Caporegime
Except I'm not? If Bulldozer was clocked 1GHZ higher, and overclocked 1GHZ higher than SB, it might be worthwhile, even with the low IPC.
IPC inherently improves performance across the board however.
-
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 Bulldozer Finally!
- Thread starter AceTK
- Start date
Except I'm not? If Bulldozer was clocked 1GHZ higher, and overclocked 1GHZ higher than SB, it might be worthwhile, even with the low IPC.
IPC inherently improves performance across the board however.
Caporegime
and won't ever be.Indeed, but it's not.
Caporegime
And I was people would stop bringing up server stuff, this isn't the server variant, yes, it might make a fine server chip, but what good is that in desktop space?
The chip is around 20% away from being a decent chip at the £200 mark. Chill with the hyperbole my man.
To remind people 2500K is £180 and 2600K is £240, so if AMD can bring a chip that is around 2500K performance in some scenarios and exceeds 2600K in others whats the problem for £200? Why does AMD need to beat Intels higher priced chips? The market for £250 CPUs is relatively small, whereas AMD's design, once perfected will suit the lower / mid end of the market perfectly and should scale nicely once they need to start throwing more cores at the design to up performance, across their entire range.
The internet tinkerers are already finding decent performance gains from software alone so hopefully AMD can bring it all together to provide a decent package with hardware tweaks too.
Caporegime
The 2500k often goes on deal for about 150-160.
a lot of the SB prices are inflated, for example, the 2600k is quite a fair bit more expensive than your 240 guess =/.
AMD's Phenom II/Athlon II suited the lower/mid just fine as it was.
a lot of the SB prices are inflated, for example, the 2600k is quite a fair bit more expensive than your 240 guess =/.
AMD's Phenom II/Athlon II suited the lower/mid just fine as it was.
Last edited:
Soldato
exactly, the fact people are finding optimisations that get more performance out of the chip using just software suggests it has way more to give yet, also the fact it does so well in some ALU/FPU based tests suggests that the instructions per clock have the potential to be higher than they are at the moment. there is a lot of speculation on other sites that poor cache caused by the dodgy 32NM process, so it still has some maturing to be done, perhaps stepping B3 might help to deal with some of these ailments.
considering it is a chip that is suffering from some serious power consumption issues and potential cache issues running at clock speed lower than it was intended to, all in all I think the performance of Bulldozer shows some light at the end of the tunnel, fix the power consumption issues and I would consider giving one a go, tons of tweaks about that show performance increase leads you to think that bringing them all together could give a fairly sizeable boost. lets just wait and see...
considering it is a chip that is suffering from some serious power consumption issues and potential cache issues running at clock speed lower than it was intended to, all in all I think the performance of Bulldozer shows some light at the end of the tunnel, fix the power consumption issues and I would consider giving one a go, tons of tweaks about that show performance increase leads you to think that bringing them all together could give a fairly sizeable boost. lets just wait and see...
Caporegime
It's software that can't utilise the threads.
Pretty much in the majority of situations, only half of the integer cores will be getting used lol.
Pretty much in the majority of situations, only half of the integer cores will be getting used lol.
Do you really think AMD hit their targets with BD?. JF-AMD was given bad info. AMD aimed for slightly better IPC with higher clocks and missed both targets by some margin. That doesn't make the guy a liar- just wrong and he was big enough to come on forums and admit it.
Caporegime
You act like he never knew the situation.
He wasn't a small fry employee at all.
No, I don't think AMD hit their targets, I'm also sure they weren't hitting their targets when JF-AMD was saying what he did, ergo AMD allowed lies.
Last edited:
AMD's problems will have hit when they first attempted full production, as it would have been at that point they found out Glo Flo couldn't deliver none leaky silicon that didn't need lots of power. I don't believe at the time he posted that comment they would have been at that point, as that was over 14 months ago.
Caporegime
You're allowed to believe what you want, as will I.
But given AMD's marketing regime, them lying about their IPC isn't a massive stretch.
I'm on about the youtube video which misrepresented the 980x, and was actually a 2500k etc.
But given AMD's marketing regime, them lying about their IPC isn't a massive stretch.
I'm on about the youtube video which misrepresented the 980x, and was actually a 2500k etc.
That post was from over a year ago. The chip had just taped out and like all new CPU architectures it was probably riddled with bugs so AMD couldn't be certain on performance. I am sure the engineers were telling management, and management telling marketing, that everything would be fixed for launch which was still nearly a year away(even longer in reality).
Caporegime
That's one example, he'd been saying it consistently, with posts in 2011.
Did he go back and change it? No.
AMD misrepresented the 980x in a youtube video they made on the 12th of October of this year.
Also, some Bulldozer ES were kicking around about June etc, as we did get some "leaks", although JF-AMD put them all down, although looking back, they were pretty accurate.
AMD was going to allow the sale of 1BD sale at a competitor during a charity event some months ago. But they couldn't get it ready (Doesn't make much sense, don't try and analyse)
Did he go back and change it? No.
AMD misrepresented the 980x in a youtube video they made on the 12th of October of this year.
Also, some Bulldozer ES were kicking around about June etc, as we did get some "leaks", although JF-AMD put them all down, although looking back, they were pretty accurate.
AMD was going to allow the sale of 1BD sale at a competitor during a charity event some months ago. But they couldn't get it ready (Doesn't make much sense, don't try and analyse)
Last edited:
Soldato
I do agree Martini, don't at all think AMD hit their targets with Bulldozer, which has got to be a right kick in the nuts considering it has always been a David v. Goliath story between the two of them, where sometimes AMD manage some sucker punches, but Intel are usually quick to recover. the thing is, I still don't think JF-AMD lied as such, think AMD aimed to keep their single thread performance at the Phenom II level or slightly above but it just hasn't worked out quite like they expected.
though it is worth saying that in ALU and FPU benchmarks like Dhrystone and Whetstone the Bulldozer architecture fairs really well, usually in the same sort of region as the 2600K. notable though it also usually waltzes past the 1100T without much trouble at all, which does indicate that the ALU and FPU performance have indeed been improved at an architectural level. the problem is those two benchmarks do almost nothing to test things like branch prediction, which has been an Intel strong point for ages now, so obviously something is going wrong in some part of Bulldozer.
but tests like that for me show that removing the un-needed ALU from each core has had little impact on theoretical per core performance, it is just coming undone somewhere else.
look at the areas where Bulldozer does really well, Whetstone (FPU), Dhrystone (ALU), Sandra (Bandwidth, etc.) though it doesn't do as well as the Intel line-up in the bandwidth test it is a fair whack better than Phenom II was, there are lots of aspects where Bulldozer is a decent performer, its just when it gets to gaming and such where per core performance is needed, it should theoretically speaking be fine (as shown by the fact it convincingly beats Phenom II in ALU and FPU benchmarks) but there is some sort of stumbling block preventing it from operating at its peak, perhaps cache related who knows, but it doesn't take a genius to see that Bulldozer still has potential.
hence why I don't think JF-AMD lied, because he would have looked at theoretical benchmarks, like the mentioned ALU/FPU performance benches and thought 'nice, so we have indeed improved ALU/FPU performance across the board vs. Phenom II' its only in real world gaming (at non real world resolutions) that has shown something like a spanner in the works, lots of folks on other forums are saying it seems as though the pipelines are being starved of data, based on the randomness of the performance at the moment I for one think to some extent they could indeed have a point.
also not stating this as remotely 'fact' of the matter, not saying it will be a fantastic processor, or a terrible processor, not saying anything except opinion, just like most other people on here, also not a fanboy to either side, hell before this 1055T for the last two iterations of my system I have been Intel, hell started off with Intel and the Pentium I series, just trying to see positives of the situation rather than negatives because I want competition to return to the market and if Bulldozer flops (hasn't yet!) then that competition is effectively gone and I don't understand Intel fan or not how that is a good thing!
though it is worth saying that in ALU and FPU benchmarks like Dhrystone and Whetstone the Bulldozer architecture fairs really well, usually in the same sort of region as the 2600K. notable though it also usually waltzes past the 1100T without much trouble at all, which does indicate that the ALU and FPU performance have indeed been improved at an architectural level. the problem is those two benchmarks do almost nothing to test things like branch prediction, which has been an Intel strong point for ages now, so obviously something is going wrong in some part of Bulldozer.
but tests like that for me show that removing the un-needed ALU from each core has had little impact on theoretical per core performance, it is just coming undone somewhere else.
look at the areas where Bulldozer does really well, Whetstone (FPU), Dhrystone (ALU), Sandra (Bandwidth, etc.) though it doesn't do as well as the Intel line-up in the bandwidth test it is a fair whack better than Phenom II was, there are lots of aspects where Bulldozer is a decent performer, its just when it gets to gaming and such where per core performance is needed, it should theoretically speaking be fine (as shown by the fact it convincingly beats Phenom II in ALU and FPU benchmarks) but there is some sort of stumbling block preventing it from operating at its peak, perhaps cache related who knows, but it doesn't take a genius to see that Bulldozer still has potential.
hence why I don't think JF-AMD lied, because he would have looked at theoretical benchmarks, like the mentioned ALU/FPU performance benches and thought 'nice, so we have indeed improved ALU/FPU performance across the board vs. Phenom II' its only in real world gaming (at non real world resolutions) that has shown something like a spanner in the works, lots of folks on other forums are saying it seems as though the pipelines are being starved of data, based on the randomness of the performance at the moment I for one think to some extent they could indeed have a point.
also not stating this as remotely 'fact' of the matter, not saying it will be a fantastic processor, or a terrible processor, not saying anything except opinion, just like most other people on here, also not a fanboy to either side, hell before this 1055T for the last two iterations of my system I have been Intel, hell started off with Intel and the Pentium I series, just trying to see positives of the situation rather than negatives because I want competition to return to the market and if Bulldozer flops (hasn't yet!) then that competition is effectively gone and I don't understand Intel fan or not how that is a good thing!
Caporegime
Sorry, your posts are becoming lengthy lol.
Addressing the first part, if AMD had aimed to keep their single threaded performance at Phenom II level, to say IPC has increased would still be a lie, as AMD are bringing these parts out with massive turbo's, so single threaded performance meant what exactly? One core at its default frequency would equal the similar clocked Phenom II, or that the core with turbo boost would equal that of Phenom II performance?
To address the last parts. I too want AMD to be competetive, I've said this twice now. I want Piledriver to be to Zambezi, what Deneb was to Agena. That would make Piledriver quite mighty, but I can't see it.
As it is, in a lot of peoples eyes, the current crop of FX CPU's has flopped, they are replacing Phenom II's at the same price point, but where it matters, they're performing worse at the same money, games, encoding, etc.
Bulldozer obviously has some problems, take the Cinebench result for example, its scaling is horrendous, I feel that the CPU is just falling flat on its face in that instance. If you saw some of Gareths real world comparisons between a Phenom II 1090T at 3.6GHZ with a 2.6GHZ CPU NB, against the stock FX8, it's VERY disappointing. His Phenom II is able to catch up in the majority of programs.
Addressing the first part, if AMD had aimed to keep their single threaded performance at Phenom II level, to say IPC has increased would still be a lie, as AMD are bringing these parts out with massive turbo's, so single threaded performance meant what exactly? One core at its default frequency would equal the similar clocked Phenom II, or that the core with turbo boost would equal that of Phenom II performance?
To address the last parts. I too want AMD to be competetive, I've said this twice now. I want Piledriver to be to Zambezi, what Deneb was to Agena. That would make Piledriver quite mighty, but I can't see it.
As it is, in a lot of peoples eyes, the current crop of FX CPU's has flopped, they are replacing Phenom II's at the same price point, but where it matters, they're performing worse at the same money, games, encoding, etc.
Bulldozer obviously has some problems, take the Cinebench result for example, its scaling is horrendous, I feel that the CPU is just falling flat on its face in that instance. If you saw some of Gareths real world comparisons between a Phenom II 1090T at 3.6GHZ with a 2.6GHZ CPU NB, against the stock FX8, it's VERY disappointing. His Phenom II is able to catch up in the majority of programs.
Last edited:
Soldato
yeah I am somewhat notorious for long posts
they did indeed claim early on that their intention was to hold the line with IPC with regards to Phenom II, but JF-AMD has been saying toward the later part of the development process that IPC haven't only stayed the same, but improved over the Phenom II series.don't think this individual would have spent too much time looking at gaming benchmarks, just theoretical ones more than likely. there is no way with such an increase to ALU performance and FPU performance the instructions per clock should have went down, something just doesn't add up in that respect.
how a processor posts such good ALU/FPU scores then does so badly in arithmetic operations like HiperPi/SuperPi and wPrime makes no sense at all, like someone who is a mathematical genius getting stuck on something as simple as the eleven times table. so rather than arguing about how it has clearly went wrong, which I won't deny it has! how about we discuss theories on why it has went wrong since on paper the architecture is still a strong one?
Edit: who thinks that AMD are as baffled as all of us are to why this has occured? even Piledriver won't be that competitive unless they figure out what is stumbling Bulldozer, what is a spanner in the tracks so to speak, it is so essential that they figure that out and fast.
Last edited:
Caporegime
Superpi isn't really relative, I don't put much stock in those.
On a purely gaming point, where most of the forumers will be buying a CPU for, the FX8 can't perform as good as the Phenom II's, unless it's highly threaded.
You seem highly theoretical. What's the problem on the gaming side?
On a purely gaming point, where most of the forumers will be buying a CPU for, the FX8 can't perform as good as the Phenom II's, unless it's highly threaded.
You seem highly theoretical. What's the problem on the gaming side?
Soldato
haven't got a clue!? not even in the slightest, that is the baffling thing about the whole affair, it seems as though based on the more basic tests that it should be much better than it is, why its not, your guess is as good as mine!? just don't understand why a processor that annihilates my 1055T in mathematical horsepower (in theory) is slower in games...
to be honest, would love JF-AMD to return from his exile and give his opinions on this all, AMD have been very tight lipped about it all for the most part, which suggests that they are stumped to an extent!?
