Discussion in 'Article Discussion' started by Lizard, 12 Oct 2011.
Spend an absolute stonking fortune on hardware just to show that bulldozer can compete as a gaming chip??
I've got an AMD3+ set up with some serious water cooling. I love to get my hands on Bulldozer just to see how far I could take it. My 1100T peaks at 41 degrees @4.1ghz on prime 95 with all the fans set to minimum. So a good Bulldozer chip could go beyond 5GHz easily. The results could prove to be quite interesting.
But that said, unless some one gives me one or I have a sudden influx of cash I don't think I'll be buying one.
As always though, how newer game engines are becoming more threaded etc is all fine and good, but the reality is that what matters is the performance in today's applicaitons, and in most BD doesn't deliver.
They really should've marketed it as a 4C/8T chip - then the expectations would be different I assume.
And what is the point of 5GHz with Bulldozer, when your 1100T@4.1GHz will still beat it ?
I think you're very right about this, as a quad people would probably have looked at it and said it's landed in the price bracket between the 2500k and the 2600k and the performance reflects this. But from a octocore I think we all expected so much more.
Yeah, fair point, but I just want to overclock it till something breaks...lol
Whats worse though, the fact that they marketed it as an octocore OR the fact that AMD had to make an octocore to try and keep up but still failled?
Not trying to be funny, genuinly would like others opinion.
You know what BD reminds me of? The original Fermi reviews. Power hungry, massive die, a huge transistor count, but ultimately a promising architecture that requires refinement. Everyone rated the 5870 as equivalent/better than the GTX480. Yet after a while, the GTX480 was seen as performance king, as it had massive overclock potential in the right situations, and people figured out how to get the best from it.
This doesn't quite translate to BD, but I suspect the promise is there. As with any brand new architecture there are issues (obviously), but as stuff becomes more threaded, and if programs can be written that make advantage of BD's features, it could be seriously quick. Like in some situations Deneb was faster than i7 920s et al, but was generally slower.
I suspect they've got themselves another niche market here...
Yeah good point - only I still believe that the gtx 480 was a turd all its life, it's the 580 that 'fixed' the arch there.
Hopefully piledriver can bring single thread performance up a little. Sadly, I have little hope that they can reach deneb's level of IPC, it's simply a too big a difference.
So, amd, change the packaging to AMD FX-4250, and say it's a quad core. Ppl will like it then
The difference is that the 480 was much better than their previous line-up, despite being hot and power hungry...
These BD fail miserably because they can't even make a clear win against their old architecture. They bring nothing new to the table.
actually, Bulldozer reminds me of Intel Pentium 4. massive marketing campaign, hot, power hungry, bad clock-for-clock performance, bad architecture overall.
it really is just a quad core processor with slightly more advanced hyperthreading. it really is!
Was waiting for BD all summer. Now have just bought i7 2600k + P67 board. Bye Bye AMD was good using Phenom the last few years...
BD doesn't scale at all with memory speeds unlike Llano,
And for those hoping for Piledriver which is supposed to come in 2012... if it comes at the start of 2012 it will have go to against Ivy Bridge, if it comes towards the end of 2012 only then it will have to face Haswell, no one will care anymore if Piledriver might manage to catch up to the Phenom II x6 at the end of 2012.
I think a lot of people feel like that. I know the Prescott comparison crossed my mind more than once.
What AMD planned to do is maintain IPC and increase frequency to increase overall performance. What they actually seem to have done is (accidentally) reduce IPC and increase frequency thereby maintaining overall performance.
Such a shame. We can only hope that next time they'll try and massively increase IPC to catch up with Intel.
bulldog, i am finding your comments about bd to be quite vexatious...please discuss this cpu in street-wise manner...
still find nothing out there that tells me this chip is $30 better than a 2500k...
OMG. I thought we had heard the last of Chap....lol
Trig, you have been awarded rep in a non vexatious street wise manner
So it is a server chip that looks like it, MIGHT, beat out some of the intel server chips in some common server workloads, and maybe even by a fair amount and will likely be priced lower than the Intel server chips. However, as a desktop chip it doesn't beat out the 2500k or 2600k and probably doesn't even beat out a 2400 in most COMMON workloads a workstation and home computer would see. It would win in a few workloads, but not many. Its single and light threaded performance is abysmal, and that is still what a fair number of workloads are.
AMD stated that a goal was to hold the line on IPC and get improvements through parallelization and clock speed. They got extra clock speed, but they seem to have lost 10-20% in IPC over the old Phenoms. That is frankly embaressing.
Their stated goal of 10-15% IPC improvements per year from here on out doesn't seem like enough. Intel is looking at probably a 8-12% improvement in Ivy Bridge, let alone that clock rates are likely to be able to go significantly, significantly higher and power draw is going to go down a lot. Bulldozer is actually worse in power draw, and possibly not even as good in overall efficiency compared to phenom in a lot workloads.
Again embaressing and potentially fatal in a server environment where power draw and efficiency are very critical.
They should've done a sort of pipe cleaner like they did for the R8xx series.
Ah well, hopefully they'll be able to make something out of it.
Separate names with a comma.