Update: NVIDIA has confirmed that the Twitter account is indeed theirs, so this information is official.

With a skeptical eye towards Twitter, a post was made on the NVIDIAGeForce account 4 hours ago announcing the names of the first two GF100 cards. As we’re largely sure this is a legitimate NVIDIA account we’re going to go ahead and post this information, but please keep in mind that we have not yet been able to confirm that this is indeed an official NVIDIA posting (it’s 10PM on the West Coast).

With that out of the way, the post reads as follows:

Fun Fact of the Week: GeForce GTX 480 and GeForce GTX 470 will be the names of the first two GPUs shipped based on our new GF100 chip!

It’s a very small piece of information so we don’t have a lot of commentary here, but the names are a little bit surprising. The names are consistent with NVIDIA’s G/GT/GTX naming scheme, but we’re not quite sure what happened with the numbers. Technically speaking NVIDIA launched their 300 series late last year with the GeForce 310, an OEM-only rebadge of the GeForce 210. But we had expected that NVIDIA would fill the rest of the 300 series with GF100 and GF100-derrived parts, similar to how the 200 series is organized with a mix of DX10 and DX10.1 capable parts. Such an expectation is also consistent with the earlier rumors on the GTX 380 and GTX 360.

Instead they've gone ahead and started the 400 series of cards with GF100, not that we’re complaining. This is great news for developers and buyers since it prevents a repeat of the GeForce 4 Ti/MX situation, but it does make us wonder what the point was in burning off a whole series number with a single bottom-tier card. And although this is an article about the 400 series, we're left wondering what the purpose is of a rebadged 300 series is since that clearly has an impact on the naming of the 400 series.

At any rate, no further information was announced. We still don't know what the performance will be like or the clock speeds. It's a good bet however that the GTX 470 will have some CUDA Cores disabled, if NVIDIA's GTX 280/270/260 naming scheme is anything to go by.

AMD’s often perilous financial situation usually bears watching, but this past quarter is of particular interest. On the business side we have seen AMD and Intel settle their long-standing feud over accusations of anti-competitive behavior by Intel, which had several big outcomes for AMD this past quarter. As part of the settlement terms Intel paid AMD a cool 1.25 billion USD, officially to make up for past transgressions – and unofficially as life support for a company that lost money for the last 13 quarters.

On the technology side we’ve seen AMD’s graphics division spend an entire quarter driving their own destiny. With Fermi/GF100 delayed well in to Q1 of this year, Q4 was all about the Radeon 5000 series – and it would have been an even better quarter if only it weren’t for that pesky TSMC 40nm supply problem. AMD’s CPU division didn’t have quite as rosy of a time, but Intel’s price gaps below $200 have left AMD with a viable value market.

So with that in mind, how did AMD fare for the quarter and for the year? It’s a mixed bag, particularly if you throw out the Intel settlement. Perhaps a table will best get at this.

AMD booked the final Intel settlement at $1,242 million, giving us the difference between AMD’s actual results for the quarter and year, and what it would have been without the Intel settlement. Officially AMD booked a profit even without the Intel settlement if you go by their non-GAAP (Generally Accepted Accounting Principles) numbers, but we’re sticking with the GAAP numbers here and just yanking out the Intel settlement.

To put things in perspective, last year AMD lost $1.4 billion just on Q4’08, and $3.1B for all of 2008, so this is a massive turnaround for the company. At this point both of their core divisions are turning a small profit, and the company’s reduced exposure to the foundry business has greatly improved their bottom line. AMD took a loss of $99M in Q4 from their share of Global Foundries – so if they were able to drop the foundry business entirely, they would have likely turned a true GAAP profit. Although AMD has several problems, at the moment it’s the foundry business that continues to be doing the most harm to them.

On a broader market outlook, AMD was expected to lose some 18 cents a share this quarter (not factoring in the Intel settlement), whereas they actually only lost 8 cents a share, so the company beat the street in losing less than half as much as was expected. Certainly that’s good news for AMD, whose long streak of losing quarters has made it an unpopular item.

One thing of interest from AMD’s results was the average selling prices and gross margins. AMD doesn’t give us the exact ASP of their products, but they do compare it to previous quarters and years.

In spite of AMD’s overall better financial position, the ASPs in both the processor and graphics division has fallen both on a quarterly and yearly basis. This isn’t particularly surprising for the processor division with Nehalem regulating AMD to the sub-$200 market, but it is more surprising on the graphics side where AMD had not been able to sell a single-GPU video card for over $300 for years until now. It just goes to show you that the bulk of products really are sold at the low-end, where prices on older 4000-series products continue to slowly drop. Graphics revenue is a similar story: Up slightly on the year, but up significantly for the quarter compared both to last year and to Q3 of this year.

What is surprising is that considering AMD’s financial situation, their processor gross margin is pretty good. At 45% they’re no Intel (64.7%) but it’s the kind of margin the company needs to pay for things like R&D, particularly since they don’t move massive volumes of chips like Intel does.

Wrapping things up, the big thing for the company in 2010 is going to be that they’re finally going to escape the black hole that has been Global Foundries. ATIC – whose long-term plan has always been to completely buy-out GF from AMD – is expected to go through with that transaction soon now that Intel is no longer going to block the sale through the use of the x86 license. Since GF has been a losing proposition for AMD, in the short-term they stand to finally stop losing money on it, meaning they stand a good chance of turning at least a small profit overall since both processors and graphics are profit-generating. Of course it remains to be seen whether this is a great long-term strategy, but you can’t argue with the short-term effects on the balance sheet.

On the tech side, this quarter will see the launch of the rest of the Radeon 5000 series (Evergreen) chi p stack. 55nm products will continue to sell for quite some time, but it should be good for the ASP of the graphics division. Fuzion (CPU+GPU) products are still on-schedule for a launch late this year, so the CPU side of things should see a decent shakeup even though Bulldozer won’t make it this year. AMD mentioned that the first Fuzion products will be SOI-based, so it sounds like they’ll be making them at GF, with later products fabbed at additional foundries.

From our Radeon HD 5770 Review:

Interestingly enough, we’ve been told that the Phoenix shroud isn’t going to be sticking around for long. The first wave of cards launching today and for the near future will be using the shroud, but once AMD’s vendors begin using their own designs, AMD doesn’t expect most of the vendors to stick with the shroud. XFX has specifically been named as a party that will keep using the shroud on products, but anyone else is subject to change. With a TDP of only 108W, the Phoenix shroud is probably overbuilt and certainly more expensive than vendors would like, where mainstream products come with thinner margins. We would expect the vendors that do switch to move to more traditional dual-slot coolers, likely ones that aren’t shrouded at all and would not blow hot air outside of the case.

What AMD explained to us quickly came to pass, and once the first wave of 5770s sold out, the replacement waves started coming with coolers besides the Phoenix shroud. Since then we’ve had a number of people ask us how the later coolers compare to the Phoenix, and this is something that we can finally answer today.

XFX's 5770 Rev 2

XFX – who AMD named as one of the companies who would be selling the Phoenix along with other cooler designs – was kind enough to send over one of their alternative 5770 models for this cooler evaluation. Our particular interest in this card is that it uses the same egg-shrouded cooler that is appearing on the majority of 5770s seen in retail, making it the de-facto standard 5770 cooler. For the sake of simplicity, we are calling this the 5770 Revision 2 cooler. In XFX’s case, this is sold alongside cards using the Phoenix and another card using an open cooler similar to the Rev 2.

When first looking at the Rev 2, looks can be deceiving. While it does use an egg-shaped shroud very similar to that found on the 5750, the actual HSF unit is entirely different, making the similarities merely superficial. Ultimately the 5750’s HSF was a simple aluminum heatsink sitting on top of the GPU, with a fan sitting on top of the heatsink. For the 5770, a much more elaborate HSF is used, consisting of a copper block with a pair of heatpipes embedded. These heatpipes then run the entire circumference of a circular aluminum heatsink that sits on top of the copper block. Finally, sitting in the middle of the heatsink (and not above it) is the fan.

A side view of the Rev 2 cooler

The original Phoenix cooler

In short, in spite of the visual similarities to the 5750’s cooler, the Rev 2 cooler entails a much larger heatsink using heatpipes and better fan placement. If you’ve ever seen Zalman’s VF900 GPU cooler, then the design is very similar to that.

Overall the Rev 2 is very similar to the Zalman VF900 (image courtesy Zalman)

With the change in coolers comes some space savings for the card. As the Phoenix shroud added another half-inch to the total length of the 5770, it pushed an 8.25” design out to 8.75”. With the Rev 2 cooler there is no overhang, bringing the card in at just the PCB length of 8.25”. This also means that the 6-pin PCIe power plug is no longer partially hidden by the shroud, making it easily accessible on the Rev 2.

We should note that while the cooler has changed, the PCB has not. The PCB on this card is exactly identical to the PCB on our 5770 reference card. In fact the only difference besides the cooler is the use of Samsung GDDR5 rather than Hynix GDDR5. In benchmarks there is absolutely no difference in performance between our Rev 2 card and our original reference card. So the only practical difference is the cooler.

That brings us to what kind of difference the change in coolers brings. When we first discussed the 5770, we noted that the Phoenix was probably overbuilt for the relatively low 108W TDP of a 5770. Indeed that appears to be the case, as you can see with the results of our Rev 2 cooler.

We’ll start with a look at temperatures. Our expectations for the Rev 2 cooler were for higher overall temperatures since its design means that it has to recycle some hot air rather than blowing it entirely out of computer like the Phoenix cooler does, and it’s here where we found our first surprise. Rather than being warmer than the Phoenix, the Rev 2 is cooler (if ever so slightly) in both idle and load situations. At a 2C and 1C different at idle and load respectively it’s not a huge difference, but it’s an extremely notable difference since the otherwise simpler Rev 2 cooler is clearly keeping up with and otherwise outperforming the Phoenix.

Our other surprise came in our noise tests .Surely the Phoenix would be quieter thanks to its shrouding, right? Wrong we were once again. In fact the difference is even more pronounced than the temperature differences. The Rev 2 cooler not only manages to stay quieter than the rest of the computer (something the Phoenix can’t claim) but at load the difference becomes 6dB. While the Phoenix cooler is by no means a loud cooler, the Rev 2 in comparison is whisper-silent. Even compared to itself, the Rev 2 is only 3dB louder under load than it is while idling.

Final Thoughts

Frankly, based on this data we have a hard time justifying the Phoenix over the Rev 2 cooler. The Rev 2 makes for a smaller board, a slightly cooler GPU, and a significantly quieter video card. Thusly the only advantage the Phoenix cooler has is that it completely exhausts all hot air, which in the case of our well ventilated Thermaltake case isn’t doing it any favors. Unless this card is placed in a case with extremely poor airflow, we can’t think of a situation where the Phoenix cooler is superior (and in which case, we’re left wondering whether there would be enough fresh air to satisfy the Phoenix regardless).

The Phoenix cooler may be the better looking (and 5800-series matching) cooler, but ultimately it’s not the best cooler. AMD and their partners were wise to ditch the Phoenix for the Rev 2 in most of their design, as it offers better thermal and acoustic performance, not to mention a lower resulting price (we’d peg the difference at about $10 retail). In fact at this point we’re left wondering why they launched the Phoenix at all –ultimately it would have been all the more impressive for the 5770 and cheaper for consumers if they had just launched with the Rev 2 cooler in the first place.

On a quick note this morning, along with yesterday’s release of the Catalyst 9.12 drivers, AMD has also published a 9.12 hotfix driver that has added a couple of interesting things.

Along with refreshing their line of OpenCL-capable drivers (OpenCL is still not in the mainline driver), AMD has added support for Crossfire Eyefinity. We first saw Crossfire Eyefinity in our Radeon 5970 review, where the feature was enabled solely for the 5970 so that the complete card could be used for Eyefinity. At the time AMD had promised that they would be enabling Crossfire Eyefinity for true Crossfire-paired cards soon, and this is the first step of fulfilling that promise. The need for AMD to whitelist games for Crossfire Eyefinity has not changed, so while it works on Crossfire-paired cards, it still does not work for all games.

The second interesting addition is support for DisplayPort audio. Although we tend to think of DisplayPort as a replacement for DVI rather than HDMI, technically it can serve as a replacement for both. Several of you have been asking us if the 5000 series supported DisplayPort Audio, and we did not have a good answer until now. If you do have a DisplayPort setup that supports audio, then this driver will allow you to finally get audio out of that DisplayPort on your 5000 series card.

On the performance front we have heard that these drivers may offer some respectable performance boosts for the 5000 series, but this isn’t something we’ve had a chance to test. The timing would be good however, since the supply of RV700 cards (4870/4850) have completely dried up, so buying a cheap 4870 in place of a 5700 series card has ceased to be an option.

• "The total number of x86 instructions is well above one thousand" (!!)
  
• "CPU dispatching ... makes the code bigger, and it is so costly in terms of development time and maintenance costs that it is almost never done in a way that adequately optimizes for all brands of CPUs."
• "the decoding of instructions can be a serious bottleneck, and it becomes worse the more complicated the instruction codes are"
• The costs of supporting obsolete instructions is not negligible. You need large execution units to support a large number of instructions. This means more silicon space, longer data paths, more power consumption, and slower execution.

We just got off the phone with Nick Knupffer of Intel, who confirmed something that has long been speculated upon: the fate of Larrabee. As of today, the first Larrabee chip’s retail release has been canceled. This means that Intel will not be releasing a Larrabee video card or a Larrabee HPC/GPGPU compute part.

The Larrabee project itself has not been canceled however, and Intel is still hard at work developing their first entirely in-house discrete GPU. The first Larrabee chip (which for lack of an official name, we’re going to be calling Larrabee Prime) will be used for the R&D of future Larrabee chips in the form of development kits for internal and external use.

The big question of course is “why?” Officially, the reason why Larrabee Prime was scrubbed was that both the hardware and the software were behind schedule. Intel has left the finer details up to speculation in true Intel fashion, but it has been widely rumored in the last few months that Larrabee Prime has not been performing as well as Intel had been expecting it to, which is consistent with the chip being behind schedule.

Bear in mind that Larrabee Prime’s launch was originally scheduled to be in the 2009-2010 timeframe, so Intel has already missed the first year of their launch window. Even with TSMC’s 40nm problems, Intel would have been launching after NVIDIA’s Fermi and AMD’s Cypress, if not after Cypress’ 2010 successor too. If the chip was underperforming, then the time element would only make things worse for Intel, as they would be setting up Larrabee Prime against successively more powerful products from NVIDIA and AMD.

The software side leaves us a bit more curious, as Intel normally has a strong track record here. Their x86 compiler technology is second to none, and as Larrabee Prime is x86 based, this would have left them in a good starting position for software development. What we’re left wondering is whether the software setback was for overall HPC/GPGPU use, or if it was for graphics. Certainly the harder part of Larrabee Prime’s software development would be the need to write graphics drivers from scratch that were capable of harnessing the chip as a video card, taking in to consideration the need to support older APIs such as DX9 that make implicit assumptions about the layout of the hardware. Could it be that Intel couldn’t get Larrabee Prime working as a video card? That’s going to be a big question that’s going to hang over Intel’s heads right up to the day that they finally launch a Larrabee video card.

Ultimately when we took our first look at Larrabee Prime’s architecture, there were 3 things that we believed could go wrong: manufacturing/yield problems, performance problems, and driver problems. Based on what Intel has said, we can’t write off any of those scenarios. Larrabee Prime is certainly suffering from something that can be classified as driver problems, and it may very well be suffering from both manufacturing and performance problems too.

To Intel’s credit, even if Larrabee Prime will never see the light of day as a retail product, it has been turning in some impressive numbers at trade shows. At SC09 last month, Intel demonstrated Larrabee Prime running the SGEMM HPC benchmark at 1 TeraFLOP, a notable accomplishment as the actual performance of any GPU is usually a fraction of its theoretical performance. 1TF is close to the theoretical performance of NVIDIA’s GT200 and AMD’s RV770 chips, so Larrabee was no slouch. But then again its competition would not be GT220 and RV770, it’s Fermi and Cypress.

Next, this brings us to the future of Larrabee. Larrabee Prime may be canceled, but the Larrabee project is not. As Intel puts it, Larrabee is a “complex multi-year project” and development will be continuing. Intel still wants a piece of the HPC/GPGPU pie (least NVIDIA and AMD get it all to themselves) and they still want in to the video card space given the collision between those markets. For Intel, their plans have just been delayed.

The Larrabee architecture lives on

For the immediate future, as we mentioned earlier Larrabee Prime is still going to be used by Intel for R&D purposes, as a software development platform. This is a very good use of the hardware (however troubled it may be) as it allows Intel to bootstrap the software side of Larrabee so that developers can get started programming for real hardware while Intel works on the next iteration of Larrabee. Much like how NVIDIA and AMD sample their video cards months ahead of time to game developers, we expect that Larrabee Prime SDKs would be limited to Intel’s closest software partners, so don’t expect to see much if anything leak about Larrabee Prime once chips start leaving Intel’s hands, or to see extensive software development initially. Widespread Larrabee software development will still not start until Intel ships the next iteration of Larrabee, if this is the case.

We should know more about the Larrabee situation next year, as Intel is already planning on an announcement at some point in 2010. Our best guess is that Intel will announce the next Larrabee chip at that time, with a product release in 2011 or 2012. Much of this will depend on what the hardware problem was and what process node Intel wants to use. If Intel just needs the ability to pack more cores on to a Larrabee chip then 2011 is a reasonable target, otherwise if there’s a more fundamental issue then 2012 is more likely. This lines up with the process nodes for those years: if they go for 2011 they hit the 2^nd year of their 32nm process, otherwise if they launched in 2012 they would be able to launch it as one of the first products on the 22nm process.

For that matter, Since the Larrabee project was not killed, it’s a safe assumption that any future Larrabee chips are going to be based on the same architectural design. The vibe from Intel is that the problem is Larrabee Prime and not the Larrabee architecture itself. The idea of an x86 many-cores GPU is still alive and well.

On-Chip GMA-based GPUs: Still On Schedule For 2010

Finally, there’s the matter of Intel’s competition. For AMD and NVIDIA, this is just about the best possible announcement they could hope for. On the video card front it means they won’t be facing any new competitors through 2010 and most of 2011. That doesn’t mean that Intel isn’t going to be a challenge for them – Intel is still launching Carkdale and Arrandale with on-chip GPUs next year – but they won’t be facing competition at the high-end too. For NVIDIA in particular, this means that Fermi has a clear shot at the HPC/GPGPU space without competition from Intel, which is exactly the kind of break NVIDIA needed since Fermi is running late.

Raja and I have been working on a Holiday Guide the past couple of weeks and hopefully we will complete it this week. Our emphasis has been on finding components that offer a great bang for the buck even though they might not be the absolute best in their class. While we will offer our opinions on what is best in class, our focus has been on balanced performance, support, and features versus cost. A really good example is the ASRock X58 and P55 Extreme boards that offer a great set of features and performance for excellent pricing in each category. While they will not satisfy the needs of extreme overclockers, for the other 98% of us, they offer a really great value. Just like the Gigabyte GA-EP45-UD3P, MSI 790FX-GD70, and ASUS M4A77TD Pro have in their respective categories.

That said, there are a lot of great choices currently in the lower end market, especially in the AMD 785G camp. The 785G boards are just terrific values for building a SOHO centric platform that will be primarily used for office applications, Internet, communications, and casual gaming. This is especially true when paired up with an Athlon II based processor. Really, current Intel S775 processors in the sub $100 market teamed with a G41 based board just do not have what it takes to compete with the AMD products in this price sector. Intel has a competitive answer coming early next year, but until then AMD is the wise choice.

J&W is one of our favorite second tier suppliers and they have a winner in the 785G market with their new JW-A785GMT-Extreme board. This product compares very favorably with the Gigabyte and ASUS 785G offerings and actually makes for a great SFF gaming platform for those who just use a single graphics card. Our testing to date with a Phenom II X4 965BE and AMD HD 5850 leads us to believe this a great combination for the gamer centric user on a budget or as an all around gaming/media center with a HD 5770 and Athlon II X2 550BE.

The JW-A785GMT-Extreme features a great five-phase PWM design that fully supports 140W processors, Realtek ALC 888 8-channel HD audio, Realtek RTL8111D Gigabit LAN, IEEE-1394a via the JMicron JMB381 chipset, 12 USB 2.0 ports, HDMI/DVI/VGA output, 16GB DDR3 support, six SATA 3G ports via the SB710, and on-board graphics capability thanks to the update HD 4200 IG engine with full DX10.1 and UVD 2.0 support. J&W also throws in 128MB of DDR3 side port memory. There are three fan headers, with the CPU fan header having speed and temperature settings.

The layout of the board is very good and includes an LED Debug display along with power on and reset buttons.

Initial performance ranks right with the Gigabyte and ASUS 785G boards when it comes to overclocking. Our results with the 965BE were superb as the board allowed stock voltage overclocks to 3.87GHz with our GSkill Ripjaws DDR3-1600 4GB C8 kit reaching DDR3-1720 at 9-9-9-24 2T timings.

Bumping up the CPU VID to 1.43V resulted in 24/7 stable 4.07GHz clocks under Windows 7 64-bit with NB speeds at 2.86GHz, something we have not been able to do under a 64-bit operating system until now. The BIOS is geared for the overclocker with significant settings available for memory and voltage tuning. That said, memory timings are not as aggressive past 1600 as the Gigabyte or ASUS boards but this board typically clocked our CPU about 70MHz higher at like voltage settings. We will take a further look at this board in the near future but for now, anyone looking for a very high quality 785G motherboard should place the JW-A785GMT-Extreme at the top of their list.

In our Radeon HD 5970 review, we ran in to some issues when trying to overclock the card to 5870 speeds of 850MHz/1200MHz. At the time this is something we attributed to the VRMs, meanwhile AMD suggested that it was cooling related, and that we should manually increase the fan speed.

As it turns out, we were both right, we just didn’t have the tools at the time to properly identify and isolate the issue. Late last week we got our hands on a beta version of Everest Ultimate, which added preliminary support for the 5970. With that, we could read and log the voltages and temperatures of the various components of the 5970, and properly isolate the issue.

From that, we’ve discovered a few interesting things about the 5970. Let’s start things off with the cooler removed from the 5970.

We’ve gone ahead and circled the VRMs in red. There are 9 altogether; 6 on the right side, and 3 near the left side of the card. We aren’t able to track down what each specific VRM is connected to, but we believe that each GPU is attached to 3, each GPU’s RAM is attached to 1, and finally the PLX PCIe bridge is attached to 1. Regardless, pay attention to the location of these VRMs for later discussion.

As we previously noted in our 5970 review, when overclocked the card was throttling down in two cases. One was when running OCCT/FurMark, members of AMD’s “power virus” list by virtue of the fact that they put a card under a greater load than AMD believes to be realistically possible. Our 5800 series cards never throttled under these applications, so to see the 5970 throttle here was a bit surprising but not wholly unexpected.

The second case was using Distributed.net’s pre-release GPU client for use with AMD’s GPUs. Since this is a real program, this was absolutely unexpected, and is what instigated our look in to the matter.

In both cases, the key was the overall load on the GPU cores, and consequently the amount of power required to drive the GPUs. When a bank of VRMs reached roughly 120C (this being averaged among all the VRMs in that bank), overcurrent protection kicked in and throttling began. In the case of FurMark this was very quick and even at 100% fan speed the cooler could still not keep the VRMs cool enough to allow full-time 850MHz operation. The Dnet client on the other hand was much slower to ramp up, and we ultimately found that 70% fan speed was enough to keep our hottest bank of VRMs below the threshold, stabilizing at 116C.

Notably, during this whole period the GPU cores themselves stayed at or under 94C, which is still a few degrees below their own throttle point. AMD’s fan quickly ramped up, and in our testing it only needed to go to 59%. So if the cores did get hotter there was still plenty of room to go with the fan.

This brings us to our first point of concern for the 5970, which is the fan speed. Clearly it’s adequate for the GPU cores themselves, but we cannot find any proof that the fan speed is adjusted based on the temperature of the RAM or the VRMs. If the fan speed were to ramp up in the case of near-critical temperatures in the VRMs, then the Dnet client likely would have ran without an issue the first time, as this would have pushed the fan to 70%.

We asked AMD about whether the fan speed is affected by VRM temperatures at all, but we didn’t receive a response. This isn’t particularly surprising since post-launch periods are a good time to take a vacation and there’s a holiday this week for their American employees, but it means we couldn’t get a confirmation of our assumption. So for the time being, we’re working on the assumption that only GPU core temperatures drive fan speed.

It also bears mentioning that the 5970 gets quite a bit louder when the fan goes up to 70%. We went ahead and captured the noise data for it at 70% and 100%, which is in the chart below. At the 70% fan speeds needed to run the Dnet client at 5870 speeds, you’re looking at 70dB, which is quite a bit louder than the fan noise at stock speeds. It is in fact uncomfortably loud by this point.

Our second point of concern goes beyond just the fan, and is the overall cooling of the VRMs. When we looked at our Everest logs after running the Dnet client, we noticed something interesting with respect to which VRMs were overheating. The VRM bank attached to GPU 1 was some 25C hotter under load, but it wasn’t GPU 1 that was the hottest. GPU 2 was consistently a couple of C warmer. We don’t believe this to be in error, so to understand why this is, we refer back to our disassembled 5970.

As the fan is on the right, the right side of the heatsink the vapor chamber dumps its heat in to is going to be cooler than the left side by the virtue of the fact that the left side is effectively using the already hot-air of the right side to cool. The heatsink and vapor chamber mitigate this some, but the right side of the card – and consequently the right GPU– should be cooler than the left side. This leads us to believe that GPU 1 is the right GPU, and GPU 2 is the left GPU.

This is important since if we look at the VRMs, the VRMs feeding GPU 2 sit under the vapor chamber, while the VRMs feeding GPU 1 (along with the RAM and PCIe bridge) are not. We haven’t been able to fully dissect the cooler, but the VRMs on the right side sit right underneath the fan, and we don’t believe there to be a significant heatsink in the metal bar that sits above them. So while the VRMs feeding GPU 2 are being cooled by the vapor chamber, the VRMs feeding GPU 1 are only being cooled by the heat dissipation properties of a metal bar.

From this, we can conclude that the VRM banks are receiving wildly different amounts of cooling. The VRMs on the right side are not cooled nearly as well as those on the left and as a result the card is being held back by the VRMs on that right side. To that extent, we believe that if all the VRMs received the same level of cooling as the VRMs on the left side, then the card would have no problem maintaining 5870 speeds while running the Dnet client, and likely even FurMark. It’s also worth noting that all the 5800 series cards share the design of placing the VRMs under a metal bar under the fan, but the 5970 seems to suffer more for it compared to the 5800 series.

Finally, there’s the matter of whether this is even going to matter for most users. After catching the VRMs hitting 120C under the Dnet client, we went looking at other applications and games to see where else the card was throttling. The result of that inquiry was that we couldn’t find anything else that could match the Dnet client in total load. The Dnet client is a bit of a special case here, since crunching encryption keys makes exceedingly good use of the 5-wide SIMD design in the 2000-5000 series cards. When we took a look at something similar to the Dnet client, in this case the Folding@Home GPU client, we couldn’t break 100C. The significance of that result remains to be seen though, since the Folding@Home GPU client hasn’t been optimized for the 5800/5900 series yet like the Dnet client has. Our ultimate concern is that this card is going to repeatedly fall flat on its face at 5870 speeds with more GPGPU applications as OpenCL and DirectCompute take off, and the number of such applications bloom.

Meanwhile in games it was a similar story. Crysis and the STALKER benchmark are two of the most demanding games we’ve tested on the 5970, and in both cases the VRMs again peaked at near 100C. As games aren’t going to hammer the SIMDs like GPGPU applications do, the power load from games should be lower than for GPGPU applications.

As far as our opinion on the 5970 is concerned though, this doesn’t change anything. While we’ll buy AMD’s “power virus” rationale for FurMark and OCCT, the Dnet client is not a power virus. It’s a real application, one that AMD even used in their 5800 presentation back in September. Thus as far as we’re concerned, our 5970 is only good for 775MHz, the lowest clock speed where the VRMs stayed under 120C. Granted, AMD will never officially promise that the 5970 can reach 5870 speeds, but based on how the 5970 was promoted and presented the fact of the matter is that the card can’t meet its advertised capabilities – this card is clearly meant for 5870 clockspeeds.

With that in mind, we’ll end on two thoughts. The first of which is that in spite of our experience, for pure gaming scenarios we don’t have any data to bring in to doubt the idea that the card can run at 5870 speeds without throttling. So long as you only intend to play games, those speeds should be fine.

Our second thought is that cards from vendors with custom overclocking utilities will be better able to maintain 5870 speeds at all times. These are cherry-picked chips, so there’s no reason why they absolutely need 1.1625v core voltage to run at 850MHz; we suspect that they could do with less. Since voltage is our main enemy here, even a small drop in voltage should have a noticeable impact on VRM temperatures. But you’re going to need a utility with a full suite of voltage options to take advantage of that.

As we mentioned in our previous edition of The Cable Chronicles, Microsoft and CableLabs have come to an agreement to allow the installation and use of CableCARDs on unapproved and non-OEM systems, allowing for the wider proliferation of CableCARD equipped HTPCs beyond the handful of OEM systems that CableLabs had previously approved. With Windows 7 implementing a complete DRM scheme for TV tuners – the Protected Broadcast Driver Architecture – computers running Win7 would be the first to be able to take advantage of these relaxed restrictions.

The limitation at the time was that computers did not come CableCARD-capable out of the box. A Digital Cable Advisor tool was to be released by Microsoft, which would check a computer to make sure it meets all of MS’s and CableLab’s requirements, before going ahead and enabling CableCARD access. That tool was supposed to be released in time for Win7’s launch, but it ended up being AWOL at the time. As Microsoft was not going to publish the complete system requirements for using a CableCARD, this tool was the only way to find out the system requirements.

The tool was finally released this weekend, allowing us to get an idea of what the system requirements are.

The tool comes as a Windows Media Center Extra, and needs to be downloaded, installed, and used from within Windows Media Center. If you don’t regularly run MCE, then it probably won’t show up in the Extras Gallery, as MCE picks up on it when it regularly checks in for updates. To force MCE to check in, go to Settings -> General -> Automatic Download Options -> Download Now.

Once installed. It shows up in your Extras Gallery.

When activated, the tool will analyze the system to determine if it meets MS/CableLab’s requirements, reporting back whether the computer passes or fails the requirements, and if it fails, offers a short explanation why.

Digging through the tool’s support files, for most of the system requirements the tool appears to just be looking at the Windows Experience Index of the computer. The chart below lists the scores we’ve found, and the text attached to them if the computer fails

In spite of the attached text, the tool doesn’t appear to actually be looking for specific pieces of hardware. We ran this test on a single-core computer with 1.5GB of RAM, and it passed anyhow, as it met the WEI scores required. So our best guess is that only WEI scores matter here.

The second test tests the Content Protection (Read: DRM) capabilities of the hardware. It appears to be looking for the various MCE DRM components (PBDA and PlayReady), HDCP support, and DXVA acceleration of MPEG-2. We suspect the DXVA check is just a warning rather than a hard error, but we don’t have any appropriate hardware to test this.

From our testing, we believe the tool will fail if HDCP support is not present in the video card, regardless of if a digital connection is being used. The big question we have, and one we haven’t been able to find an answer to, is whether HDCP is required for digital connections (e.g. an older TV using HDMI), as we don’t have a monitor on-hand that doesn’t support HDCP.

The analog situation looks better. On a PC hooked up to a TV via Component, it passed the check and was allowed to enable CableCARD support.

However once a PC passes the test, the final screen leaves us scratching our heads as to whether the tool actually knows what it’s doing.

The screen is a list of recommended hardware:

• Sufficient disk space on your PC to record TV. For example, you’ll need about 160GB to record 20 hours of TV
• A minimum of 4GB of RAM
• A digital output connected to an HDCP-compliant display to receive HD-quality video.

We’re not sure where the disagreement lies. It could very well be that the tool isn’t checking for an HDCP-attached monitor and is just blindly approving everything so long as there’s HDCP support, or that it’s downscaling content going out an analog output (ala the Image Constraint Token on Blu-Ray), or it could be that this is an empty recommendation. This tool was supposed to clear up confusion about what’s required for CableCARD use, and it hasn’t really achieved that.

Finally, in the strangest occurrence, one of our systems was already authorized according to the tool. The Core i7 rig we use for benchmarking shows up as authorized, even on a fresh install of Windows 7. This is expected behavior for OEM systems (or rather, motherboards) that were previously approved by CableLabs, but we have no idea why this would be showing up on an Asus Rampage II Extreme.

Finally, along with the release of the advisor tool, the updated firmware for the ATI TV Wonder Digital Cable Tuner needed to go with these looser restrictions has finally been released. As this was originally intended as an OEM-only product, the only place we’ve seen it thus far is at Dell’s website, where they want $210 for a tuner. Since it’s a single-stream tuner, you’ll want 2 (or more) to watch & record multiple channels. Other vendors will have CableCARD tuners out, including multi-stream tuners, but not until next year. For the time being, CableCARD on Win7 comes with a high early adopter tax.

It’s not often we write about prices going up.

Last week there was a rumor going around that AMD intended to raise prices on the 5800 series. At the time we wrote this off as yet another highly-speculative rumor based on shaky evidence. Official price hikes are virtually unprecedented, after all.

Then things changed.

We’ve talked previously about TSMC – the foundry both NVIDIA and AMD GPUs are manufactured at – having yield issues with their 40nm process. This first surfaced with the Radeon 4770, which at the time of its introduction was being built while TSMC’s yields were below 40%, and this coupled with its popularity made for a significant shortage around its introduction. TSMC continued to improve their yields, and by the time of the Radeon 5000 series launch, AMD told us that they weren’t concerned with yields. As of this summer, TSMC was reporting yields of 60%.

On Friday the 30^th, Digitimes broke the word that TSMC’s yields were back down to 40%. This we believe is due to issues TSMC is having ramping up overall 40nm production, but regardless of the reason it represents a 33% drop in usable chips per 40nm wafer. When you’re AMD and you’re rolling out a top-to-bottom 40nm product line in a 6 month period, this is a problem.

The 5870 and 5850: Out Of Stock Everywhere

When the 5800 series launched, we knew supplies would initially be tight, but we had been expecting them to pick up. With these yield problems, that has not happened. Instead 5800 cards continue to be out of stock near-universally, even with the fact that most OEMs have yet to start using these cards. AMD’s current 5800 supplies are being exhausted just by Dell and self-builders.

Meanwhile NVIDIA started the end-of-life process for the GTX 200 series some time ago, with production of the GT200 GPU ramping down. So NVIDIA doesn’t need to play pricing games with AMD, as they’ve already planned on selling out anyhow.

With low supplies, no (single-GPU) performance competition, and no price competition, you have the perfect storm for a price hike.

All of a sudden that rumor about an AMD price hike became far more realistic. Checking around, virtually none of the 5800 series cards are listed at their MSRP. Although they’ve continued to be in low supply since launch, it’s only recently that there’s been a breakaway from the $379 and $259 MSRP of the 5870 and 5850 respectively.

After our latest round of price checks, we talked with AMD about the situation and asked them if there was any truth to the rumor of an official price hike. The news is not good: 5850 prices are officially going up. AMD is citing supply issues of components (including memory) amidst the heavy demand for the 5850, and ultimately deciding to pass the cost on to the consumer. Meanwhile there is no official price hike for the 5870, although it’s going to be affected by any increased component costs just as much as the 5850.

Bear in mind that the 5850 is also a special case. AMD can’t keep the 5870 in stock, never mind the 5850. For every fully-functional Cypress die they get, the only reasonable option is to build a 5870 out of it. The only things that should be going in to the 5850 are dice with a defective functional unit, making them ineligible for use in a 5870. Without an idea of how many harvestable dice TSMC is spitting out, we can’t get any real numbers, but the most reasonable assumption is that most of them are either fully-functional or unsalvageable, so we expect AMD and their vendors to be producing many more 5870s than they will 5850s. In other words, the 5850 shortage is going to be worse than the 5870 shortage.

The result of all of this is, is that regardless of the reason, there’s a price hike across the entire 5800 series – an official hike for the 5850, and an unofficial hike for the 5870. AMD has not established a new MSRP for the 5850, but their best guess is $20; ultimately it’s up to vendors (and retailers) to determine pricing. It’s hard to get an idea of what the price is going to be on a card that’s always out of stock, but an MSRP of $279 is probably too low. $300 (or more) is a more realistic target for the 5850. As for the 5870, it seems to be settling around $400.

Our best guess is that these new prices will continue through the rest of the year, even if supplies pick up as TSMC gets their yields back in order. Without any serious competition from NVIDIA, these cards can be priced anywhere between $300 and $500 based on performance alone, and no one has any incentive to keep prices down so long as 5800 series cards keep flying off of the shelves. It’s Economics 101 in action.

We can’t say we’re happy with any of this, but we can’t accuse AMD and their vendors of acting irrationally here. It’s a lousy situation for consumers, but that’s a shortage for you. When has there ever been a good shortage?

Finally, with these price hikes, our product recommendations are changing some. The 5870 is still the card to get if money is no object, but the 5850 is far more situational since it’s no longer the great bargain it once was. We can get 1GB 4890s for $170 right now, which have become downright cheap compared to our projected $300 for a 5850. Certainly the 5850 whips the 4890 by upwards of 40%, not to mention DX11 and Eyefinity, but at that level it’s commanding a 75% price premium. It’s a $300 card and performs accordingly, but don’t break the bank in order to get a 5850 at these prices.

If you want a cheap 5800 series card, then it looks like you’re out of luck until 2010.

The Biggest 5850/4890 Performance Gap

"Xen 3.2.0 which can be found in the newest Novell SLES 10 SP2, is capable of running Windows 2003 R2 under heavy stress."

For those of you sticking with Vista, Microsoft has finally officially released DirectX 11 for Vista, after having spent the last couple of months in beta. This final release looks to be the same as the last beta released earlier this month.

The update is KB971512, which is being released as part of a larger Platform Update for Vista that includes a few other things that are being backported for Vista. Vista SP2 is the prerequisite, so if you aren’t already on SP2 you’ll need to update.

All of these updates should be available on Windows Update.

We ran a quick sanity check on our Vista install from our Win7 Performance Guide from earlier this week, and there are no surprises. Just like with DX10, DX11 titles (all 2 of them ) perform the same between the two OSes. In this case we’re using BattleForge, along with Unigine’s DX11 Heaven benchmark (it’s synthetic, but pickings are slim for DX11). We’ve also thrown in Crysis for good measure, although it's not a DX11 title.