OLPC XO 1.5 Beta 2 Version Available For Contributors

As the release of the new XO 1.5 nears, 30 B2 versions are available for hardware and developers and they start shipping next week to the ones with the best proposals.

The revamp to the current Geode LX based OLPC XO-1 laptop features various improvements, including an updated C7-M @ 1GHz, 1GiB for DDR2 RAM, replaceable internal flash memory and the VX855 chipset with hardware HD video decoding capabilities. For the original announcement and the full specs, look at the previous post about the XO-1.5.

The revamp is much needed, because while the XO-1 fullfills his mission remarkably well, the Geode LX was unable to run things like Youtube videos, Hulu, etc. Videos still play fine using MPlayer and other applications work fine enough. Wireless range is good and battery life is very good for such a cheap device, similar to most netbooks of the same price range: around 5 hours when using just the internal memory, 3 hours running the OS from an SD card and around 9-10 hours if used in ebook mode, where the laptop can suspend itself often.
The device has a remarkable suspend/resume time of around 150ms, which allows it to retain the image in the DCON(a custom chip you won't find elsewhere), output it to the screen while most of the system is suspended and still wake up at a press of a key before you notice it. The reflective screen is definitely another strong point of the machine and one of my favorite ones.

As for the new one, motherboard pictures have already been provided for reference and it's a pretty compact one, in the same form factor as the previous XO:


The casing remains the same but there are a few changes going on the outside. The new XO-1.5 is sold exclusively with the simplified trackpad and green buttons on power, rotate and the battery holders below the laptop:


There are 30 laptops available for contributors that want to help out the OLPC Foundation, proposals should be sent contributors@laptop.org and should follow the guidelines for project submission. For more information please visit blog.laptop.org
Before you apply, also have a look at the wiki page about the Xo 1.5 B2, as there are still some issues to sort out before it will go into production.

There isn't any Give One Get One program planned for this year yet, but more information will be released soon and I hope to have one XO 1.5 to review here in the coming months.

Nvidia Ditches LGA 1156 Chipsets

I had cautioned last month that we might not ever see an Nvidia chipset for LGA 1156 CPUs. We now have a confirmation of this from Nvidia but the chipset woes don't stop there.

The story about the ditching is from Fudzilla and goes on like this: "Nvidia confirms to us that it won't make a DMI chipset due to legal pressure from Intel".
Given that Fuad has been very pro-Nvidia lately and this is bad news, this time I'll take is word for it, or part of it. Nvidia calling the quits is not just as big of a legal problem as it is an engineering one, one with bandwidth starved bus, one which Intel was carefull enough to craft even though it end up delaying their integrated offerings, known as "Clarkdale".
Nvidia could build a discrete chipset, or I/O Hub, for the LGA 1156 processors but this theory was mostly buried when they announced that would be licensing SLI support for P55 chipsets.

There's a bigger issue on the horizon, which is called Ion 2. It's supposed to be coupled with Atoms yet again but I wonder with what. Here's the diagram of Intel's Atom N2xx successor, "Pineview":


Will the Ion 2 be a thing for the current Atom? One that it will stop manufacturing as soon as "Pineview" is out? The new Atom is expected already this year. The story is the same as before: no FSB, no bandwidth to access the embedded memory controller, no bandwidth for graphics. Hence, no DMI chipsets. Everybody seems quiet about this right now, so I'm going to revisit rumor: Nvidia needs to seriously take a look at VIA. Like now.
It either needs to buy VIA to acquire it's IP and x86 license or it needs to make an investment on it to have Nano processors coupled to Nvidia's chipsets. The notebook market is as big as it's ever been and a combination of both would deliver very interesting performance, albeit at a higher TDP than Atom's. That hasn't been a problem for VIA right now, imagine with an actual Nvidia graphics core (and compute core) behind it. Plus, the dual core Nano is expected next year.

Nvidia GT300 ''Fermi'' Architecture Unveiled

Image courtesy of bit-tech.net

Yesterday, on the new GPU Technology Conference organized by Nvidia, the new Tesla cards were shown and the chip's architecture was exposed - the focus is again on CUDA.

The new chip is huge:


It's expected to be close to 500mm² while AMD's Radeon 5800 Series weighs in 334mm². It is manufactured at 40nm at TSMC but we still don't know any specifications regarding core, shader or memory clock. What we do know is that the card has a 384-bit bus, capable of delivering 50% more bandwidth than AMD's Radeon 5870. The presentation today was mostly to scare off potential Radeon 5000 buyers and to address the main target of this new architecture: High Performance Computing.
So is the card shown by Huang, which is smaller than current generation high-end cards. This isn't strange, as Tesla cards have always had a lower power consumption due to having disabled parts of the die dedicated to graphics rendering, hence the smaller form factor.

There's no doubt about the HPC/CUDA focus, just look at what the new architecture brings:

• ECC support - like everywhere, registers, memory bus, memory and caches. This is straight from Jen-Hsun Huang's mouth as he presented the chip. R800 chips from AMD only have ECC for memory bus errors.
  
• 8x double precision performance on floating point calculations over GT200, half of peak single precision performance.
• Unified address space, which enabled support of object oriented programming(C++).
• L1 and shared memory are interchangeable so one of them can be 16 KiB or 48 KiB, as the programmer chooses.
• Addition of an L2 cache(the blue in the middle of the chip), of 768 KiB, shared by all cores.
  
• 384-bit width of the memory bus ensures further scalability in compute but also in games over the previous generation. (AMD forgot this and just touts peak performance numbers)
• Can execute more than one compute kernel from the same context at the same time.
• Can switch between contexts(full programs) up to 20x faster than the GT200 architecture.
• Full IEE 754-2008 compliance for all floating point calculations.
• 64bit memory addressing capability, up from 32bit(4GiB), although limited to 1TiB for now.
  

The addition of an L2 cache is major: it brings general purpose programming on GPUs to a level not possible before. Memory alignment requirements are relaxed and you can actually do some efficient programming of short loops and some serial code without having too move data back to the CPU. In GT200 running serial code on the GPU is as costly as moving data back to the CPU, mostly due to memory controller design. Some pieces of code in an algorithm can't be parallelized and previous generations were very bad at dealing with such code. Moving the data back to the CPU to run them is the last thing you want to do and a big limitation to general purpose GPU computing.

While I personally don't care about ECC, I've done some scientifc programming on GPUs and I've noticed no problems even with G92 based GPUs. I'm sure someone does, as this is a requirement for most HPC machines. For those, Nvidia has that covered.
IEEE 754-2008 compliance is very important and another hindrance that was removed. The previous architecture only allowed for full compliance on 64-bit calculations - no one wants that. The loss of half the performance even on this architecure is too much. Double precision will be strictly used only where needed, which was the main reason for the inclusion of some DP units on the GT200 architecture. Sadly, they lacked full precision on 32bits which complicated comparing calculations from CPUs to calculations from GPUs, among other problems.

Just like with the GT200 chip, this new architecture is heavily built towards HPC, a market Nvidia wants to address all by itself. The compute power is there, the architecture is more refined than ever. This focus, however, may hurt the ability of Nvidia to compete in the graphics card market, just as with the GT200. The architectural changes done to support these applications increased die size and took up space from other featrues that could have been used for gaming applications. It remains to be seen if Nvidia will be able to compete with AMD on price and performance. I do expect the chip to perform at least 60% faster than the GT200 at the same clock, possibly more due to the addition of the L2 cache.
The Radeon 5870X2 is looking very able to fend off Nvidia's new card, although with all the problems that come from using multi GPU setups.
At the time the "Fermi" is released, the crown of fastest single card on the market will be something very tough to take from Nvidia. Still, AMD might be able to strike back with a cheaper, updated R800 card just like the Radeon 4890 - timing is everything and despite the somewhat under performing 5870, AMD has the advantage right now.

ASUS Sabertooth 55i Released

ASUS delivers a new high-end motherboard for LGA 1156 processors.

The new motherboard is part of the "TUF" series - The Ultimate Force - which ASUS targets at computer enthusiasts mainly - that roughly translates to a big price tag.
The higher price does bring high quality along though, as the motherboard supports SLI(licensed from Nvidia), high end components like all solid capacitors and a 16 phase CPU power supply.

Specifications:

Model	SABERTOOTH 55i
CPU	LGA1156 socket for Intel® Core™ i7/Core™ i5 Processors
Chipset	Intel® P55 Express Supports Intel® Turbo Boost Technology
Memory	4 x DIMM, max. 16GB, DDR3 2000/1800/1600/1333/1066MHz, non-ECC, unbuffered memory Dual channel memory architecture Supports Intel® Extreme Memory Profile (XMP)
Expansion slots	2 x PCI Express 2.0 x16 slots (single at x16 or dual at x8/x8 mode)
	3 x PCI Express 2.0 x1 (2.5GT/s)
	2 x PCI
Multi-GPU	Supports both NVIDIA® Quad-GPU SLI™ Technology and ATI® Quad-GPU CrossFireX™ Technology
Storage	Intel® P55 Express Chipset - 6 x SATA 3.0 Gb/s ports (black) - Intel Matrix Storage Technology supports RAID 0, 1, 5 and 10 JMicron® JMB363 SATA & PATA controller: - 1 x Ultra DMA 133/100/66 for up to 2 PATA devices - 1 x External SATA 3Gb/s port (SATA On-the-Go) JMicron® JMB322 (DriveXpert Technology) : - 2 x SATA 3.0 Gb/s ports (orange and white ) - Supports EZ Backup and SuperSpeed functions
LAN	Realtek® 8112L PCIe Gigabit LAN controller featuring AI NET2
Audio	VIA® VT2020 10-channel High Definition Audio CODEC - Absolute Pitch BD192/24 featuring ENVY HD - DTS Surround Sensation UltraPC - Supports Jack-Detection, Multi-streaming, Front Panel Jack-retasking - Optical S/PDIF out port at back I/O - ASUS Noise Filter
USB / IEEE1394a	14 x USB 2.0 ports (6 ports at mid-board, 8 ports at back panel) VIA® 6308P controller supports 2 x IEEE 1394a ports (one at mid-board; one at back panel)
TUF Exclusive Features	Power Design 12-phase CPU Power, 2-phase Memory Power, 2-phase VTT_CPU Power, Efficient Switching Power Design for 3VSB/1.8V/1.05V, TUF Capacitors and MOSFETs (certified military-standard) Thermal Solution CeraM!X Heatsink Coating Tech, CoolMem! Fan Frame, T.Probe Technology, ASUS Fan Xpert, Guardian feature MemOK!, DriveXpert, ESD Guards, EMI Eliminator
Form Factor	ATX Form Factor, 12 inches x 9.6 inches ( 30.5cm x 24.4cm )

The support to add a fan to cool memory is a worthy addition on any motherboard and ASUS has done this while still leaving some room between it and the graphics card.


The ceramic technology is mostly marketing than anything else. Some manufacturers have tried this approach a few years ago but the combination of heatpipes and more cooling surface area has won. It may result in MOSFETs a few degrees cooler but nothing spectacular.

Price for the newcomer has not been announced yet.

On The Green Camp...

This week Nvidia has been on the news twice, for rather unusual(or should I say desperate) business practices:

• Batman: Arkham Asylum, a game part of "The way it's meant to be played" program from Nvidia, only let's you enable Anti Aliasing features on Nvidia cards.
• PhysX support has been disabled on systems that also have am AMD graphics card installed, to ensure you don't buy a cheap green card just for PhysX.

I'm appalled at the first move and Batman: AA's developers should also be embarrassed to be doing this. One thing is accepting money to optimize the game for said manufacturer's graphics card, while providing "unique" features like hardware PhysX support, but cutting out a feature that's general to every graphics card out there? That's been around for years? I'm at a loss of words.

On the other hand, the PhysX support getting disabled on systems with AMD cards was delivered close to the launch of Windows 7. This is no coincidence, as Win 7 differs from Vista by allowing the OS to support two graphics cards running two different drivers. This allows the use of a cheap Nvidia card for PhysX - which is enough for most stuff - and an AMD card to drive graphics.

It is a rather strange move for two reaons:

• Nvidia can probably make money from their cheaper cards but is - or will be soon - at a loss with bigger cards that can't even compete with AMD's offerings right now.
• Game developers will see hardware support with an even narrower view, when it should have at least widen a bit given Win7 support for two vendor drivers.

This move puts further constraints on a technology that's seeing difficult enough market adoption as it is right now. One year from now, when everyone has moved to other vendor agnostic physics middleware, Nvidia will have been left with a very expensive investment in PhysX and no one to support it.
If Nvidia wants to be so aggressive regarding PhysX, it should've at least waited for decent market adoption before pulling a move like this. While I don't agree with such move, it would at least have made some sence from a business standpoint.
Perhaps it has all went down the drain already and this is a desperate move to secure some retail sales before PhysX is finally put to sleep.

Radeon HD 5850 Reviewed

Smaller, cheaper and performs admirably.

The Radeon HD 5850 is an interesting card: it features all the capabilities of the big brother but a cut down memory bandwidth, TMUs and shader units. The card only has 1440 active shader cores, 72 TMUs and memory bandwidth is provided by slower 4GHz GDDR5 chips, vs 4.8GHz for the 5870. The lower specs aren't enough to slow down performance much but the card is considerably cheaper.

Anandtech's Ryan Smith has reviewed the card, performance in Crysis Warhead is more than ok:


I had already anticipated that the card would perform well, based on AMD's Radeon HD 5870 performance that was unveiled last week. The card does indeed live up to the hype, it can around 20% of the performance of the 5870 while costing $120 less, a more palatable $259. This also means that the newcomer is still able to take on the Nvidia GeForce GTX 285 pretty fiercely. DX11 support and Eyefinity is just icing on the cake.



Power consumption is surprising: 86w less than what the Radeon 5870 churs up, which is 160w officially, vs 188w on the 5870. This card is a more sensible one than the 5870 on all levels, even here.
Temperature on the GPU is also better, at 80ºC, 9ºC less than the 5870. Although still quite high, it's at more reasonable levels.

Conclusion

AMD's Radeon HD 5850 1 GiB is an excellent card by all measures: it features DX11 support, better performance than more expensive cards from Nvidia, triple monitor Eyefinity support, excellent power consumption and a decent price tag. If you overclock it, you'll probably be able to squeeze 5870 performance out of it. It really can't get any better than this right now, and one can only wonder what would've happened if AMD had cut ridiculous ratio of shaders/bandwidth(see 5870 reviews) and delivered a more balanced card for the same $199 as the previous generation. The only problem with the card is Linux support, in which case you should go green.
The ball is in Nvidia's court and it's a very hot one.

Weekly Memory Price Update

Memory prices have risen again since last week's update, to the tune of 5% on average, as the DXI shows on the graph. DDR2 memory rose as much as 8%, with a rise of 2% accounted today. DDR3 prices have stabilized and dropped 0.5% today although they are still up around 3% since last week.
DDR2 and DDR3 1Gb chip average prices are now within 2.5% of each other with DDR3 being more expensive than DDR2.
The cost of a DDR3 2GB DIMM is of $36 in memory chips only (16 1Gb chips), to the manufacturer, and $32.8 for a 2GB DDR2 DIMM. Fully built memory modules can still be bought for less than that, but that situation is expected to change soon, if the market keeps rising as it has been lately.