AMD Lifts The Hood On Bulldozer At ISSCC

AMD's upcoming Bulldozer processor has been an increasingly hot topic as its launch date creeps nearer, but the company has kept a great deal of information under wraps. That's now beginning to change; AMD plans to discuss Bulldozer more in-depth at the ISSCC conference currently underway.

According to AMD Fellow Tim Fischer, Bulldozer was designed from the ground up to reduce power consumption. He writes:
Changes in clocking, latching, power management and on-chip memories are part of the comprehensive circuit updates incorporated into Bulldozer.  These are detailed in the paper, along with significant power reduction improvements, including clock gating, a new low-power flop design, and L1 cache power improvements.

AMD's high-level Bulldozer diagram, showing the independent integer units and shared FPU

These changes should have a significant impact on AMD's entire product line. Both desktop and Opteron-based chips will be more finely tuned. This should allow AMD to better meet the needs of both ultra-low power and high-performance market segments. As for the chip's performance, Fischer states:
High performance computing relies heavily on vector (packed integer) and floating point operations, both handled in the FPU.  Bulldozer was designed to execute these operations at higher performance and using less power than the current generation of microprocessors.  Key to Bulldozer’s performance and power improvements are FPU changes, including completely redesigned arithmetic units and control structures. As previously described... the Bulldozer FPU supports new instructions including SSSE3, SSE4.1, SSE4.2, AVX, AES, and advanced Multiply-Add/Accumulate operations.

The lower-level, more detailed diagram. AMD claims that each Bulldozer module contains "two tightly linked processor cores." In reality, the module contains something like 1.5 - 1.8 CPU cores which are linked together and operate using chip-multithreading, or CMT.

AMD has released a block diagram with more information on it than we've previously seen. The company has also hinted that Orochi will be capable of operating at 3.5GHz at launch, though it's not clear if this is an average clockspeed or the highest speed the company is targeting. The CPU designer claims that Bulldozer "improves performance and frequency while reducing area and power over a previous AMD x86-64 CPU in the same process. The design reduces the number of gates/cycle relative to prior designs, achieving 3.5GHz+ operation."

Just The Facts

Based on the data AMD has released to date it's clear that Bulldozer will offer better power management and performance than what we've previously seen in K10. More importantly, at least for AMD's profit margin, it offers these advantages on a significantly smaller die than a traditional quad core. This means AMD can build more processors on a single wafer, thus lowering its cost of goods sold.

We can safely assume that Bulldozer will outperform Phenom II and current Opterons both in terms of performance-per-watt and raw performance. There may be a few corner cases where the "true" quad-cores outperform Orochi thanks to their individually dedicated FPUs, but Bulldozer's additional features and higher clockspeeds should close the gap.

The bigger question, of course, is whether or not Bulldozer will be able to close the gap with Sandy Bridge. Later presentations at ISSCC may offer insight on this topic, but for now we're still mostly in the dark. It stands to reason that Bulldozer will at least shrink the gap—but how much is still unknown. 
Via:  AMD
SammyHayabuza 3 years ago

Funny how AMD always talks about the Bulldozer outperforming its own Phenom chips, but don't talk about Intel much . To tell the truth I don't expect BD to compete core per core in raw power against SB or Ivy B. But probably BD will offer more cores for the money in their higher end chips as an attractive buy for the consumer.

Also, I find the GPU on Desktop Sandy Bridge processors to be a waste money or resources, seeing that you cant enable it on higher end motherboards, and people that are buying the H67 boards will buy a Mid range GPU anyways.

Joel H 3 years ago


Two points. First, rumor has it that AMD has kept quiet about BD deliberately, so as not to tip its hand. AMD won't confirm or deny of course, but after the original Phenom debacle the company has taken a "less is more" approach to pre-launch hype.

Second, AMD doesn't need to match Sandy Bridge or Nehalem at every price point to be competitive. SB chips are very competitively priced but there may be room to maneuver. Right now, a 3.33GHz dual-core Clarkdale is $149. AMD's closest equivalents are a quad-core Black Edition Phenom II 955 (3.2GHz) at $139 or a Black Edition Phenom II 965 (3.4GHz) at $159.

It's already clear that AMD is aggressively pushing more cores against Intel. If we assume that a Bulldozer chip is a modest 10% faster than a Phenom II at the same clockspeed, it would be very, very competitive with Intel at that price point.

Bulldozer would have to be a miracle processor to re-establish AMD at the very top of the market, but that doesn't mean it and Llano can't substantially strengthen AMD's position in the value, low-end, and mass market segments.

rapid1 3 years ago

Yeah; Sammy I get your points exactly, and of course we won't know anything until Dave/Marco/One of us etc get our hands on one to see it in action. I do applaud AMD for finally getting this thing near to release, but also find it to be somewhat lacking. I mean Intel has released how many new processors in the last 2 years V. AMD really? So it strikes me as it will be behind the game unless they really do something advanced beyond Intel. Seeing this I kind of doubt it, and of course Intel will have a new tock in what 4 months now.

coolice 3 years ago

Though i'm a computer enthusiast, i dont understand 1/2 the stuff posted in this article, so forgive me for my lack of understanding.

Can someone explain to me what this statement means:

[quote] The lower-level, more detailed diagram. AMD claims that each Bulldozer module contains "two tightly linked processor cores." In reality, the module contains something like 1.5 - 1.8 CPU cores which are linked together and operate using chip-multithreading, or CMT.[/quote]

1.5-1.8 cpu cores?? how can you have 1/2 a core? i understand if you use 1/2 the processing power of the core, but 1.5-1.8 cores?

CDeeter 3 years ago

Looking at the diagram in question there are 3 sections in the middle. Left and right sections are the non shared parts (Int Scheduler, Instr Retire and so forth) of the two "cores". In the middle (FP Scheduler and so forth) are shared components. Normally these are not shared, and are duplicated in each core. So in essence you don't have 2 complete cores, but 1.5 or 1.8 depending on how you weight the shared components.

This enables AMD to do a kind of multi core hardware based hyper threading instead of Intel's single core logic based hyper threading.

coolice 3 years ago

Ohh, i see... so is that in edition to the # of cores on a particular die? for example a dual core is really a 2.5 core? Sorry for my noobness... i'm not That technical.

Bulldozers the reason i've been putting off my budget build for the last 3-4 months.... and if it fails to outperform s/b i'd be a little disappointed. Then again, i'm not going for a crazy powerful as i've mentioned before. I just need a computer that will last me 4-5 years without the need to upgrade.

I purchased and built my current desktop around 4 years ago.... amd am2 x2 5000, and i've been slowly upgrading the parts just to keep up (ram, hdd, psu, vid card). I want to build something like in the next few months that will last me for another 4-5 years. At that time my current pc was a top of the line device (though within budget) and thats what i'm aiming for now, something that'll do me good by lasting me long without major upgrades to the core components (mobo+cpu)

CDeeter 3 years ago

I'm in the same boat as you are as I'm still rockin my Athlon x2 4400, 4GB ram and a 7200gs. The thing just won't Hopefully we won't need to wait much longer if Bulldozer launches as scheduled.

Or maybe one of us will win that sweet system HH is giving away!

RMikula 3 years ago

Instead of having two distinct cores, you have two cores feeding on the same thread. This prevents each core from reaching 100% of it's ability. Therefore, you have 1.5 - 1.8 core equivalent. What makes bulldozer special is that you can apply two cores to a single thread like that. Typically, you can only apply one core to one thread and utililze 100% of the core. Bulldozer applies two cores to one thread at approximately 80% on each core. Most applications are written in 1 or 2 threads. The best that any current processor can do on a two thread application is use 2 cores. Bulldozer can use 4. Current processors on a two thread application: 1 + 1 = 2. Bulldozer processor on a two thread application: .8 + .8 +.8 +.8 = 3.2. You therefore have more applied processing power with a bulldozer even though each core is a little less powerful.

CDeeter 3 years ago

Nice explanation! Thanks

For a more detailed explanation check out this article fro PC Perspective:

soldier1969 3 years ago

Give me 8 cores at 4Ghz + and 20 - 40% boost in overall speed than my current X4 955 at 3.6 Ghz for less than $350 and Im sold AMD !

rapid1 3 years ago

I guess we will just have to wait and see what AMD gives us, other than that this is all just conjecture anyway, my statements included. AMD is obviously trying to keep quiet about this so as has been said they do not tip there hand, so it is all to be seen, and of course it also looks like we do not have long to wait.

dodgers2213 3 years ago

again.....competition is good for the consumer Cool

realneil 3 years ago

[quote user="dodgers2213"]again.....competition is good for the consumer [/quote]

IdeaYesIdea, this is what makes them relevant in the market. They provide competition for Intel, and even outdo them every once in a while. They do this at a better price point as well,....and that keeps Intel from being able to charge higher prices for their products than they do now. This is why I still buy from AMD whenever I can, even if Intel has something that's just a little bit better for a little more cash.

The last three people who came to me asking for my advice on what Dell to buy all ended up with home built Phenoms, and they're all loving it because we were able to add larger HDD's and more RAM to the systems and stay within the amount that they wanted to spend.

I feel that they're worth supporting as often as possible.



rapid1 3 years ago

The thing is for the general user the AMD parts are more than adequate. Also as realneil and dodgers point out the competition is good for us all.

I actually got my mom and off market Acer/Gateway PC with a Phenom in it, 4 Gb memory, and replaced the GPU and PSU with components I had in my basement as well as a couple fan's. The kicker is the it came with a 24" LCD display that is nicer than me or anyone I know has on there PC for $425 shipped.

At a price level like that you really cannot beat it, I also could not really beat that price for the components. The other thing is the only place a general user would be able to really see the difference would be in a benchmark if you explained it to them. They surf the web, check there email, and do Facebook. They may even have some pictures they took on there PC, but will only edit them in the instant correct way, not in the true Photoshop or GIMP way. They also don't burn CD's, don't write software, burn movies to DVD, or even know how to up there graphics setting's if they even play any real games.

So the general users benefit's immensely, and it keeps Intel and AMD busy trying to provide more within there price line, and of course Intel has to keep there price's as low as they can do so while still making enough money.

Joel H 3 years ago


That's almost certainly *not* going to happen unless we're talking about an application where Bulldozer can use AVX in a manner Phenom can't. You want eight cores at 4GHz and a chip that's 20-40 percent faster.Unless you're a power user who renders a lot, you're not going to pick up 40% performance from four more cores. That leaves inherent core architectural improvements + clockspeed jumps.

I won't be surprised if 3.5GHz is the maximum launch speed. (I actually wouldn't be surprised by a 3GHz launch speed). Even if I knew 3.5GHz was base, I'd bet that the maximum would be 3.8GHz, max.

That's just 8.5 percent faster than 3500MHz.

If you look back at historical figures, you'll note that while larger jumps are possible (K7 - K8, Willamette - Northwood, Wolfdale - Nehalem), the typical best-case performance jump in aggregate from architecture to architecture is maybe 15%. In all of the cases above, we can point to very specific reasons why performance jumped more than that:

K7 - K8: Integrated memory controller, SSE2 support.

Willamette / Northwood: P4 was badly hurt by just 256K of L2, Northwood scaled MUCH better than Willamette).

Wolfdale - Nehalem - Integrated memory controller, QPI.

Everyone Else:

The one big thing I think ya'll have collectedly failed to mention is that "Keeping Intel Honest" is not a business strategy. Granted, AMD's BoD has raised serious questions about whether or not it's actually fit to run the company of late, but AMD has just barely survived more times than I can count. Keeping Intel honest might be great for consumers, but AMD's financials are rock solid proof that it's a terrible goal for a company.

Post a Comment
or Register to comment