Bulldozers first screens

[Shadov]

I expect Bulldozer to be on par of SB or slightly below in single threaded applications, while it might excel at multi-threaded workloads. If you consider that fanboy talk is your choice.

Bulldozer will fit into AM3/AM3+, while Intel will be cashing in on different sockets (same crippled chipset). No matter how many customers will see the light when it comes to motherboards - we will have competition again to lower prices especially at the (higher) end of the scale.

[Manicdan]

Why does everyone put so much worth in SuperPi? I mean for crying out loud, its not like it's serving you cheerios in the morning upon rising

Not Good at SuperPi makes for a terrible chip????? Funny I am typing on my 3 year old Phenom 940 BE at STOCK and this processor handles anything you throw at it just fine!

superpi is the dead horse of a benchmark we all like to keep beating at. i hope you saw the sarcasm in my post

[sdsdv10]

Intel is worried, they are selling unlocked cpu's at $220. They have never done that.

The K series have been around for about a year and where a reaction to the Amd Black series so don't think that this as a worry about BD but a hedge for the PII unlocked.

What qurious63ss said. The first K series (the 655K and 875K) were both launched in 2Q'10, I believe primarily as a response to the plethora of AMD Black CPU's, with the 655K launch price ~$220 ($216per unit in lots of 1000). You can see the explanation of this and the launch price in this review from last year. http://techgage.com/article/intels_c...ed_processors/
I don't believe this has anything to do with BD or SB.

http://ark.intel.com/Product.aspx?id=48499
http://ark.intel.com/Product.aspx?id=48750

[XRL8]

I am not expecting something disastrous. Phenom II was definitely not a disaster, It help AMD catch up alot.

I think it is easier to expect something like phenom II type improvement than expect a crazy improvement. How is expecting Phenom II performance increase vs phenom Intel fanboyish. It is pretty realistic.

I think expecting the second coming of athlon is fanboying and too optimistic. They have so much ground to make up at this point. As for some people not thinking thinking it will match or beat ivy bridge. Look at some people signatures or avatar. E.g Intel is about to get athloned. Phenom II clock for clock,clock for clock is a bit worse than core 2 quad(very slight). So they need to make up for two or three generations. If they can match sandy bridge that would be great but this is a huge endeavor. New Architech dont always pay off. Look at phenom and to a lesser extent the 6970 and Fermi. AMD's track record for CPU's isn't the best lately so its hard to expect a blockbuster.

A Intel fanboy will say BD will disappoint like Phenom I because they are not saying anything and keeping everything under wraps.

You see, phenom I vs phenom II AT SAME CLOCKS AND CORES wasnt that much of a improvement really.What it did bring to the table tho, is clocks, much more of it.And later, cores, 50% more of it.So if you take a look at it YOUR way, its a small decent improvement.But from where im sitting, its pretty much a big one (add higher clocks ,more cores,ddr3,higher attanaible NB speed).
However going back to your first post, you were clearly making it to be a little better than phenom I .Not my choice of words .
Getting back to the point, JF told us that BD will have HIGHER IPC than PHII.
We know from lotsa different sources that beside that it will have:
-More cores
-New instructions
-Higher memory throughput
-Higher clocks
-Big turbo boost

So if youre trying to tell me, that this cpu is going to be just a "bump" from PHII.Thats very unrealistic.Keep in mind tho, im not talking bout the superpi.
Because its quite concevaible that its gonna be just 10% faster at that.Thats possible.Im not working with or playing it tho.
As for the "expecting too much" ,youre saying people here have unrealistic expectations for BD.Well 20% higher IPC on ivybridge is hard to believe also .
Mayeb i should clarify as to what i would consider big upgrade:
-Up to 5ghz overclockability
-10-15% ipc gain
-Core gain (thats a given right now) which gives me 33% more power vs my current X6
-New instruction sets which give me additional boosts here and there (transcoding, pcsx2 etc).

+High speed DDR3

Im expecting all this to give me 40-50% more cpu raw power than my current setup.If its not going to deliver that, im waiting for socket 2011 and examining it.

[BeepBeep2]

You see, phenom I vs phenom II AT SAME CLOCKS AND CORES wasnt that much of a improvement really.What it did bring to the table tho, is clocks, much more of it.And later, cores, 50% more of it.So if you take a look at it YOUR way, its a small decent improvement.But from where im sitting, its pretty much a big one (add higher clocks ,more cores,ddr3,higher attanaible NB speed).

Phenom II at release was DDR2 only AM2+, we knew AM3 was coming but there was a small wait... some CPU's struggled above 3.6-3.7 Ghz and 2400-2600NB and also had trouble around DDR3-1600.
Same design, AMD managed to push in 2 cores, run speeds 400-600 Mhz higher on both CPU and CPUNB, and gain about 400 Mhz DDR3 with just a stepping change or two...

Bad thing is that Phenom II X6 is slower IPC than Phenom II X4 but hits higher clock speeds...

[w0mbat]

x6 has exactly the same ipc as x4.

[crazydiamond]

cache.. less l3 per core

[RussC]

BD will be 40-50% higher speed than and x6 1100? Mmm, while that would be real nice, I dont' believe thats the case.

RussC

[BeepBeep2]

BD will be 40-50% higher speed than and x6 1100? Mmm, while that would be real nice, I dont' believe thats the case.

RussC

Add two cores. That's 33% right there, so you are saying you dont think they could find 7% either in clock headroom or CPC?
You've got to be kidding me...
8 Cores at 3.5 Ghz /3.9 or 4.0 Turbo is 46% faster at the same efficiency of Phenom II X6...

x6 has exactly the same ipc as x4.

This is not true

[mAJORD]

given little Bobcat is only a few % slower than K10 in superpi, maybe BD won't be so bad at it any more either :p

..not that it matters to anyone with a clue.


Anyhow, I believe the only thing that will spoil Bulldozers launch would be SB-E hex core, OR a failure to reach the frequencies we're assuming it will launch at.

A 4M / 8core BD should dispatch a 2600K enough in MT apps to 'crown' it as faster. IF the target frequencies are met, in particular the 4+Ghz Maximum Turbo (as per the leaked doc) I'd also expect it to be a competitor for the 990X, both MT and Single threaded.

[Chumbucket843]

Add two cores. That's 33% right there, so you are saying you dont think they could find 7% either in clock headroom or CPC?
You've got to be kidding me...
8 Cores at 3.5 Ghz /3.9 or 4.0 Turbo is 46% faster at the same efficiency of Phenom II X6...

who said anything about linear scaling?

given little Bobcat is only a few % slower than K10 in superpi, maybe BD won't be so bad at it any more either :p

..not that it matters to anyone with a clue.

i still run applications along with a lot of other people on this forum that require x87. go down to the WCG section for example.

essentially superpi is slow single threaded matrix multiplication, which is quite useful for a lot of things. an upgrade from superpi would be something like LinX although i cant find any reviews that anything like it.

edit: here are some benches. similar outcomes for the euler3d app along with superpi, myrimatch, etc show that AMD has a weak spot for scientific computations.http://techreport.com/articles.x/20188/15

[jimbo75]

Be realistic people. Something like 5 weeks from release and theres still no performance numbers for Bulldozer, don't ignore whats staring you in the face and thats BD won't be the miracle chip people are praying for like A64 was.

You could consider that it might be in the OEM's best interests not to leak benchmarks. You know, they have the same partners who still have SB systems to sell.

[ichime]

All I care about is that Bulldozer better Bulldoze, or it's a bust.

[Dimitriman]

I don't think anyone is expecting for something as disastrous as the original Phenom, I think people are expecting something more akin to something like Phenom II which was decent for a company with the resources AMD has. I think with all the framework with current CPU's, it is hard to gain single core performance because the biggest improvements have already been implemented. Sandy bridge for example isn't as impressive an improvement clock for clock as nehalem and nehalem was as impressive as core 2 duo vs P4 architecture.

I think expecting something that's going to beat for example ivy bridge for example or even match it clock for clock core for core is super unlikely. It would be the biggest improvement of all time, bigger than core 2 duo vs p4. Ivy bridge is supposed to be 20 percent faster and with sandy bridges lead over phenom II it would have to be a monster gain that is pretty much impossible.

I used to love AMD for their processors but they have continued to disappoint over and over again. Their overclocking ability is less consistent than Intels which make them less fun. Unless your working with LN2 i guess or liquid helium with the most recent processors. They are so much slower clock for clock too. So that why I wouldn't set my expectation too high for bulldozer as well as a lack of anything to show. Intel shows things months ahead of time because they are confidence about their processors, AMD just seems less confident.

I don't want to compare bulldozer to ivy bridge yet because ivy is still not existant. But I honestly just find it hard to believe how the 2600k would trash BD like so many predict. Sure we can talk single threaded performance all day, but how many games and apps still rely only on 1 thread? Plus let's not forget one crutial detail - Zambezi will pack 8 physical cores at 3+ ghz stock. I don't see how 2600k will be better with half the cores at anything other than single threaded. Sure the 6 cores will rock. But discarding BD vs the 4 cores i7 seems a premature decision to me.

[Motiv]

I don't want to compare bulldozer to ivy bridge yet because ivy is still not existant. But I honestly just find it hard to believe how the 2600k would trash BD like so many predict. Sure we can talk single threaded performance all day, but how many games and apps still rely only on 1 thread? Plus let's not forget one crutial detail - Zambezi will pack 8 physical cores at 3+ ghz stock. I don't see how 2600k will be better with half the cores at anything other than single threaded. Sure the 6 cores will rock. But discarding BD vs the 4 cores i7 seems a premature decision to me.

Sorry I must disagree. Unless AMD pulls something out of the bag, this chip will be priced high in the server market, while low in the desktop market (relatively speaking).

Single threaded performance is very important, as core counts of 8 will not be utilised in the desktop arena. Unless you suddenly think gaming systems will stop being console ports, or an insane amount of grid computer users suddenly appear.

I doubt very much if AMD is top of the tree when it's out in the desktop area, the server market is a completely different situation.

[pipoo]

I dont like this AMD' silence. I dont know if its just me... that's why im not putting to much expectation on this.

[tbone8ty]

damn it this thread is like facebook i keep checking it but nothing useful is on here

[slaveondope]

damn it this thread is like facebook i keep checking it but nothing useful is on here

And done

[MrMojoZ]

I dont like this AMD' silence. I dont know if its just me... that's why im not putting to much expectation on this.

People will see what they want to see.

[KiSUAN]

People will see what they want to see.

[OCX600RR]

Lol

[STEvil]

Cant sleep, CPU's will burn me?

[Micutzu]

Add two cores. That's 33% right there,
...

More like 0% in desktop-specific applications, or even -1..3% because of architecture "tweaks" in order to accomodate more cores, the same way Thuban is a bit slower per clock than Deneb.

On server parts sure, Bulldozer will be very nice, up to 16 cores per socket in a decent TDP looks like a winner, but on desktop i doubt it will even get close to SB, unfortunately.

[-Boris-]

More like 0% in desktop-specific applications, or even -1..3% because of architecture "tweaks" in order to accomodate more cores, the same way Thuban is a bit slower per clock than Deneb.

Proof?

[Micutzu]

We did a quick test on this, you can find it here, but i'm sure there are more out there.