nINJAkECILThey are indeed a real core (to be exact, an integer core. AMD has moved away from a "real core" marketing slogan and what we see as a "real core" right now).
4 module, each module has 2 integer core, and each integer core has its own L1 data cache, integer scheduler and integer datapath, while sharing L2 and L3 cache between integer core in one module.

This is the best die shot I have on BD architecture:

img.techpowerup.org/110925/BD die.jpg

sure its not from google earth? :p
j/k

YautjaLordBTW: 2 all those that say 2600K outperforms 980X (let alone 990X) - i think it's on a contrary: don't remember @ which sites, but in most cases 990X & 980X outperform the afformentioned CPU stock & OC'd; in games & synthetic benchies i think it's better than 2600K excluding apps that specifically desingend for SB if there's any. No fanboy s*** on my side, but this Westmere 980/990X is still heck of CPU compared to SB; not that 2500/2600K's are bad: if i had enough ca$h i would buy one BD-based gaming rig & one SB/SB-E-based rig & than pit one gainst the other to see which beats the f*** out of the other. :laugh:

Regardless of what i said above, these news completely acceptable by me, since this is how i hoped the BD (this FX-8150) will fair out. Now the only thing left to know is how it'll run when paired with C5F, Sabertooth 990FX or any other 990FX/X-based mobo, 2x4GB DDR3 1600MHz RAM, 2xGTX 4xx/5xx or 2xHD 6xxx, 1kW+ PSU & fast HDD or SSD & tested here. If you'll pit it (FX-8150) gainst both 990X & 2600K - ace. :toast:

P.S. Is there a CPU-based Photoshop benchie to test multithreading? I might check this one out once i'll buy this CPU; probably end of October/middle of November.

Nope, 2600K beats the 980x in almost all game tests. In just 1(one!) game test, the 980x wins.

Proof:
www.anandtech.com/bench/Product/142?vs=287

Crap DaddyThis is a marketing smokescreen by AMD. The funny/sad part is that they present the FX as the ultimate gaming chip and they fail to confirm this in all those slides. Comparisons are made with their own Phenom and a general presentation comparing it with an Intel last gen chip (by the way, discontinued) which is proven to be under the current Sandy Bridge generation in gaming. Furthermore we don't know if they used an xfire setup or a medium class single GPU since this might alter drastically the results.

Anyway, BD has to be better than Phenom II x6 otherwise what's the purpose? It's 315 mm2 die size compared to Sandy's 216 mm2 (on which we also have an IGP) should provide some performance, isn't it?

Ummm don't they show a test in games against the 2500k in eyefinity? and the 8150 shows up to a 30% performance over the 2500k. Its in one of the slides, ive got it open now. :confused:

I don't see why people compare 8 AMD cores to 4 Intel cores - Remember that for each BD module, two FAMC units are shared between the "cores". I guess with only 4 cores active, those cores have access to all the FAMC's...

We really should be comparing 4 active cores on a BD 8-core CPU to the 4 cores in a core i5 CPU...

And before we see any official benches, we can hardly even make qualified guesses as to how the two competitors compare in regards to real world performance.

The performance of BD decides my next CPU... but I think that the mere option to freely tweak and OC the CPU's are going to win over a lot of semi-enthusiasts that "plan on" venturing into overclocking.
Let's not forget that the BD lineup will contain the cheapest unlocked CPU of this generation (so far).

I've never liked HT. The only reason it works is because of shoddy software. If you fully stress the 2600's cores with 4 threads, you will not see any benefits from HT, because what HT does in very basic terms is widen the processor pipeline, so that if a thread stalls, it can be parked and the next thread after it can be picked up and processed while the stalled step is flushed.

However BD's design does not widen the pipeline, it literally adds a second parallel pipeline next to it, so that two threads can run simultaneously through the module. The thing is that for you to truly experience this extra performance, you need properly written software and that is somethig we rarely find in today's world...

I applaud AMD for what they're trying to do. It may not be the best solution right now but it sure is the one better suited for the future. There are new algorithms found every day and new and better libraries made for the popular compilers. Sooner or later we will not be able to scale hardware any further or with the same pace as today and when that time comes we will need better software. And I believe AMD's design will be better suited for that software.

Damn_SmoothWho has a Xeon?

Mac users. :rolleyes:

The CPUs (4xxx/6xxx/8xxx Dozers) will be released October 12, right? Like i said: prior to or at it (release) TPU staff have to pit these (especially FX-8150) gainst both 990X & 2600K, that's when it'll be unbiased & legit. Waiting to see that happen; if the results are identical or 90%+ same - ace twice. Still hoping BD to wipe the floor with 990X & 2600K in most gaming/synthetic/system benchies suite you got guys. :toast:

HalfAHertzSooner or later we will not be able to scale hardware any further or with the same pace as today and when that time comes we will need better software. And I believe AMD's design will be better suited for that software.

You do realize that wont happen in the near future. Not for at least 10 years.

I'm not so sure why all the debate about hyperthreading and vityual cores.

What matters for gaming will be core speed and instruction efficiency. None of that other rubbish will matter for nearly every game on the market.

BundyI'm not so sure why all the debate about hyperthreading and vityual cores.

What matters for gaming will be core speed and instruction efficiency. None of that other rubbish will matter for nearly every game on the market.

True.

If:

BD is better than SB 2500 for majority of general use including gaming it is good.
BD is cheaper than SB 2600 it is good.
BD core count (whatever it is) allows it to perform better than SB in multi threaded non cherry picked tasks, it is good.

BD is 125w versus SB 95w TDP:

it uses 30w more at top spec, which is 30/95 = 32%.

For ref as well, accord to other marketing slides (www.techpowerup.com/149464/Core-i7-3960X-About-47-Faster-On-Average-Than-Core-i7-990X-Intel.html) the SB-E top cpu (which is a more relevant competitor to use than 980x) is 47% faster than the 990. Given the unrealistic price point, the 1000 dollar cpu shouldn't be used and if it is, like i say, use the upcoming 3960, not an outdated 980.

Price point versus price point, BD looks good enough BUT not the winner. But consider if it keeps up with (or near enough to) the SB line up, then it is an excellent step for AMD. If AMD have got a competitively priced cpu that comes fairly close to SB then it has succeeded.

However, one caveat. Going back to power, its akin to the 6970 versus 580 debates. The 580 wins but at a considerable power cost. With a component drawing 32% more juice than it's rival, you'd want it to hammer it. BD draws far more power but doesn't exceed it's competitor (well not by the vague PR slides).

I game and surf. If BD games well (within 10% of SB) and costs much less (system wide), I'd happily put it in my PC. If only IB wasn't coming in 6 months.

the54thvoidit uses 30w more at top spec, which is 30/95 = 32%.

It produces 30 watts of heat more at top spec*

TDP = Heat, Joules per second

Max Consumption of power can either be higher than the TDP if the CPU is inefficient or it can be lower if the CPU is efficient(actually, I might have flipped these but oh well hahahahaha)

The estimation is that Bulldozer consumes in max workloads is

105 to 110 watts @ turbo clocks while it produces the bracket of 125 watt heat

TDP wattage tends to be bracket based

95 = 94.9999 and below
125 = 124.99999 and below(ends at 94.9 where it can be called 95 watt TDP)

105 to 110 watts @ turbo clocks while it produces the bracket of 125 watt heat

This is physically impossible. Every single Watt that is "consumed" by the CPU and is not output as data signals will be converted directly to heat.
If the CPU consumes 125W at full load, then it produces 125W of heat energy. - if it consumes less than 110 watts, there's no way, it can produce more than 110 watts of thermal energy (first law of thermodynamics).

RookienoobThis is physically impossible. Every single Watt that is "consumed" by the CPU and is not output as data signals will be converted directly to heat.
If the CPU consumes 125W at full load, then it produces 125W of heat energy. - if it consumes less than 110 watts, there's no way, it can produce more than 110 watts of thermal energy (first law of thermodynamics).

I think what he meant is it draws 110 watts but runs in a bracket category of 125 watts (for safety margins).

Either way, a 30% difference in TDP is still mighty big. And by seronx's own statement SB could be far lower than 95 watts (it fits into the 95 watt bracket for TDP).

So using apples to apples in logical terms what AMD say is "My CPU requires 125 watts of cooling power" and Intel say "My CPU requires 95 watts of cooling power". Which in essence says, my cpu runs 30% cooler because it creates less heat because it uses less power.

Even taking in to account inter company variances in calculations, 30% is still a large difference.

the54thvoidBD is 125w versus SB 95w TDP

AMD's TDP ≠ Intel's TDP

repman244AMD's TDP ≠ Intel's TDP

it'd be nice if they did use equal TDP values, but sadly they dont compare directly.

the54thvoidI think what he meant is it draws 110 watts but runs in a bracket category of 125 watts (for safety margins).

AH... This makes a lot more sense. Sorry for misunderstanding that :(
So basically, the two 8-core CPU's will probably have different power consumptions even if they're in the same 125 TDP bracket? Does this mean that some of the 8120 or even 8100-chips may be able to run at 8150 speeds without breaking the 125TDP?

If I were to sell an Intel chip, I'd clearly point out that it has a superior performance/power consumption ratio and that the procssing power is concentrated in just 4 cores.

If, however, I were to sell an AMD CPU, I'd point out that it supports a new instruction set, that Intel doesn't, that it provides better value for the money and that it's a good way to get into overclocking.

Musselsit'd be nice if they did use equal TDP values, but sadly they dont compare directly.

Yes it would be nice, AMD also has ACP which AFAIK is mostly used for server CPU's.

RookienoobAH... This makes a lot more sense. Sorry for misunderstanding that :(
So basically, the two 8-core CPU's will probably have different power consumptions even if they're in the same 125 TDP bracket?

this is always the case. if a CPU is labelled as 95W, it doesnt use 95W perfectly. - it just means it fits under 95W TDP.


you can have a dozen chips in a lineup rated for the same TDP, but individual power consumption will vary depending on clock speed and voltage. they really just list a maximum (and thanks to them TDP asshattery, its more of a maximum average :rolleyes:)

Musselsand thanks to them TDP asshattery, its more of a maximum average

And that is where the differences are between the two...

*AMD's ACP uses a "round down" average of power measurements performed with industry standard benchmarks (usually running at 100% CPU load, with the exception of Stream).

*AMD's TDP is close to the electrical maximum a CPU can draw (when it is operating at its maximum voltage).

*Intel's TDP is a "round up" average of power measurements of processor intensive benchmarks.

www.anandtech.com/show/2807/2

Damn_SmoothWho has a Xeon?

Two Xeon 5310, in my server. :p

nINJAkECILThey are indeed a real core (to be exact, an integer core. AMD has moved away from a "real core" marketing slogan and what we see as a "real core" right now).
4 module, each module has 2 integer core, and each integer core has its own L1 data cache, integer scheduler and integer datapath, while sharing L2 and L3 cache between integer core in one module.

This is the best die shot I have on BD architecture:

img.techpowerup.org/110925/BD die.jpg

The unique thing about AMD modules is that two cores share a floating point unit. Not exactly sure how that provides an advantage but as you said, they are "real cores." They don't fit the description of logical cores like Hyper-Threading or IBM's implementation of logical cores. They're closer to physical cores.

Musselsit'd be nice if they did use equal TDP values, but sadly they dont compare directly.

AMD is the odd one out. Everyone else uses Intel's method of measuring TDP, including Via.

Damn_SmoothAnd that matters how? I don't care what's inside the CPU as long as it performs. It's stupid to say that what is inside the CPU is more important than how the CPU performs.

Wow they got u pretty worked up there....
:confused: isn't what inside a cpu what gives it its performance?

So much hate for calling them 8 cores??

AMD can save a bundle of silicon by doing it this way and hopefully not compromise performance much, if at all.

It makes sense and they pass the savings on to you. So, whats the problem?

MikeMurphySo much hate for calling them 8 cores??

AMD can save a bundle of silicon by doing it this way and hopefully not compromise performance much, if at all.

It makes sense and they pass the savings on to you. So, whats the problem?

maybe thats BD's secret to the low prices, maybe it really did cut costs a lot.

Well isn't it like a 25% die increase for 80% of the performance ?

( something like that) Bound to save some pennies.