-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD Piledriver, finally coming soon!
- Thread starter littlepuppy
- Start date
Associate
- Joined
- 10 Sep 2006
- Posts
- 1,907
- Location
- UK
Really, anyone who has an ounce of common sense would know that the Bulldozer architecture is solid, and future revisions and die shrinks will show that. A patchy launch was almost inevitable for this kind of design, when you take all things into account.
Associate
- Joined
- 5 Nov 2005
- Posts
- 2,242
So when's piledriver coming then?
Exactly - Vista was the same when it released, after it matured a bit it was a nice, solid OS.
I posted this in another thread but figured i'd repost it here as it seems more relevant:
This is the link Mollari posted: http://forums.extremeoverclocking.com/showthread.php?t=358508&page=3
A better process with less leakage should help bring down power consumption by quite a lot (just look at GTX480/580).
So yh i'm expecting piledriver to easily beat ivy bridge quad core i7s in highly threaded tasks (hell bulldozer beats a 2600k in some tasks). And in lightly threaded tasks it almost definitely wont be as quick, but should be close enough (unlike bulldozer which is generally miles behind).
And regarding the architecture, its actually not bad, just let down by software, process and the fact its gen 1, so was always going to have some teething problems.
From what i've heard piledriver will still be up to 8 cores, with a 3-5% ipc improvement and a 7-10% clock speed improvement, as well as lower power consumption across all uses. Combine that with improved software (interesting part of toms hardware review: http://www.tomshardware.com/reviews/fx-8150-zambezi-bulldozer-990fx,3043-23.html) and piledriver looks quite good. Best case scenario would be 5% ipc improvement, 10% clock speed improvement, and another 15% from better optimised software. So in a best case scenario piledriver could be 25% (edit: maths fail, make that 30%) better than current bulldozer performance, worst case about 10%. Doesn't sound bad
Edit 2: the 15% best case software improvement would apply to current chips as well, so bulldozer could be a sound buy for anyone expecting to keep it for a few years or more.
Edit 3: Mollari's link below shows some tests getting a 5% or sometimes even more performance increase just from a different mobo. If I assume thats because mobo manufacturers haven't completely figured out how to get the best from bulldozer, but probably will have by the time piledriver is out (fairly big assumption I know), that bumps up the best case performance increase to 35%. Now that is definitely not bad.
This is the link Mollari posted: http://forums.extremeoverclocking.com/showthread.php?t=358508&page=3
A better process with less leakage should help bring down power consumption by quite a lot (just look at GTX480/580).
So yh i'm expecting piledriver to easily beat ivy bridge quad core i7s in highly threaded tasks (hell bulldozer beats a 2600k in some tasks). And in lightly threaded tasks it almost definitely wont be as quick, but should be close enough (unlike bulldozer which is generally miles behind).
And regarding the architecture, its actually not bad, just let down by software, process and the fact its gen 1, so was always going to have some teething problems.
First there was BullDozer. That turned out to be BullS..
Then there's PileDriver. That might turn out to be a Pile of S..
Then there'll be Steamroller. That'll probably turn out to be a Steaming Pile of S..
Then there'll be Excavator. That'll possibly turn out to be complete Excre..
Well, then again, they might be awesome
Then there's PileDriver. That might turn out to be a Pile of S..
Then there'll be Steamroller. That'll probably turn out to be a Steaming Pile of S..
Then there'll be Excavator. That'll possibly turn out to be complete Excre..
Well, then again, they might be awesome
Bulldozer isn't terrible! its just a very immature architecture, from an architectural point of view it is still a pretty decent design, performance will come to the design rather than the other way around, as applications get more threaded the attributes of the design will be shown. sure it isn't great in single threaded applications, but it was never ever going to be, that was never the point of the exercise.
honestly, all they need to do is get the 32NM process sorted out as soon as possible so the thing can run at the frequencies it was supposed to be running at (lets not forget its running slower than intended!), the temperatures will stabilise and overclocking will improve a fair amount, since in the applications its designed to do it performs well, would imagine in multi-threaded workloads it'll be untouchable in the price bracket. watch this space I would say...
honestly, all they need to do is get the 32NM process sorted out as soon as possible so the thing can run at the frequencies it was supposed to be running at (lets not forget its running slower than intended!), the temperatures will stabilise and overclocking will improve a fair amount, since in the applications its designed to do it performs well, would imagine in multi-threaded workloads it'll be untouchable in the price bracket. watch this space I would say...
Bulldozer isn't terrible! its just a very immature architecture, from an architectural point of view it is still a pretty decent design, performance will come to the design rather than the other way around, as applications get more threaded the attributes of the design will be shown. sure it isn't great in single threaded applications, but it was never ever going to be, that was never the point of the exercise.
honestly, all they need to do is get the 32NM process sorted out as soon as possible so the thing can run at the frequencies it was supposed to be running at (lets not forget its running slower than intended!), the temperatures will stabilise and overclocking will improve a fair amount, since in the applications its designed to do it performs well, would imagine in multi-threaded workloads it'll be untouchable in the price bracket. watch this space I would say...
I hope they sort it. I'm bored of Intel and wanna change next time round. I never thought my i7 920 would last this long.
30 days!
lol burell 
Whilst I am just as dissapointed as most of you, theres a couple of points to bear in mind here. AMD are trying to 'think ahead'. Maybe too far ahead! - Theres obviously issues that need resolving:
A)- New Architecture - new process etc, will need to be refined
B)- Windows 7 scheduler - Will be addressed by Windows 8
C)- More software developers stop being lazy and using the multi-threads that we all have in our CPU's.
This isn't realisticly going to happen until next year - Ivybridge isn't going to be the bee all and end all that most people are making it out to be - its only a 10% improvement tops! A tock in the tick-tock scenario. Intel can have the high-end tbh - its not that big a market in reality but you should all think twice about how Intel is 'controlling the cards'. OEM's are being 'bribed' into using Intel stuff in the mid-market - which we are all aware of (think anti-monopoly cases worldwide against intel), so AMD will find it hard to flourish. Its a shame in the short term they also took the 'bribe' from intel in the US and settled out of court - its going to damage them in the long run tbh
I'm still going to support AMD as I believe without competition Intel will stuff us all - simple as. Problems have to be expected with new architecture - Manage your expectations, BD was not designed to beat SB outright, but simply compete. Figures wise it's not so bad, falling between 2500k and 2600k in most scenarios - just that darn power consumption issue which should hopefully be addressed soon..... Sure AMD will have had a hard enough kick up their rears from us all
Whilst I am just as dissapointed as most of you, theres a couple of points to bear in mind here. AMD are trying to 'think ahead'. Maybe too far ahead! - Theres obviously issues that need resolving:
A)- New Architecture - new process etc, will need to be refined
B)- Windows 7 scheduler - Will be addressed by Windows 8
C)- More software developers stop being lazy and using the multi-threads that we all have in our CPU's.
This isn't realisticly going to happen until next year - Ivybridge isn't going to be the bee all and end all that most people are making it out to be - its only a 10% improvement tops! A tock in the tick-tock scenario. Intel can have the high-end tbh - its not that big a market in reality but you should all think twice about how Intel is 'controlling the cards'. OEM's are being 'bribed' into using Intel stuff in the mid-market - which we are all aware of (think anti-monopoly cases worldwide against intel), so AMD will find it hard to flourish. Its a shame in the short term they also took the 'bribe' from intel in the US and settled out of court - its going to damage them in the long run tbh
I'm still going to support AMD as I believe without competition Intel will stuff us all - simple as. Problems have to be expected with new architecture - Manage your expectations, BD was not designed to beat SB outright, but simply compete. Figures wise it's not so bad, falling between 2500k and 2600k in most scenarios - just that darn power consumption issue which should hopefully be addressed soon..... Sure AMD will have had a hard enough kick up their rears from us all
Last edited:
Having seen some of the power consumption figures for the server parts (starting at 35W, which is good even at 2.x GHz) the high power consumption, and consequently poor upper clocks, do seem to stink of problems in the manufacturing. Feels like the chips that are sent for server use are the ones that actually come out right - everything that makes it to a desktop is thus a "factory second," or at best something that fails to meet expectations
But there are still major errors in how they've marketed it... AMD are trying to sell what they hoped for instead of what they're actually making. I can't imagine pricing a Micra based on how good a 350Z is
Having seen some of the power consumption figures for the server parts (starting at 35W, which is good even at 2.x GHz) the high power consumption, and consequently poor upper clocks, do seem to stink of problems in the manufacturing. Feels like the chips that are sent for server use are the ones that actually come out right - everything that makes it to a desktop is thus a "factory second," or at best something that fails to meet expectations
But there are still major errors in how they've marketed it... AMD are trying to sell what they hoped for instead of what they're actually making. I can't imagine pricing a Micra based on how good a 350Z is
well since Bulldozer is designed to make an impact in the server market (where it will do real well I would guess) it makes sense unfortunately for them to keep their Opteron chips the best silicon, what they have been doing for years though because servers need to be cool and reliable. hopefully as time goes on an they sort out the fabrication problems these cooler and less power hungry chips will filter through to the desktop lines.
for me personally the manufacturing process is a weak link as far as Bulldozer goes, hell even Llano suffers from dodgy manufacture, based on how Phenoms can get upto 4.4GHZ sometimes on 45NM, would expect a 32NM Llano processor to reach much higher frequencies however we aren't seeing that happen at the moment.
as far as marketing goes, Bulldozer is a good thing to market, it has a couple of points that make it appealing to those who don't spend a lot of time reading up on such things so I would expect them to shift quite a lot of them. also under the hood it is still a powerful architecture that does well in a number of examples, don't think we've seen the best of it yet.
so considering Bulldozer has came out, its had pretty mixed reviews and has performed well in some benchmarks, bad in others, performs worse on ASUS boards at the moment (which is what all the reviewers use!) I think they are alright in the desktop market, Llano is still doing well in the lower end market and Zacate in notebooks is still the best choice, so as much doom and gloom that are flying around not sure how so many people are still so ignorant to the fact AMD have set themselves a nice foundation to build on, to take the fight to Intel in the coming years.
Edit: also if anyone ever managed to make that whole concept of reverse Hyper-Threading work, Bulldozer would be a monstrous processor in that sort of an environment, loads of cores with pretty good multi-threaded performance and multiple integer cores working in unison in single-threaded workloads, not a likely happening but hell would be interesting to see. can anyone explain why that concept doesn't work? is there any particular reason that something cannot be coded to work in that way, where one thread is fragmented to run across numerous cores? would love to hear opinions on that from anyone who is 'in the know' about that sort of thing!
Last edited:
Bulldozer isn't terrible! its just a very immature architecture, from an architectural point of view it is still a pretty decent design, performance will come to the design rather than the other way around, as applications get more threaded the attributes of the design will be shown. sure it isn't great in single threaded applications, but it was never ever going to be, that was never the point of the exercise.
honestly, all they need to do is get the 32NM process sorted out as soon as possible so the thing can run at the frequencies it was supposed to be running at (lets not forget its running slower than intended!), the temperatures will stabilise and overclocking will improve a fair amount, since in the applications its designed to do it performs well, would imagine in multi-threaded workloads it'll be untouchable in the price bracket. watch this space I would say...
There will be some improvement, sure, but I don't really buy this argument. All the really good architectural releases were at least good right from the start.
Last edited:
Um, bulldozer is really good, it's just being held back by other factors.
An immature 32nm process is greatly increasing power consumption and reducing clock speeds by a few hundred Mhz (some are speculating by as much as 600Mhz, but it's hard to guess without more details from AMD).
Problems with the ASUS Crosshair V motherboards most reviewers are using are also effecting performance, reducing it by about 5-10%.
Then there's Windows' scheduler, which is reducing performance in lightly threaded tasks by 5-10%, and increasing power consumption (Windows 8 will fix this, but it's a while off).
Then there's individual applications which are not optimised for Bulldozer, and reducing performance again (I don't have any figures for this, but I wouldn't be surprised if some apps can get a 5% boost).
So thats a fairly big boost for Bulldozer, Piledriver should (if the road maps are correct) be 10-15% faster still just from architectural changes.
The thing is people see Bulldozer at/near the bottom of benchmarks, and assume the architecture is ****. It's not, it's just not being used to anywhere near it's full potential.
Last edited:
Bulldozer isn't terrible! its just a very immature architecture, from an architectural point of view it is still a pretty decent design, performance will come to the design rather than the other way around, as applications get more threaded the attributes of the design will be shown. sure it isn't great in single threaded applications, but it was never ever going to be, that was never the point of the exercise.
honestly, all they need to do is get the 32NM process sorted out as soon as possible so the thing can run at the frequencies it was supposed to be running at (lets not forget its running slower than intended!), the temperatures will stabilise and overclocking will improve a fair amount, since in the applications its designed to do it performs well, would imagine in multi-threaded workloads it'll be untouchable in the price bracket. watch this space I would say...
I guess the main problem with pinning hopes on Windows scheduler enhancements pulling Bulldozer from the clutches of failure, is that these improvements will also benefit Intel CPU's to one degree or another.
