News AMD Bulldozer B3 Stepping Due

I hope so - seems like whenever the quarterly numbers come out, it's always AMD losing money. Even if it feels a little BD was an afterthought on the desktop - "crap, you guys, Llano isn't going to cut it this time. It's been 10 years since we've had a brand new architecture!" Hence, turn BD into the new higher-end desktop processor. Now, turns out maybe that wasn't the best idea, when you see it benched beside SB. But, if they survive long enough to get PileDriver out the door, then I'll upgrade. I'll stick with my Phenom II for now.

Be nice to see them get some server marketshare back, though - that slide has been painful to watch. At work, when the IT manager asks my advice on hardware (he's almost strictly a network guy), it kills me to have to recommend Intel, but it hasn't made sense to recommend AMD for a while now, in terms of bang for the buck. Got even harder once SB got released, and the appeal of Llano is not what businesses need. Neither is BD, for office PCs. Maybe next go-round speccing new servers there might be some competition.

 

Is BD going to be a success in the server market? As I understand it performance/watt is a crucial factor in this sector and another area in which BD is soundly beaten.
Don't get me wrong, I want to see AMD being successful but BD just seems an evolutionary dead end in its current form.

 

Look - we need to know the eol for Phenom II - as far as I'm concerned for my AMD platforms my next upgrade will be hex-core Phenom II rather than Bulldozer...

If Bulldozer had been what I'd joked about - just a die shrink with more cache it probably would have preformed better in benchmarks and energy use than Bulldozer. So that's £1m you owe me AMD in consultancy fees...

 

Nice title.

I couldn't bring myself to recommend BD to anyone and when I can afford to jump to some i5 or i7 goodness I will, but you really can't help but trash talk BD can you?

The actual news itself was delivered with less negativity than Anandtech, so Ican only assume you were trying to stir the pot.

 

I may be wrong here, but it seems like AMD is no longer interested in single thread applications. Also, this is a totally new architecture, and as such will have "bugs". There are some who have argued that they should have waited until they were competitive with Intel. They are in the server arena. (for now) So the argument would be "If not now,, when??" (release of new architecture) Any new stepping will probably address the power issue, but unless it is incidental, I don't see single thread performance getting any better. The advent of the Linux/windows fixes will help some, but it is what it is, I think.

 

Hmm, not really trying to stir things up. AMD has added features via a number-update to stepping before (C1E support, for example) while Intel usually uses number updates for minor process/manufacturing tweaks. Letters tend to signify a big change, as with the G0 Core 2 Quad Q6600. I think that was worth pointing out.

I don't think the new stepping will make a big change to Bulldozer overall though. Maybe drop power consumption a bit, or improve overclocking (we could have got 5.1GHz out of our chip if Core 8 had been better, for example)

 

Piledriver is supposedly going to be nicer. If they manage 10% better IPC numbers like they are shooting for, that'll help out a bit. However, IPC seems to be down more like 20% over what Phenom managed!

Clock rates are up, and threads are up, but IPC is so far down that single threaded and low multithreader performance even on Piledriver is going to be nothing to write home about. Worse Ivy Bridge will be its competitor when it comes out, which looks like it promises 3-6% IPC improvements over Sandybridge, and might well be allowing for anything from 10-30% higher clock rates...Piledriver might well not be enough of a performance improvement to even hold the current SB vs Bulldozer relatively levels of performance, IB versus Piledriver might make AMD look even worse.

In the server market, the little data that has come out so far, AMD has not grabbed a huge market share with Bulldozer coming out. First the Opteron chips haven't hit full release to the best of my knowledge (so how can they grab a huge market share) and if you look in the super computer market, Intel for CPU shipments outstrips AMD by a huge margin. Nvidia outstrips AMD in super computer GPU sales also.

For servers, Intel has a huge market lead and Xenon SB versus Opteron Bulldozer it doesn't look like Opteron Bulldozer is likely to offer enough benifits to be worth most customers switching, though it may be an upgrade path for some existing customers already using Opteron servers.

 

What you're seeing is the problems with the lack of FMA instructions, for one, and even a way to process it in Win7. In server Linux, for example, it exists. But in heavily parallel workloads, with the proper instructions, it ought to shine. AMD brought out a processor that has new instructions that really would have a hard time being inserted into Win7, and it relies on those extra instructions to do its thing. I don't think that was the wisest thing, but it does mean there's more performance to be had.

 

I can't believe how many spelling mistakes and mistake and bad english are in this article.

 

The desktop BD models are currently a bit too hairy on the clocks and voltages. Drop clocks by 20% you might drop power consumption by 40%, something along those lines is going on with the Opteron versions.

 

I do recall seeing some early benchmarks for Linux-based systems running on BD (on Phoronix) - that were really not great. Obviously these probably weren't running specially optimised code. But even if BD has good support for FMA4 and you bother to compile stuff optimally - the releatively high power consumption is also going to be a problem for the Enterprise/HPC space...

 

As a final for me, what is annoying the heck out of me lately is seeing AMD and others (fanbois or not) talking about how, well if anything is wrong with it, its that it is before its time.

That is a serious issue! If you can't take advantage of what it CAN do well, than it is not great. A Ferrari in the era before paved roads does a fat lot of good when everyone else is driving lifted Wranglers and Land Rovers with locking differentials. What good does it do to buy one a couple of years before anyone bothers paving roads?

New instructions and optimizations for potential upcoming software and methods of execution is great, but you can't handicap how most things work today without still having a crappy product.

Windows 8 gives a small performance boost in benchmarking. I suspect it'll be a bit better once Windows 8 RC gets let out of the bag let alone the full version. However, I don't think we are going to see 50% performance improvements. BD seems to get 3-8% performance improvements...Intel SB chips were seeing 2-5% performance improvements on those same Win 7 vs Win 8 benchmarks...so BD gets a little more than SB does...but frankly not an enormous boost. Win 8 is also at least 6 months away really and probably more like 9 months.

Again, by the time Win 8 is out, Ivy Bridge will be and probably not long after IB-E as well and then Pile Driver right before Hasswell.

AMD might on their road map look to be improving their CPU performance in each iteration, but it doesn't look like it is going to improve even as fast as Intel will be, which means AMD is going to be even further behind unless they take a radical new course.

 

Considering it takes 5-6 years to brew up a new architecture, I think they'll be well on that course.

Let me see if I can find the more optimized benches.

http://www.phoronix.com/scan.php?page=article&item=amd_fx8150_bulldozer&num=1

http://www.phoronix.com/scan.php?page=article&item=amd_bulldozer_scaling&num=1

Full Linux article, and some of the real world boosts were impressive in more parallel workloads.

Add to it that the ASUS CHV is getting a lot of flak for unoptimized BIOS (as in EVERYONE beats its numbers, Biostar, ASRock, ECS) and while it's not gonna turn into a barn-burner, it's not getting a fair implementation either.

It all comes down to implementation. I'm saving my pennies for dual G34 to replace my quad Socket F system, because a pair of 24-core procs will really shine in some of the workloads I use it for. A similar Intel system would have a lot fewer cores and cost significantly more for no real benefit.

Eh, the numbers in Linux are decent, though I'd still have to flog this i7 and board to even think about changing over. But things like that become important if you have systems in the house that never see Windows.

 

Here's my point of view for what it's worth - please forgive my technical know-how, I try and put it in a way I can kinda understand:

Bulldozer right now is not an octo-core, it's not a quad core with hyperthreding either, it's a hydrid multicore CPU with advanced extension set (in my mind I think of the old 386 cpu's with a co-processor).

The problem is that there's no OS available today that recognises or optimises application to suit - apparently Windows 8 fixes 'some' of that, what AMD Actually need is optimised driver sets for other Windows platforms too...... point is the dataprocessing pipeline in unused part of the cpu must be a huge waste of steps - hence the massive performance hit in these single threaded tests.

So right now we don't know exactly what Bulldozer can do in the right environment, because the software is not around, all know is what it can do today, simply put, thats jack schit.

B3 step on exisiting architecture will be little more than optimising it's power, clocks and leakage, other changes to the cores will be the Piledriver series, which may or may not suffer the same way as the current CPU unless there's some heavy driver work to be released with it.

 

I am an amd fans boy and the reason why is simple. Intel own 95 percent of all the computer market, if things carry on as they are amd will fold and then intel can price even higher than they already are. Intel and amd are under the supervision of the us government and the competition is just ******** to keep the chinses in check

 

clock for clock the BD CPU is 10-20% slower then an Phenom II, never mind comparing it to an Intel SB

The BD will most likely be remembered as an P4 with HT (good for video stuff, heavy threaded work, slower and hotter for other stuff that most users do day to day )

I doing this on my phone (i update the hotter part later on as think the idle temps on BD is better then phenom)

 

The thing is though is that BD is it from now on from AMD. This is THE design they will be sticking with from now until kingdom come. We will just have to hope that they mould what they have now into a more cohesive whole (because at present I'm not convinced BD is anything like AMD imagined it would be).

Time will tell if they iron out the terrible L1 cache copy and write performance and pull down the L2 cache latency.

That said my next components I will buy will be a 2700k and suitable motherboard. If BD version 1 could have matched PhII for IPC I would have bought it. But sigh it didn't happen.

 

intel and microsoft are best buds, amd gets the leftovers, if I were at amd, I would recomend them attempting to corner the integrated graphics sector, those apu's are sexy, and combined with a discrete graphics chip, powerfull enough to run demanding games and apps, intel has nothing to compete with those apu's, so amd win there, but fail with bulldozer, which is like a big fat american muscle car, I would love to see amd release an apu powered tablet, now that would be interesting.....

 

Still advocating for AMD processors, especially the new ones, are a bit like the people in the 1920´s, that - equally unreal - kept on advocating for gaslight instead of electric.

 

AMD FX – Series B3 revision is more than just a basic stepping:
I have direct knowledge of a possible B3 revision for the AMD FX line of CPUs. I cannot disclose performance projections at this time, but be assured AMDs processor division is working vigorously on a (B3) stepping revision with minor architectural tweaks. The base architecture will not be changed at this time.

See Other thoughts below.
Slower than the Phenom II X6 CPUs in a Clock for clock review(s). B3 revision coming to resolve this issue among a few more.

Other Thoughts: Within the B3 stepping revision, expect minor tweaks to the following:
1) L1, L2 and L3 latencies
2) Cache Thrashing Issues
3) Modified Algorithms for Branch Prediction
4) Healthy Bump in Processor Frequency
5) Slight Frequency increase via NB Controller
6) “Total Intelligent Controlâ€￾ For example programs and applications should look at the module design approach and the ability for the processor to intelligently turn off and/or turn on specific cores that it believes is hindering performance for maximum performance. (May be for Socket FM2, not sure at this time).
7) Power will be improved but not my much. We will have to wait for Socket FM2 or a future B4 revision for the AM3+ platform for better power efficiency especially when Over-clocked.

By: ATInsider
http://www.newegg.ca/Product/Produc...rMark=&IsFeedbackTab=true&Keywords=(keywords)