haha
I've built over 20 PCs, half of which were AMD based. They were all sluggish unfortunately, so not "fanboyism", just bitter experience.
PEBKAC.
-
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 Says Bulldozer Is 50% Faster than Core i7
- Thread starter opethdisciple
- Start date
PEBKAC.
That is a clear mistake by AMD to not have AM3+ boards out by now, we are only what 2-3 months away from BD launch?
LOL for sure.
Having some out now would be a good idea.
Yet diehards should know there's always something else faster around the corner after you've bought it.
Oh very true, but if you can afford to wait you wait.
Besides, the issue is not so much the CPU but the whole new socket. I don't mind upgrading CPU now and then, but to have to upgrade motherboard as well sucks.
I'm one of these people who likes to have a motherboard that can be kept for a while
OEMs come first because a.) they are 80%+ of all of the shipments and b.) they take twice as long to do their validation and testing.
OEMs care because they build systems, put them in distribution and then distribution moves them to retail where customers buy them. When benchmarks come out that make people reconsider their purchases and stall sales, it impacts the millions of systems already built and sitting in the channels. Weeks or months worth of inventory.
There is always something right around the corner for any company. But the product "right around the corner" becomes infinitely more real when an actual benchmark comes out. If you don't know whether something is better or not you might be inclined to go ahead and buy. But a benchmark adds clarity.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
Yup because seemingly rushing a mobo out with a new chipset without proper validation works soooooo well doesn't it
The simple fact is, if AM3+ boards came out tomorrow, Asus/Gigabyte/etc would be selling them for more than AM3 boards, because they'll have millions worth of boards out at OEM's, and at retailers that all need to be sold.
Yes some people would buy the new £160 Crosshair 58, with 47 pci-e slots and a hexcore Phenom, then in a few months replace it with a Bulldozer, the percentage of people that would do that is tiny, essentially insignificant to AMD's bottom lines.
But then if AMD had planned to have AM3+ available earlier it would appease enthusiasts and would swing a few sales, and gives a genuine interim option thats "good enough" with the option for potentially the best chip on the block to be added in a few months.
It does tend to work that enthusiasts have a HUGE sway on all their non techy friends purchases. Most people I know are aware of how much I know about computers and so the majority of them ask me what to get, if AMD is the best choice, a few enthusiasts can translate quickly to a lot of Dell boxes with AMD chips inside being sold.
Then again in most cases I'd still recommend AMD now as the vast majority of "normal" uses barely need a dual core and value for money is more important tha all out performance, and AMD are incredibly competitive there still.
But we KNOW that AM3+ & BD is better than AM3 and current CPUs, unless you are going to "officially" tells us that it is not
The only thing the benchmark will do is clarify where BD stands against Intel chips ... but if you are confident that BD can stack up against Intel, why not hurt them even more given the issues with SB by showing that your chips, that will be released around the same time as fixed SB boards will be just as good.
If AM3+ board launch was planned to be a fair few months ahead of BD launch then the manufacturers would have already been reducing output of AM3 and clearing that stock ready to kick off the AM3+ stuff.
It would have helped AMD to have something out there right now, to tempt people to get it and then take BD later on ... as opposed to just taking the latest Intel offering (which outperforms the AMD stuff really in raw power)
I think it was already stated on another forum (by JF-AMD if im not mistaken) that it is an improvement over the current AM3 processors in terms of core vs. core performance, the fact that Intels current line-up are superior doesn't really factor too much into it. think about it this way, Bulldozer is clearly designed to excel in heavily multi-threaded workloads, if it was 50% faster than Intels Sandy Bridge processors in that environment (which is hugely impressive considering each dual-core module only requires 13% more die-space for another complete thread) then I think Bulldozer would have served its purpose, fulfilled its designers aspirations if you will, if the Sandy Bridge was say 10% faster in single-threaded applications that would be irrelevant, due to the fact the Bulldozer, in an ever more threaded world comfortably outperforms it in the enviroment its designed to operate in. comparing them to Intels Extreme Edition is just pointless, since they are so shockingly priced, if Bulldozer performs as well as current i7s I think its been a success, only problem in my eyes is how to overturn Intels insane marketting machine?
indeed, we shall have to take the information available to us and draw a conclusion from there. been trying to sum up a general performance for a while now, based on what we see from the architectural diagrams, etc. but its not easy when we don't have an exact idea of whats even in a Bulldozer Module, the only information we really have is:
1) it has two 128-bit FPUs that can be combined to act as a single 256-bit FPU.
2) it has two 'interger cores' / module, though there is no details on exactly what in the way of ALUs, etc. we have in there?
3) shared front-end, shared L2 cache and shared (but also not) FPUs.
4) according to 'document' from ISSCC we have 213 million transistors on 32nm high-K metal-gate SOI designed to operate from 0.8V (seriously?) to 1.3V, operating at 3.5Ghz+ on a die-area of 30.9 square millimetres (including L2 cache)
so the conclusions I have drawn from this is that Bulldozer is designed as a high-frequency, multi-cored, modular and efficient architecture, emphasising multi-threaded performance and efficiency over single-threaded performance and raw power. 0.8V has to be a power saving mode and 1.3V is likely using Turbo Core, so perhaps 1.1/1.15V for standard frequency, thats pretty impressive if the stock frequency is as high as 3.5Ghz (with 0.5Ghz Turbo), would imagine with the design being geared toward reducing the number of gates/cycles we should expect quite high clock speeds to be achievable, so 4.2Ghz at 1.3V, who knows? without knowing whats actually in those Interger Cores, specifically ALUs, etc. I don't see how we can really draw any serious conclusions in that respect since some diagrams have two ALUs and two AGUs, some even have four ALU/four AGU, so would just like to assume at least as fast as the current Phenom II cores, for the sake of argument we'll say 10% faster, due to refinement. a 4.0Ghz+ processor with eight 'cores' seems like an decent proposition to me personally. as for price, I would reckon that die-space is a critical factor in pricing, for the documents we know the quad-module Bulldozer should be no bigger than current Phenom II X6 die, so I wouldn't imagine the prices would be massively higher, unless of course performance is killer!
Edit: and I wouldn't think JF-AMD would be willing to give us a 'hint' of what exactly is in each module, in terms of ALUs, etc. so I guess much is going to remain a mystery until the official unveiling, so lets say JF-AMD should we expect core vs. core performance exceeding the current Phenom II line, wouldn't ask you to reveal anything currently disclosed, but maybe a nod or a shake?
Last edited:
well personally, I am waiting to see what Bulldozer brings to the table and see how things look then!
Edit: another one for JF-AMD, by Q2 are we talking second quarter of this year, i.e. summer or second quarter of the US fiscal calendar? since thats runs till March right?
Edit: another one for JF-AMD, by Q2 are we talking second quarter of this year, i.e. summer or second quarter of the US fiscal calendar? since thats runs till March right?
- Joined
- 18 May 2010
- Posts
- 23,636
- Location
- London
A bit more detail on Bulldozer:
http://news.softpedia.com/news/AMD-Details-Bulldozer-Module-Throughput-185948.shtml
http://news.softpedia.com/news/AMD-Details-Bulldozer-Module-Throughput-185948.shtml
Nice. Lets hope AMD are on to a winner.
Associate
- Joined
- 19 Dec 2008
- Posts
- 663
Your point is far too broad and generalised. There are plenty of algorithms and tasks for which it is not possible infinitely parallelise (i.e. not all algorithms are embarrassingly parallel [yes that's the technical term!]). For such tasks, providing extra cores/CPUs doesn't help, certain tasks are simply linear.
For most large applications there are _often_ sufficiently many tasks occurring that could happen in parallel that more cores could be used, but it is not always possibly to provide an idealised load balance across multiple cores. For instance a single tasks which must be completed procedurally may consume all of one core, whilst all other tasks occurring may be relatively lightweight. It isn't easy or simple laziness in many cases.
Anyway, almost any program will by obvious logic only be able to break itself into a very finite number of parallel tasks at any given moment, therefore your point about "up to infinity" is wrong.
People also get a bit obsessed with threads, there are plenty of languages that can use event based programming or use message-passing, to achieve parallelism. Often these are actually far easier to develop highly parallel systems in because you can avoid many of the nasties associated with thread management (many of them surrounding locking and mutual exclusion nightmares). See: Erlang (message passing), Message-based Programming
Thanks for bringing the point up though
On topic: Hope Bulldozer is as good as it is alleged to be, this would finally ramp up the competition.
Last edited:
Yup because seemingly rushing a mobo out with a new chipset without proper validation works soooooo well doesn't it
The simple fact is, if AM3+ boards came out tomorrow, Asus/Gigabyte/etc would be selling them for more than AM3 boards, because they'll have millions worth of boards out at OEM's, and at retailers that all need to be sold.
Yes some people would buy the new £160 Crosshair 58, with 47 pci-e slots and a hexcore Phenom, then in a few months replace it with a Bulldozer, the percentage of people that would do that is tiny, essentially insignificant to AMD's bottom lines.
But then if AMD had planned to have AM3+ available earlier it would appease enthusiasts and would swing a few sales, and gives a genuine interim option thats "good enough" with the option for potentially the best chip on the block to be added in a few months.
It does tend to work that enthusiasts have a HUGE sway on all their non techy friends purchases. Most people I know are aware of how much I know about computers and so the majority of them ask me what to get, if AMD is the best choice, a few enthusiasts can translate quickly to a lot of Dell boxes with AMD chips inside being sold.
Then again in most cases I'd still recommend AMD now as the vast majority of "normal" uses barely need a dual core and value for money is more important tha all out performance, and AMD are incredibly competitive there still.
+1
Let's hope that AMD can deliver, as it has been said throughout it would help the competition and hopefully lower prices. Sadly I have recently bought an i5-2500k, so maybe in a few years I'll give AMD a shot. Anyone know (sorry if it has been asked/answered before) what the price range of these will be?
Pv123
Pv123