CPU/Mobo AMD Bulldozer Discussion Thread

I had no idea AMD was so far behind in terms of IPC
Click to expand...

Actually, let's do the math so we can be sure :) even I said that off the top of my head.

We need to have a pure CPU benchmark, so it'll be synthetic. There are too many variables otherwise.

Here's one graph:

Charts, benchmarks Desktop CPU Charts 2010, ALU Performance: SiSoftware Sandra 2010 Pro (ALU)

The i5/2500 is 85.30/3.3GHz and the x4/970 is 54.10/3.5GHz.

Let's normalise to 3GHz, the figures are 77.5/46.3.

That's 59% IPC, close enough to what I estimated it was. Lucky me :)

Assuming they get a 10% improvement in IPC, you're looking at a final figure of close to 70% of Intel's IPC. I'm not sure BD will make more headway than that.

Edit: Of course there's more to life than IPC. TCO, power efficiency and core count all add up rapidly. So it's not really the end of the road for AMD, it's not going to be a quantum jump as far as anyone can see.

We could all be proven wrong (one way or the other) and now only time will tell. If they hit just below the level of Intel's IPC and have more cores to bridge the gap, manage smart core switching and retain compatibility with older motherboards to some extent, you could actually end up with a decent seller.
 
the cores share resources but the resources aren't tidbits.if considered properly we find that there is about 1Mb L2 for each core and 8Mb L3 to be shared with all the modules.but the sharing isn't so linear and i think it will be more intelligent,like the modules getting more load getting more resources etc.although i am still confused how their version of the turbo boost will work.no amount of cores can compete against intel if the per cycle performance doesn't increase.BD's future in this aspect doesn't look very appealing although the no. of cores may seem lucrative.the only scenario which is in probable favor of AMD is when apps begin using full 8 cores and then probably will BD's sail past sandy as the hyper threading won't prove as efficient as real physical cores,how do you fill the execution slots that efficiently?you don't.But Ivy and sandy bridge E would probably close this gap too,and this is only considering when apps use full 8 cores.as intel follows a tick-tock strategy with each tick referring to a die shrink and a tock referring to a new microarchitecture we would see Ivy as a tick and haswell as a tock,then braodwell as a tick and skylark as a tock.i doubt if AMD will be able to compete with haswell although 2nd generation BD could prove to be good performers.on seeing the gap between the Phenom II and sandy,BD really seems like a product to decrease the gap rather than made to compete which is reflected in it's probable lower pricing.i think it will be more of a sell rather than a sale.too bad really.
 
Considering the pathetic performance for my two magny-cours, anything would be an improvement. Only problem is, who would want my two 6128s once I opt to go for the Interlagos procs. Not in TE market for sure. No one would want to buy them :(

Still waiting for the Bulldozers. Let us hope it lives up to the expectations that have been built up and a fraction of the hype generated so far.
 
OMG to an untrained eye like mine, the charts are just awesome :D.

Waiting for Pros to shed some light on those benchmarks :)

but yet again its just too late :( hope AMD stays alive !
 
if first gen can match the 2nd gen sandy bridge then there are good chances to have 2nd gen fx processors to be @ equal or better of next gen of intel :D
 
Back
Top