AMD's Radeon HD 4870 X2 - Testing the Multi-GPU Waters
by Anand Lal Shimpi & Derek Wilson on August 12, 2008 12:00 AM EST- Posted in
- GPUs
These Aren't the Sideports You're Looking For
Remember this diagram from the Radeon HD 4850/4870 review?
I do. It was one of the last block diagrams I drew for that article, and I did it at the very last minute and wasn't really happy with the final outcome. But it was necessary because of that little red box labeled CrossFire Sideport.
AMD made a huge deal out of making sure we knew about the CrossFire Sideport, promising that it meant something special for single-card, multi-GPU configurations. It also made sense that AMD would do something like this, after all the whole point of AMD's small-die strategy is to exploit the benefits of pairing multiple small GPUs. It's supposed to be more efficient than designing a single large GPU and if you're going to build your entire GPU strategy around it, you had better design your chips from the start to be used in multi-GPU environments - even more so than your competitors.
AMD wouldn't tell us much initially about the CrossFire Sideport other than it meant some very special things for CrossFire performance. We were intrigued but before we could ever get excited AMD let us know that its beloved Sideport didn't work. Here's how it would work if it were enabled:
The CrossFire Sideport is simply another high bandwidth link between the GPUs. Data can be sent between them via a PCIe switch on the board, or via the Sideport. The two aren't mutually exclusive, using the Sideport doubles the amount of GPU-to-GPU bandwidth on a single Radeon HD 4870 X2. So why disable it?
According to AMD the performance impact is negligible, while average frame rates don't see a gain every now and then you'll see a boost in minimum frame rates. There's also an issue where power consumption could go up enough that you'd run out of power on the two PCIe power connectors on the board. Board manufacturers also have to lay out the additional lanes on the graphics card connecting the two GPUs, which does increase board costs (although ever so slightly).
AMD decided that since there's relatively no performance increase yet there's an increase in power consumption and board costs that it would make more sense to leave the feature disabled.
The reference 4870 X2 design includes hardware support for the CrossFire Sideport, assuming AMD would ever want to enable it via a software update. However, there's no hardware requirement that the GPU-to-GPU connection is included on partner designs. My concern is that in an effort to reduce costs we'll see some X2s ship without the Sideport traces laid out on the PCB, and then if AMD happens to enable the feature in its drivers later on some X2 users will be left in the dark.
I pushed AMD for a firm commitment on how it was going to handle future support for Sideport and honestly, right now, it's looking like the feature will never get enabled. AMD should have never mentioned that it ever existed, especially if there was a good chance that it wouldn't be enabled. AMD (or more specifically ATI) does have a history of making a big deal of GPU features that never get used (Truform anyone?), so it's not too unexpected but still annoying.
The lack of anything special on the 4870 X2 to make the two GPUs work better together is bothersome. You would expect a company who has built its GPU philosophy on going after the high end market with multi-GPU configurations to have done something more than NVIDIA when it comes to actually shipping a multi-GPU card. AMD insists that a unified frame buffer is coming, it just needs to make economic sense first. The concern here is that NVIDIA could just as easily adopt AMD's small-die strategy going forward if AMD isn't investing more R&D dollars into enabling multi-GPU specific features than NVIDIA.
The lack of CrossFire Sideport support or any other AMD-only multi-GPU specific features reaffirms what we said in our Radeon HD 4800 launch article: AMD and NVIDIA don't really have different GPU strategies, they simply target different markets with their baseline GPU designs. NVIDIA aims at the $400 - $600 market while AMD shoots for the $200 - $300 market. And both companies have similar multi-GPU strategies, AMD simply needs to rely on its more.
Facebook
Twitter
RSS
93 Comments
View All Comments
helldrell666 - Wednesday, August 13, 2008 - link
Anandtech hates DAAMIT.Have you checked the review of the 4870/x2 cards at techreport.com?The cards scored much better than here.
I mean In assassins creed it's well know that ATI cards do much better than nvidia's
It seems that some sites like: anandtech,tweaktown"nvidiatown",guru3d,hexus... do have some good relations with NVIDIA.
It seems that marketing these days is turning into fraud.
Odeen - Wednesday, August 13, 2008 - link
With the majority of the gaming population still running 32-bit operating systems and bound by the 4GB RAM limitation, it seems that a 2GB video card (that leaves AT MOST 2GB of system RAM addressable, and, in some cases, only 1.25-1.5GB of RAM) causes more problems than it solves.Are there tangible benefits to having 1GB of RAM per GPU in modern gaming, or does the GPU bog down before textures require such a gargantuan amount of memory? Wouldn't it really be more sensible to make the 4870x2 a 2x512MB card, which is more compatible with 32-bit OS'es?
BikeDude - Wednesday, August 13, 2008 - link
Because you can't be bothered upgrading to a 64-bit OS, the rest of the world should stop evolving?A 64-bit setup used to be a challenge. Most hw comes with 64-bit drivers now. The question now is: Why bother installing a 32-bit OS in new hardware? You have lots of Win16 apps around that you run on a daily basis?
Odeen - Thursday, August 14, 2008 - link
Actually, no. However, a significant percentage of "enthusiast" gamers at whom this card is aimed run Windows XP (with higher performance and less memory usage than Vista), for which 64-bit support is lackluster.Vista 64-bit does not allow unsigned non-WHQL drivers to be installed. That means that you cannot use beta drivers, or patched drivers released to deal with the bug-of-the-week.
Since a lot of "enthusiast" gamers update their video (and possibly sound) card drivers on a regular basis, and cannot wait until the latest drivers get Microsoft's blessing, 64-bit OS'es are not an option for them.
I'm not saying that the world should stop evolving, but I am looking forward to a single 64-bit codebase for Windows, where the driver signing restriction can be lifted, since ALL drivers will be designed for 64-bit.
rhog - Wednesday, August 13, 2008 - link
Poor Nvidia,DT and Anandtech have their heads in the sand if they don't see the writing on the wall for nvidia. The 4870X2 is the fastest video card out there, the 4870 is excellent in its price range and the 4850 is the same in its price range. The AMD chipsets are excellent (now that the SB750SB is out) and Intel Chipsets have always been a cut above also they really only support Crossfirenot SLI. Why would anyone buy Nvidia (this is why they lost a bunch of money last quarter,no surprise). For example, to get a 280SLI setup you have to buy an Nvidia chipset for either the AMD or Intel processors (the exception may be skulltrail ofr intel?) Neither Nvidia Chipset platform is really better than the equivalents from Intel or AMD so why would you buy them? Along with this Nvidia is currently having issues with their chips dying. Again why woudl you buy Nvidia? I feel that the writing is on the wall Nvidia needs to do something Quick to survive. What I also find Funny is that many people on this site and on others said AMD was stupid for buying ATI but in the end it seems that Nvidia is the one who will suffer the most. Give Nvidia a biased review they need all the help they can get!
helldrell666 - Wednesday, August 13, 2008 - link
AMD didn't get over 40% of the X86 market share when they had the best cpus "athlon 64 /x2".AMD knew back then that beating INTEL "to get over 50%" of the
x86 market share" wont happen by just having the best product.
Now,INTEL has the better cpu/cpus and 86% of the cpu market.
So,to fight such a beast with a huge power you have to change the battle ground.
AMD bought ATI to get the parallel processing technology.Why?
To get a new market where there's no INTEL.
actually, that's not the exact reason
Lately nvidia introduced cuda,"the parallel processing for general processing "And as we saw,The parallel procesing is much faster than the x86 processing in some taskes.
Like in transcoding the 280gtx with a 933 Giga flops/cycle of processing power {processing power is the number of constructions or flops a gpu can handle in a single cycle} was 14 times faster than a QX 9770 clocked at 4GHz.
NVIDIA claims that there are much more areas where the parallel processing can take over easily.
So,We have two types of processing and each one has it's adavantages over the other.
What i meant by changing the battle ground wasn't the gpu market.
AMD is woking at these seconds on the first parallel+x86 processor .
A processor that will include x86 and parallel cores working together to handle everthing much faster than a x86 processor at least in some tasks.So the x86 core will handle the tasks that they are faster at,and the parallel cores will handle tha tasks that the're faster at.
Now,Intel claims that geometry can be handled better via the x86 processing.
you can see it as a battle ground between INTEL and NVIDIA but,It's actually where AMD can win.
I think that we're going to see not only x86+parallel cpus but also
x86+parallel gpus.Easily put as much processing power of each type as it needs to make a gpu or a cpu.
I think that AMD is going to change the micro processing industry to where it can win.
lee1210mk2 - Wednesday, August 13, 2008 - link
Fastest card out - all that matters! - #1Ezareth - Wednesday, August 13, 2008 - link
I wouldn't be suprised to see the test setup done on a P45 much like Tweaktown did for their 4870X2 CF setup. Doesn't anyone realize the 2 X PCIe X8 is not the same as 2 X PCIe X16? That is the only thing that really explains the low scoring of the CF setups here.Crassus - Wednesday, August 13, 2008 - link
I think this is actually a positive sign when viewed from a little further away. Remember all the hoopla about "native quad core" with AMD's Phenom? They stuck with it, and they're barely catching up with Intel (and probably lose out big in yield).Here Sideport apparently doesn't bring the expected benefits - so they cut it out and moved on. No complaints from me here - at the end of the day the performance counts, not how you get there. And if disabling it lowers the power requirements a bit, with the power draw Anand measured I don't think it's an unreasonable choice to disable it. And if it makes the board cheaper, again, I don't mind paying less. :D
And if AMD/ATI choses to enable it one or two years down the road - by then we've probably moved on by one or two generations, and the gain is negligible compared to just replacing them.
[rant]
At any rate, I'm happy with my 7900 GT SLI - and I can run the whole setup with a 939 4200+ on a 350 W PSU. If power requirements continue to go up like that, I see the power grid going down if s/o hosts a LAN party in my block. We already had brownouts this summer with multiple ACs kicking in at the same time, and it looks like PC PSUs are moving into the same power draw ballpark. R&D seriously needs to look into GPU power efficiency.
[/rant]
My $.02
drank12quartsstrohsbeer - Wednesday, August 13, 2008 - link
My guess (before the reviews came out) was that the sideport would be used with the unified framebuffer memory. When the unified memory feature didn't work out, there was no need for it.I wonder if the non functioning unified memory was due to technical problems, or if it was disabled for strategic reasons... ie since this card already beats Nvidias, why use it. This way they can make it a feature of the firegl and GPGPU cards only.