The April Fool's Day Bulldozer Presentation

Chad Boga

Distinguished
Dec 30, 2009
1,095
0
19,290
Did you guys see the April Fool's Day Bulldozer "Performance Preview" presentation that appeared on AMDZone? :lol:

For those that missed out, I think I might be able to get my hands on the graphs in a little while.

If anyone has a copy of those graphs, please feel free to add them to this thread.

Even though the presentation was an April Fool's joke and the AMD Marketing guy who may or may not have leaked more info than he should have, freaked out and got it removed, I wonder if the "made up results", were disturbingly close to the likely results?

What kind of performance improvements in percentage terms are you expecting to see from Bulldozer?
 
me2, i dont expect bulldozer to be an intel crusher, intel has seriously been making sure they stay ahead of amd since core2, but i would love to see a drastic drop in tdp and a huge spike in ipc, i wanna see a 45watt quad @ 3.0ghz : )
 
IMO AMD has been putting just about all their design effort into BD, since the recent CPU releases from them (MC, Thuban, etc) don't appear to be particularly innovative or earth-shattering. So to be honest I would expect some big performance gains with BD. Whether it'll be better than Sandy Bridge, I dunno - there's a speculative thread on the subject over on AMDZone in case anybody wants to wade through it..
 

AMD seem to be betting everything on multithreaded applications, hoping to be able to fit twice as many cores into a die as Intel, but readily conceding the lowly threaded applications to Intel.

I suspect they have taken this approach mainly to try and shore up their position in the server market, and will try and stress that AMD are still a good choice on desktop because they will be fast enough on lowly threaded apps, and will shine on highly threaded apps like video encoding.

However, even if Bulldozer proves to be a more compelling processor than Sandy Bridge, they face the prospect of having to deal with Intel's 22nm refresh of Sandy Bridge, Ivy Bridge.
 
On paper anyway, the dual full integer pipes in the BD core looks to deliver a better hyperthreading than what Intel's current SMT does, at the expense of more transistors of course.

Plus AMD hasn't made a major screwup in what - 3 years now, since they fixed Barcelona's TLB bug. So their recent track record seems to have improved.

I sorta feel sorry for JF-AMD over on AMDZone. With friends like that, who needs enemies? 😛 I haven't seen any posts by him since April 1st however, so even he may have abandoned that site to the looney crowd..
 
Him 'acting' freaked out could have been part of the 'fooling'. *shrug I'm curious as well.

I have high hopes for BD, if there is more IPC increase rather than cores it'll be my next system. Assuming intel doesn't hold the price/performance at the time.
 

That is because they use two smaller cores.

Plus AMD hasn't made a major screwup in what - 3 years now, since they fixed Barcelona's TLB bug. So their recent track record seems to have improved.
Yes, but AMD's own information points to Bulldozer not making great strides in per core IPC improvements, they are just making smaller cores so they can fit more into a die.

I sorta feel sorry for JF-AMD over on AMDZone. With friends like that, who needs enemies? 😛 I haven't seen any posts by him since April 1st however, so even he may have abandoned that site to the looney crowd..
He got what he deserved.

He knew they were mentally unbalanced, yet he was whipping them up into a frenzy.

Like the old saying goes, "If you lay down with dogs, you have to expect to get up with fleas".
 
That is because they use two smaller cores.

1 module != 2 smaller cores and it's inaccurate to say such. It's much more accurate to compare it to Hyperthreading than saying a module is its own dual-core. AMD is just going a step further than Intel has (thus far).

Also, don't forget, AMD's "smaller cores" are going from 3-way int to 4-way int, which should bring a healthy performance increase. The Floating point unit is doubling, also (256-bit or 2 128-bit), so from 4-core to 4-module you gain ~double theoretical FP and ~2.6 times the theoretical integer power.

And the actual "core" of processors is pretty small, nowadays. Cache, IO, and the Uncore take up a huge portion of the CPU die. I think they're making a pretty smart move putting in that extra processing power. Kind of like brute force in an elegant way.
 

It is completely F#@%&g accurate and even said by AMD Used Car Salesman John "The April Fool" Fruehe himself.
 


But if JF was just playing along, he would have posted a "gotcha" the day after right? Seriously his post asking AMDZone to take down the bogus BD review that quoted him seemed pretty sincere, as well as desperate. However I see that he posted a blog over on AMD.com 3 days ago, so I guess he didn't lose his job over the prank.
 


So.... you're quoting someone you say is a fool. I have never actually seen that.

I don't give 2 cents about what anyone, repeat: A-N-Y-O-N-E said over at AMDZone. And neither should any other self-respecting geek.

And you don't have to get so bent out of shape. As a great fool once said, "I did not mean to blow your mind." :wahoo:

 

April Fool.

I don't give 2 cents about what anyone, repeat: A-N-Y-O-N-E said over at AMDZone. And neither should any other self-respecting geek.
That is fair enough, but when AMD's official marketing material refers to a Bulldozer module as consisting of two cores, then it takes some doing to dispute that, as it is actually in their interest to try and pass a module off as being only one core.

And you don't have to get so bent out of shape. As a great fool once said, "I did not mean to blow your mind." :wahoo:
Okay
 
What I thought was absolutely hilarious about the prank, was somebody posting in the thread about "how this could hurt AMDZone's credibility" 😛. As if they had any..

Almost as funny as their idol Scientia who can't seem to overclock an I5-750, earning himself lots of ridicule over on XtremeSystems. And of course Abinstein, who doesn't know the difference between rendering and encoding but feels compelled to chime in anyway on a thread.. And our ex-idiot, SighQ2, whom even the AMDZone fanbois find odd and incomprehensible. At least Ghost had the sense to perma-ban his original moniker.
 

Yes, I was roaring with laughter when I read that. :lol:

Almost as funny as their idol Scientia who can't seem to overclock an I5-750, earning himself lots of ridicule over on XtremeSystems.
It looks like that idiot didn't realise that he would get world wide publicity as the world's worst overclocker, because his pathetic attempts were reported on many forums around the world.

Now he has retreated into the World of Warcraft and who knows if he will ever be seen again. Even his cohorts on AMDZone express serious concern for his state of mind.

And of course Abinstein, who doesn't know the difference between rendering and encoding but feels compelled to chime in anyway on a thread..
I doubt that he would know the difference between his elbow and his arse.

And our ex-idiot, SighQ2, whom even the AMDZone fanbois find odd and incomprehensible. At least Ghost had the sense to perma-ban his original moniker.
I think his problems as ZootyGay started when he began insane Anti-American rants over in their "political" forum and as you can well imagine, those rants made as much sense as his rants against Intel. :lol: :lol: :lol:
 
b9vantage2-500x350.png

b8vantage-500x350.png

b7sandra3-500x350.png

b6sandra2-500x350.png

b5sandra-500x350.png

b18power-500x350.png

b17crysis-500x350.png


 


LOL - good ol' Zooty - couldn't remember the name for some reason. IIRC he posted in a thread here that he lives in the Canadian wilderness, uses dialup, & some ancient computer. Why he thinks that makes him a "tek" expert (his spelling), I dunno. I used to joke that we should post huge pix in his threads and stuff up his dialup 😀. Of course, he took great offense to that and dove headfirst into the ad hominem, even quicker than jenny! 😛
 


Hmm, now that I didn't know. IMO, WOW is like crack to certain people, such as my wife's nephew. He is on his Mac here in my basement morning, noon & night. If he didn't have to eat, poop & go to school, I'm sure he'd be playing 24/7 until he died in his seat, like that Korean gamer last summer.

I've been known to skip a few projects around the house in favor of playing Dragon Age :), but not to that extent..
 


Yeah, looks impressive, but I'm wondering about how many actual cores these estimates are based on? Note the huge jump in integer perf compared to FP, due to the full integer pipes instead of the partial 'simulated' ones that Intel uses in SMT.

Hate to admit it, but if AMD does deliver on this and Intel doesn't with SB, I'll probably jump ship and go AMD my next build. I know, Jenny & the other AMD fanbois would have a heart attack, but I've always been more agnostic than what I let on. Mainly to keep them stirred up :kaola:
 

Keeping in mind these are supposedly totally fake performance figures, but the Zambesi Bulldozer models are 8 core/4 modules and Interlagos is a 16core/8 module.
 


Oh, OK - I was thinking Interlagos was 8 cores.

But you are right of course - after that "40% improvement across a wide variety of workloads" estimation for Barfelona, it pays to take AMD's estimates with a grain of salt.
 
IIRC TH's gulftown review, crysis doesn't take advantage of more than 4 cores. You can see the architecture is much more powerfull than phenom II's stars archicetchture with that big jump in performance. You have to wonder what GPU or how many of them they are using to get those numbers.

I don't know if this can be trusted though. It seems a bit early to even have engineering samples ready. Zambelzi should be out Q1 2011, but the other chips should be even later, around Q2 of Q3.