Introducing the GeForce 500M Family (and the 485M)

When NVIDIA announced their complete 400M lineup last September, we were pleasantly pleased: top to bottom DX11 support, with reasonable performance and the option for laptop manufacturers to use Optimus Technology to provide a nice balance of battery life and performance. There was one problem with that last item, however: the high-end Clarksfield CPUs used in most gaming notebooks didn’t have integrated graphics, so you were left with a notebook continually powering a discrete GPU. The result was less than stellar battery life but good performance, illustrated by notebooks like the ASUS G73Jw.

At the lower end of the performance spectrum, Arrandale offerings provided plenty of performance for mainstream users, but no one wanted to take a chance and pair a fast dual-core Arrandale (i.e. i7-640M) with a high-end GPU (i.e. GTX 460M) and enable Optimus graphics switching technology. That’s a shame, as it would have been an interesting alternative, but with such a short shelf life it probably doesn’t matter. As we illustrated earlier this week, Sandy Bridge has completely altered the mobile landscape. Now you can get near-desktop performance, quad-core processing with Hyper-Threading, decent integrated graphics, and battery life that’s none too shabby—all at roughly the same price as the old Core i7-740QM! SNB’s integrated graphics allow for a combination “have your cake and eat it too” notebook that combines a fast CPU with a fast GPU and doesn’t have to throw battery life under the bus. That’s the theory at least; we’ll have to wait for actual hardware to see who manages to pull it off.

With the Intel HD Graphics 3000 shipping in all mobile Sandy Bridge processors, Intel also effectively killed off anything below the 420M level. Yes, you can still get DirectX 11 support and better drivers from NVIDIA and AMD GPUs, but for those who only need “good enough” graphics there’s no real need to pay more. GT 425M still ended up being 55% to 96% faster than HD 3000 on average (at low and medium details, respectively), but while twice as fast is good, that 55% number is a bit of a concern. Why not widen the gap a bit by increasing clock speeds and memory bandwidth? Enter the GeForce 500M line.

Let's get this out of the way first: this is just rebranding the old 400M series and chaning the clocks. Everyone clear on that? AMD is launching their 6000M parts as well, except they failed to provide us with any information ahead of time. I guess NVIDIA wanted to do a number update to keep people excited, but anyway....

Starting at the high-end, we also get a new 400M part to go with the 470M and 460M. The GTX 485M now uses the full GF104 chip, so you get 384 CUDA cores, and the clock speed is 575/1150MHz for the core/shaders, with 750MHz GDDR5 (3.0GHz effective). That puts it ahead of the old 480M (352 cores at 425/850/600 Core/Shader/RAM clocks), and it should have better power characteristics as well. To be specific, the GTX 485M ends up with 48% more computational power than the 480M and 25% more memory bandwidth. The 470M and 460M will continue their roles as high-end parts.

The first of the new 500M chips is the GT 555M, based on the GF106. Unlike the old GT 445M, the GT 555M is always a 192-bit interface, although there’s still the potential to curtail memory bandwidth by using DDR3 instead of GDDR5. So, while there’s plenty of confusion with the overlapping 400M and 500M parts, at least in this one instance we have some clarification. Core clocks are the same as the old GT 445M, but the GDDR5 memory is now specced to run at 900MHz (3.6GHz effective) instead of 800MHz.

The next three cores are all just a speed-bumped version of existing 400M parts. 550M, 540M, and 525M replace the 435M, 425M, and 420M respectively. All have 96 CUDA cores, just like the old 400M chips they replace, but memory bandwidth is up from 800MHz (25.6GB/s effective) to 900MHz (28.8GB/s)—a 12.5% increase. Clock speeds are also up, in this case it’s a 14% increase for the 550M vs. 435M, 20% for 540M vs. 425M, and 20% for the 525M over the 420M—not too shabby. As before, everything that doesn’t start with “GTX” lacks SLI support, which is fine by us; we’d rather have a single fast GPU over two slower GPUs in SLI just for ease of configuration if nothing else.

More New Chips!

The parts we’ve discussed so far are all clearly superior to the outgoing 400M models, but as we’ve already shown with the GTX 485M, there are new chips that aren’t part of the 500M family. Rounding out the mobile GPUs launching today, we have three more options—none of them particularly desirable as far as we’re concerned.

I’m not sure what purpose these new parts serve, other than giving notebook OEMs some “new” discrete GPUs that they can foist off on unsuspecting customers. Sure, the 520M ought to beat Intel’s HD Graphics 3000, but if you’re running where it makes sense (i.e. low detail) the 520M is going to offer less than the GT 420M, thanks to the reduced shader counts and half the memory bandwidth. Given the 420M and 425M already turned in similar performance results—an indication that most games are memory bandwidth limited—that could prove disastrous at anything more than low detail, and if you’re only gunning for low quality in the first place you can probably survive on the IGP.

Anyway, the 520M replaces the GT 415M, a product which we haven’t yet been able to test. The 410M appears to be the same basic idea, only without support for GDDR5. Both chips have the same pinout, but in looking at the chip shots from NVIDIA, and the chip appears a lot smaller, so it may me that GF119 is a native 48 cores rather than half being disabled.

Finally, we also have a GeForce 315M part, which keeps the flame alive for the old G 310M by changing the clock speeds to 606/1212/790. Ugh. Notice how we say “changed” rather than “improved”: those clocks compare to 625/1530/790 on the G 310M in the ASUS U30Jc, or 608/1468/620 on the ASUS UL80Jt. I’m sure you’ll get 1GB of slow memory standard, though, which doesn’t really do much for you. Given what we’ve shown with Sandy Bridge’s IGP, you’d really have to be desperate to want the 315M.

But let’s make it clear: NVIDIA isn’t creating these low-end parts without reason; there are OEMs out there who actually intend to use these GPUs. It’s almost like a throwback to the old S3 Virge days, where we all joked about them being “3D Decelerators”. If the G 310M performance is anything to go by, Sandy Bridge will typically offer better performance than the 315M. NVIDIA still has better driver support for games, so you can make a case for the 520M/410M. Those should at least match SNB graphics performance, and probably surpass it—especially the lower clocked HD 3000 offerings found in LV/ULV chips—but the old GT218 core really needs to be put out to pasture.

The other argument in favor of the 315M and 410M is that they’re extremely cheap to produce, which lets NVIDIA get hardware into just about any level of laptop hardware. I suppose that if you’re not doing Sandy Bridge, the 315M might still hold some merit. It does after all provide hardware accelerated H.264 decoding and better-than-Arrandale graphics. It might also end up in some netbooks, although NG-ION is basically the same chip and already covers that market. We never did get the GT 415M for testing, and it’s not in any US-bound laptops to our knowledge, but some of the other world markets have different tastes and it probably showed up in Asia or Europe. Hopefully that’s the case for the 410M and 315M as well, but I’m still skeptical that there’s much point in keeping something like the 315M around in the current laptop marketplace.

It’s Not Just About Games

NVIDIA would also like to let everyone know, once again, that it’s not just about games. They have CUDA, DirectX 11, OpenCL, and PhysX support, along with various 3D modes (including 3D TV Play). We also have HTML5 enabled browsers supporting GPU acceleration, and other browsers (i.e. Firefox 4) adding WebGL support. Most upcoming browsers are supposed to leverage the power of the GPU a lot more than they do today, making for a more interesting Internet. Let’s take a few of these items in turn, though.

The jury is still out on CUDA encoding vs. Intel’s Quick Sync (note that my quality with CUDA wasn't nearly so bad using MediaEspresso, so Anand may have encountered more of an Arcsoft Media Converter issue), but even if QS comes out ahead Optimus should at least allow NVIDIA to do the sensible thing and use the CPU where it’s better, or their GPU when it’s better. OpenCL is a bit of a gray area as well, since it’s possible to run OpenCL code on a CPU. We don’t have any good OpenCL tests right now, and we can't run it on an Intel CPU, so we’ll just leave that on the table. Even so, Quick Sync might make one of the most common uses of OpenCL redundant.

That brings up another item we should point out as well: unlike desktop Sandy Bridge, nearly all notebooks will get Intel Quick Sync support. The desktop P67 doesn’t work with integrated graphics, and it’s the chipset to get if you want overclocking support—at least until Z67 shows up and provides IGP + overclocking. That means you can’t use Quick Sync (at least not right now—maybe a driver or some other hack could provide access to the necessary CPU/GPU features). In contrast, all Sandy Bridge laptop chipsets support integrated graphics and thus Quick Sync. There’s no PM67 to muddy the waters (at least not yet). The one exception would be notebooks that forego switchable graphics and only operate off a discrete GPU, and Quick Sync would make that a very bad design decision in our book. As I said in the Mobile Sandy Bridge review, SNB is a much bigger deal for laptops than desktops in my opinion.

Next up, DX11 and PhysX are certainly wins, albeit small ones, for NVIDIA. There are games that make reasonable use of both options, and even if you end up disabling the features in most titles for performance reasons (i.e. laptop GPUs are still usually too slow, so you have to choose between higher quality graphics, DX11, and PhysX), such features are still value add-ins. We’ll lump them into the same category as better overall gaming performance and drivers for now.

And that leaves the final two items, HTML5 and WebGL. Internet Explorer 9 Beta and Firefox 4 are designed to leverage the power of your GPU as well, with IE9 focusing more on HTML5 and FF4 adding WebGL into the mix. We decided a quick test of a few current platforms would set the stage, and NVIDIA was even kind enough to list some suggestions in their reviewers’ guide. We went to the WebVizBench site (don’t even try that in something like Firefox 3.6!) and ran their benchmark at 1600x900 on IE9. Our test laptops were the new Sandy Bridge system from Compal/Intel, and the ASUS N53JF sporting an Arrandale CPU/IGP with GT 425M dGPU. We tested the ASUS laptop on both integrated graphics as well as the GT 425M. CPU load was minimal during the tests, so this does appear to be more of a graphics test. Here are the results.

Oh, snap! The HTML5 test on Arrandale appears to work as expected, offering more than double the performance when you switch from Arrandale's HD Graphics to the GT 425M. The problem is that Sandy Bridge with HD Graphics 3000 actually came in slightly ahead of the GT 425M in IE9. Granted, the SNB platform has a significantly more powerful CPU, but as a graphics test we were expecting more.

We also tried the Flight of the Navigator WebGL demo in Firefox 4 Beta 8 and Chrome 9 Beta (measuring performance with FRAPS), but the results on our Optimus system indicate it never tried to use the dGPU. Yes, WebGL, Chrome 9, Firefox 4, and Internet Explorer 9 are all still beta, so we’ll just let things sit and stew for now. We’ll revisit this topic once we have some 500M hardware in our hands, hopefully with non-Beta browsers as well, and perhaps NVIDIA can wipe the egg off their face with a more convincing win next time.

Examining the 2011 Mobile Graphics World

With the launch of Sandy Bridge, we said there was an upheaval in the mobile landscape, and we stand by that. Sandy Bridge is a bigger deal for notebooks and laptops than it is for desktops, for several reasons. First, CPU performance improved substantially more than on the desktop. Second, Intel put a decent amount of work into HD Graphics 3000, and all mobile CPUs are getting the full complement of hardware. Finally, all of the mobile chipsets support integrated graphics, so there's no PM67 getting in the way of Quick Sync. That's the good news.

As impressed as we were by the Sandy Bridge IGP, we do have to take a step back and examine what it is Intel has created. In terms of hardware, HD 3000 looks to match up rather nicely against the GeForce 320M, making it the fastest IGP currently available (at least until we see AMD's next IGP). Compatibility is also far better than with previous Intel graphics solutions...but that's not saying much. We managed to run 24 games through benchmarks in our launch articles, and we still came out with at least six games that had some sort of error. This ranged from flickering polygons to missing textures to downright horrible performance--the last coming in the only OpenGL game we happened to test. The same suite of games ran on AMD and NVIDIA hardware without a single glitch that we noticed.

Intel has also been pretty good about patching their drivers to get games running once they're notified of a problem, at least if it's the press doing the notification. (Mafia II would have joined the "does not run" list if it hadn't been for a late driver update.) However, there should be a whole suite of testers at Intel that are doing nothing other than running every single title released in the past ten years to make sure their drivers work. I suspect that as much as Intel's graphics division has improved since Arrandale, there's still plenty of work yet to be done. NVIDIA has mentioned in the past that they now have more software developers than hardware people, and we know they also have employees who work at game/application developers--as in, NVIDIA pays people who end up working at some other company most of the time. I've seen quite a few "The Way It's Meant to Be Played" titles over the past few years, and plenty of "Get in the Game" offerings from AMD/ATI as well; I have yet to see a single game touting Intel graphics.

So, what I'm getting at is there's a question of compatibility that's still largely unanswered. Are there other titles that we didn't test yet where Intel's IGP still fails? Undoubtedly. What about brand new games launching post-SNB, when all the excitement has died down--will they get updated Intel drivers to make them work properly? I can't answer that question definitively, but I can tell you that the first retail Sandy Bridge laptop I review is going to end up running a selection of games that we don't normally benchmark, purely to see how far the compatibility testing has gone. AMD and NVIDIA have pretty much figured out how much work graphics hardware and drivers requires; Intel's Larrabee 1.0 project may have been canceled more for software/driver issues than anything wrong with the hardware. And there's still the whole topic of DirectX 11, OpenGL, and OpenCL support--one Intel hasn't even touched yet, the second appears iffy, and the third is apparently being internally evaluated.

So today NVIDIA announced their new GeForce 500M parts. Yesterday (with a lot less information coming our way), AMD launched the Radeon 6000M lineup. Both offer plenty of options and good performance, but so far we haven't seen anything from AMD to match Optimus. Their switchable graphics is fine if you have a pure AMD ecosystem (IGP and dGPU), but pair an AMD dGPU with an Intel CPU+IGP and you now have drivers from two different vendors, and they don't generally play well together. You can read the following in every AMD mobile driver update: "The following notebooks are not compatible with this release.... Switchable Graphics enabled notebooks using Intel chipsets." Yes, that's a concern for anyone that wants to play games released during the coming year or two! That means we'd stick with discrete-only AMD GPUs on Intel laptops for the time being.

There are several areas that are important when you're looking at mobile graphics: performance, battery life impact, cost, and drivers.  Intel does very well on the battery life and cost areas, but they lag behind on performance and are definitely our last choice when it comes to drivers. AMD's mobile graphics get the performance and drivers aspects right, cost is obviously higher than IGP (at least for their discrete GPUs), but battery life takes a hit. Yes, the hit is smaller now than in the past, but you're still losing mobility. Finally, there's NVIDIA that gets the drivers, performance, and battery life elements, but again cost is higher (similar to AMD's GPUs). If you're in the market for a mainstream laptop or notebook that can last a while, right now NVIDIA has the edge.

Wrap-Up

As usual with a “launch” of notebook graphics, we are sadly lacking in test hardware. We hope to have plenty of notebooks ready for review in the coming weeks, and we need to take a closer look at AMD’s Mobility Radeon 6000 lineup as well. Sandy Bridge is the big news in laptops and notebooks now, though, and NVIDIA's goal is pretty clear:

NVIDIA and Intel are friends these days.

The past year has certainly been successful for NVIDIA’s mobile team. In late 2009, we had our first encounter with NVIDIA’s switchable graphics, courtesy of the ASUS UL80Vt. We liked it but the lack of regular driver updates was a splash of cold water in the face, and let’s be honest: even in late 2009 the GeForce G 210M was nothing to write home about. It managed to run most games at minimum detail and 1366x768, but asking for more than that was stretching things. The potential of switchable graphics was certainly enticing, though--get dedicated GPU performance when you need/want it, but don't sacrifice battery life for regular use.

In early 2010, NVIDIA called me and told me they wanted to fly me and a few other editors in for a Deep Dive on something they were calling Optimus. One of their PR guys told me he just referred to it internally as “the Jarred Edition of our mobile platform,” which of course got me thinking of the possibilities. While we’re still waiting for mobile GPUs that give us near-desktop performance while drawing less than 50W (which is basically what Sandy Bridge i7-2820QM does for CPUs), Optimus is certainly proving useful. Combined with the revamped Verde Driver Program, NVIDIA has become an easy mobile GPU recommendation.

ASUS’ first foray into higher-performance laptops with Optimus came with the N61Jv, and it was new enough at the time to garner a Silver Editors’ Choice award. The ASUS U30Jc also earned an Editors’ Choice—this time Bronze—and was a great replacement for the previous UL80Vt recommendation. Alienware updated the M11x to include Optimus in July, addressing one of our biggest concerns with the design and also earning a Silver award. The ASUS N82Jv continued the pattern of solid performance, but we pined for a decent LCD and withheld an Editors’ Choice award, stating that a larger battery and good LCD were the only things keeping it from Gold. It took two more months, but Dell finally delivered what we wanted with the XPS 15 L501x and Optimus GT 420M graphics—with the upgraded B-GR 1080p panel, it delivered on all fronts. (Unfortunately, the 1080p upgrade disappeared a month later, but hopefully it will come back with a Sandy Bridge refresh.)

You’ll notice a pattern in the above, and other successful laptops like the Clevo B5130M and ASUS N53Jf used Optimus as well (not to mention the Sony VAIO Z that we never did get a chance to review). All told, NVIDIA reports over 100 design wins for GeForce in 2010, with over 50 designs leveraging Optimus. In fact, the only laptops to get an Editors’ Choice award last year that didn’t have Optimus came from ASUS: the Eee PC 1001P gave us a great LCD to go along with Intel's Atom N450, while the ASUS G73Jh (later updated to GTX 460M with the G73Jw) provided another good LCD with potent gaming performance for $1500. NVIDIA is poised to continue their mobile wins in 2011, with 60% more (over 200) designs from all the major OEMs set to come out.

It’s not hard to figure out why everyone is on the Optimus bandwagon these days. While AMD has some good graphics hardware, the lack of an Optimus equivalent for Intel CPUs means you get better battery life with similar performance from NVIDIA—or you get something with an AMD GPU and switchable graphics like the HP Envy 13, and you’re left out in the cold on driver updates. That’s a problem for Intel CPU+AMD GPU, and you can get around it by using AMD CPU+GPU; unfortunately, AMD is sadly quite a ways behind Intel in CPU performance and battery life. We’ll see what Brazos can do for the netbook market soon enough, and Llano is coming later this year, but for high performance it will be very difficult to surpass Sandy Bridge. Take all of that and for the next six months, it looks like the majority of midrange and higher laptops will very likely use NVIDIA with Sandy Bridge.

The way I see it, there are really only a few reasons to dislike Optimus. The first is simple: you use Linux on your laptop and you want switchable graphics. Well, even that area is undergoing some change, and the Linux community may yet provide a solution—right now there are some laptops where you can switch between the IGP and discrete GPU in the BIOS. The other reason to dislike Optimus is similar, in that you’re heavily dependent on NVIDIA to keep updating the drivers and game/application lists. I’ve encountered a few instances where programs didn’t behave quite the way I’d like (i.e. they wouldn’t use the dGPU when I wanted, or they would use the dGPU when I didn’t want it), but most of these are simply in need of bug fixing, kind of like SLI support needs updated profiles to work sometimes. I suppose we could also state that sending the frame data over the PCIe bus uses more system resources—you use about 1GB/s for a 120Hz 1080p signal—but this so far has mostly affected games and benchmarks where you hit ultra-high frame rates. (As an example of this, the Wings of Fury test in 3DMark03 clocked 286FPS on 425M Optimus compared to 363FPS on HD 5650; normally, the HD 5650 is only about ~10% faster than the GT 425M, and since those numbers are using the AMD P520 CPU the NVIDIA chip would usually come out on top.)

None of these items are enough for us to recommend against Optimus, and while some people will feel differently it's pretty clear that notebook manufacturers like what they're seeing. Unless you're into the idea of disposable laptops, you probably want your system to last for several years. With more people purchasing laptops than desktops these days, greater flexibility is useful, and the best way to get that is to have a decent GPU in the mix. With Optimus, NVIDIA enables laptop manufactures to add a GPU without a ton of extra work (the displays are all still driven off the IGP), and they can turn it off when you don't need it. Users get the best of both worlds: performance and battery life, on demand. Manufacturers also avoid some of the technical support calls asking "why doesn't game X run on my system?" Telling them that their integrated graphics isn't fast enough or compatible doesn't win customer loyalty, even if it's true. Unless you're sure you don't need graphics, pairing Sandy Bridge with an Optimus GPU is still the most sensible choice in my opinion.

Looking forward to 2011, we now have the option to get everything we always wanted from a desktop replacement. Quad-core Sandy Bridge bested the fastest Clarksfield notebooks by over 50% in pure CPU performance, and it can still get down to ~10W power draw for a 17.3" notebook. Give us that sort of notebook with a good GPU (GT 555M or GTX 460M would be a good place to start!) and you have a notebook that truly can replace most modern desktops. Of course, while notebook manufacturers are doing the above, please quit with the lousy LCDs. Tablets are now shipping with IPS displays; can the laptops and notebooks get some luvin' as well? Also, stop with the glossy plastics, give us decent keyboards, and stop using 48Wh batteries in 15.6" and larger laptops! It's really not too much to ask.