Feeds

AMD's Bulldozer cores to push to 3.5 GHz and beyond

Big bump in oomph per socket

Maximizing your infrastructure through virtualization

ISSCC Chip maker Advanced Micro Devices lifted the veil a bit on the "Bulldozer" core module at the IEEE's International Solid-State Circuits Conference in San Francisco, showing off the power savings and faster clock speeds that the new design will allow for its workstation and server processors.

As is the case with most of the enterprise processor presentations at ISSCC, the ones done by chip engineers Hugh McIntyre – who talked about the Bulldozer core module – and Michael Golden – who talked about the out-of-order scheduler and integer unit used in the chips – are full of electrical engineering goodness, a very large portion of which goes right over El Reg's head. But there are a few things that mere journos can understand, and some pretty pictures to go along with them.

First and foremost, here is the first sighting of the actual Bulldozer core module, which McIntyre showed off:

AMD Bulldozer core module

Up until this time, AMD has only been showing an abstracted block diagram of the Bulldozer module. As El Reg has previously reported, the Bulldozer core is implemented in GlobalFoundries' 32nm, 11-metal layer, high-k metal gate, silicon-on-insulator wafer-baking processes. The idea behind the core module is to have some components shared across two cores, but to give each core its own thread (with no simultaneous multithreading) to have two strong cores. Each core – which means an integer unit and a floating point unit – in the Bulldozer module has its own integer unit scheduler and L1 data caches, but the cores share fetch and decode units as well as a floating point scheduler and L2 cache memory.

The Bulldozer modules are cookie-cuttered in two-core units, and the future "Valencia" Opteron chip will be four of these modules with a shared memory controller, L3 cache, and northbridge spanning the four modules and eight cores. Here's what the Valencia chip, presumably to be called the Opteron 4200, looks like:

AMD Bulldozer Valencia Opteron chip

AMD has not shown an Interlagos Opteron 6200 chip, and there's a reason for that. The Interlagos chip is just two Valencias in a single package, sharing the same G34 socket.

Each integer unit in each Bulldozer core has four pipelines, capable of executing one instruction per cycle. A Bulldozer core module has two 128-bit floating point units, which can do two 64-bit double-precision operations per clock or four 32-bit single precision operations. Either core in the module can see if the other core is doing floating point math. If it is not, then one core can take all 256 bits and do four double-precision or eight single-precision ops in a single clock cycle. AMD has started calling this an AVX mode.

The Bulldozer module has 2 MB of L2 cache memory and has a total of 213 million transistors, according to McIntyre. The module has an area of 30.9 square millimeters and is designed to run at between 0.8 to 1.3 volts. Here's a zoom into the Bulldozer core module:

AMD Bulldozer core module zoom

As you can see, the module has a 64 KB L1 instruction cache that is shared by both cores in the Bulldozer module, with each core having its own 16 KB L1 data cache and load/store units. The module has a shared microcode ROM, and the cores share a branch predictor, instruction fetch, and instruction decode units. The white line around the chip is a trough that allows for power to be completely shut down to each core module. So it looks like AMD will be gating power at the module, not the core level. It is not clear how this will compare to chip designs that are going to be gating at the core level.

The scheduler is capable of holding 40 entries and can issue four instructions per cycle and wake up on a single cycle. The floating point unit has new multiply-accumulate functions and supports a slew of new instructions, including SSE3, SSE4.1, and SSE4.2 SIMD extensions, on-chip AES encryption/decryption, XOP and PCLMULQDQ, which is used to perform a carry-less multiplication of two 64-bit integers.

The Power of One eBook: Top reasons to choose HP BladeSystem

More from The Register

next story
Sysadmin Day 2014: Quick, there's still time to get the beers in
He walked over the broken glass, killed the thugs... and er... reconnected the cables*
Auntie remains MYSTIFIED by that weekend BBC iPlayer and website outage
Still doing 'forensics' on the caching layer – Beeb digi wonk
SHOCK and AWS: The fall of Amazon's deflationary cloud
Just as Jeff Bezos did to books and CDs, Amazon's rivals are now doing to it
VVOL update: Are any vendors NOT leaping into bed with VMware?
It's not yet been released but everyone thinks it's the dog's danglies
BlackBerry: Toss the server, mate... BES is in the CLOUD now
BlackBerry Enterprise Services takes aim at SMEs - but there's a catch
The triumph of VVOL: Everyone's jumping into bed with VMware
'Bandwagon'? Yes, we're on it and so what, say big dogs
prev story

Whitepapers

Implementing global e-invoicing with guaranteed legal certainty
Explaining the role local tax compliance plays in successful supply chain management and e-business and how leading global brands are addressing this.
Consolidation: The Foundation for IT Business Transformation
In this whitepaper learn how effective consolidation of IT and business resources can enable multiple, meaningful business benefits.
Application security programs and practises
Follow a few strategies and your organization can gain the full benefits of open source and the cloud without compromising the security of your applications.
How modern custom applications can spur business growth
Learn how to create, deploy and manage custom applications without consuming or expanding the need for scarce, expensive IT resources.
Securing Web Applications Made Simple and Scalable
Learn how automated security testing can provide a simple and scalable way to protect your web applications.