fx is legacy and a foxconn? really.

[undone]

From OBR

[crazydiamond]

hmm wonder if its reading correctly ... looking forward to some benchies

[Daveburt714]

Seriously Folks, don't put too much faith in OBR's results...
He posted all kinds of wacky stuff when Phenom first launched, alot of which turned out to be BS.

Personally, I wouldn't trust AnAnd's results either.
I'm not suggesting they intentionally try and make AMD look bad, I'm just not sure there's anyone over there
with real skills when it comes to tweaking/testing AMD gear...

@XRL8: I'm not naive, and it does seem like there are some minor problems with the new chips.
I could be wrong, but I'm actually leaning toward a 32nm process problem since we're not seeing desktop (high speed)
Llano chips yet either. BUT, some of your implied conspiracy theories seem a little paranoid....

I'm anxiously awaiting some new AMD gear too! I'm sure they're very close, and honestly I'd rather they take their
time and release the best chip they can rather than repeat the mistakes of Phenom I.

[XRL8]

so you agree there is no weirdness. ok cool. no 790fx is legacy and a foxconn? really.

Wait, what ? Where i did agree there is no weirdness iwth B0 ?Stop putting words into my mouth.Of course 790FX is not a "legacy" hardware, its just only one generation behind,and 7xx chipsets work with even BD no problem.Yeah, foxconn sucks in support regard however thats of course true and irrelevant to the subject...

@Daveburt, what conspiracy theories ? All i said is that B0 most probably has a bug.Im not gonna bore anymore with details of that.
And yeah, im waiting for it too for some time now, i really REALLY dislike intels pracitices with regard to sockets and artificial price pumping.The fact that theyre still selling gulftown for at least 500$ is like a slap for 1366 owners.And 2011 is probably going to cost an arm and a leg, and i need those things ;-).So i have high hopes for AMD on this one.
Im loyal AMD customer from back in the k6-2 days, however this time i had enough of smalltalk.Show me the money AMD! ;-)

[Greg83]

Wait, what ? Where i did agree there is no weirdness iwth B0 ?Stop putting words into my mouth.Of course 790FX is not a "legacy" hardware, its just only one generation behind,and 7xx chipsets work with even BD no problem.Yeah, foxconn sucks in support regard however thats of course true and irrelevant to the subject...

790FX is legacy for bulldoser
it is compatible yes
but i doubt its gonna fully support all bulldoser has to offer

i think i'm gonna hold of for whats after bulldoser , the 8-10 core with tri or quad channel and new socket. that is prob the best thing to do. that way they can dump all the backwards compatibility

[demonkevy666]

From OBR

I couldn't get easytune to see more then 4 cores. On The My fathers Media center(UD3 790GX). when I put my thuban in it.

Orb kinda gives the implication that Bulldozer can only beat nehalem.

[undone]

I couldn't get easytune to see more then 4 cores. On The My fathers Media center(UD3 790GX). when I put my thuban in it.

Orb kinda gives the implication that Bulldozer can only beat nehalem.

From several claims from others about this guy, that makes me not believe him any longer.

[informal]

Zambezi 8C vs Nehalem 4C/8T in multithreaded apps using 8 cores/threads? Zambezi should have no problem disposing of Nehalem.Versus i7 Westmere it may get tough.On other hand,in singlethreaded apps it may be dead race vs Nehalem core(turbo ON for both),but nehalem is already pretty close to SB in that area anyway,so if this is the case then Zambezi will be mighty competitive.

[Dimitriman]

AMD needs to compete with 2600k this year. Yes, they will likely lose to SB-E in late 2011, but the new platform in 2012 with tripple channel ram, rev.2 BD cores and 5 modules should bring competition back on the table vs SB-E. Now, Ivy bridge will be another story but its unlikely that first generation Ivy Bridge will be high end, thus SB-E will probably be Intel's top dog for most of 2012.

[informal]

SB-E will be intel's top dog until haswell arrives in 2013.Intel may release some 6C IB models in late 2012,but it will be positioned against SB-E as 2600K is positioned Vs 6C Westmere now.
AMD will have 10C BD+ based Komodo,with 25% more cores and 25% more headroom for Turbo(counting GloFo's CTI program ).So it will be 10C komodo Vs 6C SB-E or in the worst case for AMD ,8C SB-E.