• 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.

Bulldozer quad CPU

I'd just stay with what you have at the moment and wait and see what the CPUs are like after a while. Don't want one of the B2 steppings over what you have anyway...

Can't be much you are struggling to run on the 965? What clock speed is it at?
 
If you REALLY need to upgrade, then get a 1090T and OC it, that's the best AMD processor available. If you really want Bulldozer just for the sake of it, then I suggest you wait until either AMD release a new stepping which is drastically better, or Piledriver is released. Your choice, but going BD now is stupid as even the 8150 is worse at most things than your 965
 
I'm still not too keen on the need AMD chips. I still find It funny how the phenom 6 cores got beaten by the i3 2100 in many tests.
 
at the moment there is no reason to choose a four core Bulldozer over a four core Phenom II, because without a core count advantage they loose the Phenom II and tend to need higher clock speeds to match them. also the i3 will win in thread vs. thread performance, that is just a fact but multi-threaded the Phenom II will come out ahead. the problem from my point of view, I got this Phenom II 1055T for <£80, which makes it pretty much better value than any current Intel processors to be honest, regardless of the fact that even a 'mere' i3 2100 will beat it in single threaded workloads. so it all depends on what you need from your system.
 
dont mean to be a numpty but who has ever heard of any of those benchmarks , why dont they do a proper benchmark test with what we buy this stuff for ie a benchmark of how fast it can rip a dvd using handbrake or set the levels on 100 photos in photoshop or how many fps you get in bf2 or bf3 beta i mean what the hell is clomp v 3.3 more like crap v 3.3 but what you can afford and dont belive the hype
 
They are synthetic benchmarks which illuminate very useful traits of a CPUs architecture to draw meaningful conclusions and, more importantly, make speculation of how the chip might mature with more cores/clockspeed etc.

If you do not care for such pieces of data, read different review sites. HardwareHeaven will be a good click for you if you want a feel-good story about Bulldozer in BF3 with a non-cpu centric review.
 
thing is though, at a design point of view Bulldozer quad-cores aren't meant to compete with traditional quad-cores. since from a pure design standpoint, each module is two integer cores containing 66.6% of the physical resources of a 'stars' based core and a clever shared FPU. now this combined with its caches takes up about 30.9 mm2 of die space, so four modules would take up around 122 mm2 of die space, the problem with the size of Bulldozer is the enormous amount of cache present on the chip, 8MB of L3 takes up considerable space which baffles me why they did it. so strictly speaking a four core Bulldozer should be getting compared to a dual-core standard design, that was always from day one the idea behind Bulldozer, divide and conquer. the poor single threaded performance (caused by that 66.6% of 'stars' core resources!) was supposed to be offset by a substantial increase in clock speed, the whole of the Bulldozer modular design philosophy relies entirely on how many threads can be thrown at the design at once, then it should be able to offset the IPC decrease and turn it into a increase in throughput. :)

Edit: also in the only review of the FX-4100 I have read, again they are upto the same balls comparing it to quad-core Llano processors and quad-core Sandy Bridge processors, which it was NEVER intended to be compared to! in the review I am currently looking at the FX-4100 offered reasonable performance against the Sandy Bridge dual-core (which I believe are Hyper-Threaded?) which is what they should be compared to from a design standpoint. also it is even priced to compete with the Sandy Bridge i3-2100 series, which it does reasonably well, would be better if Global sort out their fabrication issues! same thing goes for the 8120 and 8150, to be compared to the i5-2500 series on price point, since the i7 is still a little bit pricier, and again a lot of the time it does that reasonably well, other times it doesn't do so well.
 
Last edited:
thing is though, at a design point of view Bulldozer quad-cores aren't meant to compete with traditional quad-cores. since from a pure design standpoint, each module is two integer cores containing 66.6% of the physical resources of a 'stars' based core and a clever shared FPU. now this combined with its caches takes up about 30.9 mm2 of die space, so four modules would take up around 122 mm2 of die space, the problem with the size of Bulldozer is the enormous amount of cache present on the chip, 8MB of L3 takes up considerable space which baffles me why they did it. so strictly speaking a four core Bulldozer should be getting compared to a dual-core standard design, that was always from day one the idea behind Bulldozer, divide and conquer. the poor single threaded performance (caused by that 66.6% of 'stars' core resources!) was supposed to be offset by a substantial increase in clock speed, the whole of the Bulldozer modular design philosophy relies entirely on how many threads can be thrown at the design at once, then it should be able to offset the IPC decrease and turn it into a increase in throughput. :)

IIRC, supposedly the CPU cores and L2 and L3 cache take up 1.2 billion transistors.The other 800 milllon transistors are taken up by the memory controller and things like I/O which is meant to be quite big for a desktop chip.
 
Last edited:
yeah indeed, but even then people wonder whether the L3 is even needed at all on the chip? would love for them to remove it entirely and we'll find out, as it stands though its around the size of a Thuban die as far as I remember, which isn't as bad as people make it out to be, but considering the transition to 32NM should realistically be smaller.

the enormous problem is the way its being compared, the 'eight-core' 8150 should be compared more like four modules vs. four cores, which is the intention in the grand scheme of things, so in that respect I think its a decent first attempt and once they sort out the process it'll show its strengths. problem is nine times out of ten on here people are writing it off because it gets beaten in lightly threaded workloads, even by Phenom II but they always, always forget that was never the purpose of Bulldozer! lightly threaded it is behind the competition, multi-threaded it is competitive with the competition and heavily multi-threaded it starts to show its potential.

people have to start seeing Bulldozer for what it is (regardless of marketing) and that is 4 module, 8 thread processor, when compared to the 2600K for example which is 4 'core', 8 thread processor, that is how they should be compared, argued this point for ages about comparing them core vs. integer core and how it shouldn't be done! :confused:
 
It doesn't matter how you define it, it doesn't change the end performance.
It's got 2 billion transistors, the yields are horrendous, price/performance is worse than Phenom II, it consumes a lot of power and there's rumours of the chips being unstable, how in any way is it a success? The FX6 is going to replace the 1100T (At a guess) price wise, but with lower performance, that's the same with the FX4 as well, with the Phenom II x4 at that current price point. I honestly can't see why people are trying to sing its praises.
 
Last edited:
yeah indeed, but even then people wonder whether the L3 is even needed at all on the chip? would love for them to remove it entirely and we'll find out, as it stands though its around the size of a Thuban die as far as I remember, which isn't as bad as people make it out to be, but considering the transition to 32NM should realistically be smaller.

the enormous problem is the way its being compared, the 'eight-core' 8150 should be compared more like four modules vs. four cores, which is the intention in the grand scheme of things, so in that respect I think its a decent first attempt and once they sort out the process it'll show its strengths. problem is nine times out of ten on here people are writing it off because it gets beaten in lightly threaded workloads, even by Phenom II but they always, always forget that was never the purpose of Bulldozer! lightly threaded it is behind the competition, multi-threaded it is competitive with the competition and heavily multi-threaded it starts to show its potential.

people have to start seeing Bulldozer for what it is (regardless of marketing) and that is 4 module, 8 thread processor, when compared to the 2600K for example which is 4 'core', 8 thread processor, that is how they should be compared, argued this point for ages about comparing them core vs. integer core and how it shouldn't be done! :confused:

It's being marketed as the worlds first 8 core cpu. Also if what you're saying is correct then it's either a design flaw or they should not have given it the FX brand, people expect the FX series to be the latest and greatest and wipe the floor clean with intel as it has done in the past.
 
well aware that it is being marketed as 'eight core processor' which is why I said (regardless of marketing). also well aware of its shortcomings, give me a break I can read, but it was never ever going to compete in light-threaded environments, anyone who thought it was is just delusional. the problem is, people are making out that it is totally useless at everything, when it is pretty decent in heavily multi-threaded workloads just like it was intended to be, competitive against its rival product. at the end of the day though I have since the NDA disappeared recommending avoiding this first iteration of Bulldozer because of the power consumption problems, yield problems and so forth, but they are more Global Foundries issue to be honest not AMD as such.

though I do agree that the transistor count is nuts, the die-size is comparable with the six core Phenom II parts so its nothing new for AMD. for goodness sake as well, not saying its the best thing since sliced bread or anything even remotely such just trying to bring some perspective into the debate, stop comparing them in a core vs. core respect, regardless of what marketing are piping out (remember they are just trying to sell a product so make use of every numerical advantage it has!) because it isn't intended to be compared in that respect and you damned well know it.

also the FX argument is rather flawed as well, because think of it this way, Intel have their Extreme Editions, there is nothing Extreme about them except the price, four times the price for a marginal difference in performance. I know in the past FX was 'top of the tree', but then that is what Bulldozer was supposed to be, part of the AMD tree, so FX seems a sensible moniker.

it hasn't went as well as AMD would have hoped but their architecture doesn't do THAT badly in the area it was designed to perform, the biggest setback for them is the poor yields and obvious troubles with Global' fabrication, let them sort them out with a new stepping (which will bring power consumption down) and let motherboard manufacturers iron out all the kinks and come back to it, it'll still not be great in single-threaded workloads but it should compete in the multi-threaded workspace. I do however totally agree the 4100 and 6100 series make no sense, their prices are only so high because I think I am right in saying they use the same die for all their lines, so a 4100 is basically a 81** series with half its modules disabled, which is wasteful.
 
Does not do that bad in heavily threaded apps?

What? It was SUPPOSED to do well here and rarely, if ever, beats out the 2600k. I would agree with you if the 8150 was smashing everything in highly threaded situations but - it's not.

Are you telling me the plan was to formulate a hugely new architecture with 8 'cores' to demolish highly threaded applications only to match the current 2600k? This is also ignoring the fact the 2600k clocks much much better.
 
Back
Top Bottom