ASUS ROG SLI HB Bridge review; Do you really need it?

by

Introduction

Basically, the acronyms translate to ASUS Republic Of Gamers (ROG) Scalable Link Interface (SLI) High-Bandwidth (HB) Bridge. Before we dive into the details of the SLI HB Bridge, let’s take a step back and understand the way SLI works in the first place. SLI is a technology developed by Nvidia that allows us to link two or more GPUs together to produce a combined, more powerful single output. In other words, when one graphics card becomes insufficient, get another similar one, and (almost) double the graphics performance of your desktop. Bear in mind though, running SLI does not add up the VRAM of the GPUs. So if you’re running an 8GB card on SLI with another 8GB card, your VRAM output will still be 8GB.

Now here’s the important bit. SLI is basically an algorithm of parallel computing which on its own is a HUGE chapter in computer science. But down the crux, multiple electronic entities (in our case, the graphics card) break down the execution of processes by splitting the workload among available machines. So while we have multiple GPUs, we always have the master and its slave(s). The workload is evenly spread out across all GPUs, but only the master is responsible of rendering the output. This rendering frame is delivered to the master via the SLI bridge. Which is the component that we are putting to test here.

That’s how it looks like in its final form basically

The thing is, there are pros and cons to doing SLI configurations. I’ve had my fair share of system instability due to drivers not fully optimised for SLI setups. But don’t blame Nvidia on that. Before this, you were able to run up to 4 GPUs on SLI, which Nvidia killed since its 10-series GPUs. Now if you’re gonna game, at the most, you can run 2 GPUs on SLI. You see, graphics-card have rather complex graphics API including, but not limited to, the latest Vulkan API and DirectX 12. These are still not matured enough in their development, so opening up 3-way and 4-way is pretty much pointless if things are going to break more than they make.

We’re also going to compare the results with ASUS’ ROG Enthusiasts SLI Bridge which we shall call the “standard SLI Bridge” here on. The biggest difference is that the HB Bridge has a 650MHz Pixel Clock while the standard SLI Bridge has a 400MHz Pixel Clock. This means, passing our final rendering frames to the master is going to be quicker. The question is, how quick? We’ll have to really push the system’s graphics to its maximum load to find this out. Take note that this bridge only works with several high-end graphics that you can check here.

Our Machine Setup

This piece of hardware requires so much juice, even our Pokderig is not powerful enough to fully test it. So I brought in my very own LaFerrari to test this out. Our specifications are as following:

CPU: Intel Core i7 6900K 8C/16T 3.2GHz
Motherboard: ASUS Rampage 5 Edition 10
GPU: 2 x ASUS Titan X SLI
RAM: 16GB (4 x 4GB) 3000MHz Avexir Blitz
CPU Cooling: Cooler Master Nepton 240
Storage: 3 x 256GB Samsung 850 Evo SSD
1 x 2TB Western Digital Black
PSU: Cooler Master V1000
Display: ASUS ROG Swift PG278Q WQHD (2560 x 1440), 144Hz, G-Sync

I’ve set the CPU clock at 4.0GHz, and turned off Turbo Boost and Intel SpeedStep to have a consistent CPU clock speed. Both Titan X GPUs are running on stock clock and GSync is enabled. Since we want to see how quickly does the master GPU deliver the final rendered image, we are only concerned about FPS (frames-per-second) here. The monitor is set to 144Hz mode with GSync enabled so our frames rendering will be consistent with the GPU’s output. This is a beast setup, cutting out as much possible setbacks as possible. Only then are we going to be able to see whether the SLI HB Bridge works or not.

SLI HB Bridge Testing Methodologies

So this time around, testing the SLI HB Bridge would not be much like other technical reviews that we have done. Running benchmark tools will not really give us much significant results. Instead, my tests are going to be based on real world scenarios – where it will actually be used. I’m going to run the setup through several games, and push things harder in steps to see where it starts showing actual differences. Basically, this setup is so powerful, running games at Ultra settings is a piece of cake for it even at 4K. Hence, our tests will begin from QHD (2560 x 1440), then we push it to 1.5X DSR factor at 3620 x 2036 resolution, followed by 2X DSR factor at 5K/UHD+ (5120 x 2880) resolution.

So I measure the lowest FPS and the highest FPS of each of these settings at random intervals, then measure the average of high and low and finally the average of all readings.

Counter Strike GO (CS:GO) Test

You might think that CS:GO is not a demanding game, but jack up all the settings to their max, and you could actually see your system jigs at random frames. For CS:GO, my tests involved around the same map, but different scenarios. Using different weapons, pooling more players in one place and firing their weapons, throwing smoke grenades and molotovs, and several other frame intensive processes. So these are the results that I obtained.

As you can see, there is literally no difference between the bridges when it’s run at QHD or even 1.5X DSR. But you can see quite a notable difference at UHD+ resolution. It looks like the HB bridge is indeed doing its job. Then again, the lowest framerate we are seeing here is 197, at UHD+, for the standard SLI bridge. 197 is an insane framerate by itself already. There’s no room for complain at all, is there?

Overwatch Test

So we take it to a slightly heavier test with Overwatch. What I like about Overwatch testing is that when things get chaotic mid-game, you can really see noticeable behavior in frame rates change. Imagine Reinhardt with his shield open, a junkrat spamming his bomb on top of it, with a Pharah in the sky, while Lucio heals the defending team, when suddenly the attacking Zarya casts its ultimate right in the middle, as Reaper ghosts into the center for his turn and a wild Hanzo casts his dragons right in the middle of all of it. That can be a very chaotic scene and your FPS counter will be fluctuating like its trying to figure out what to do with its sorry life.

The settings are EPIC on Overwatch with everything turned on. Anti-Alias, Reflections, Texture, all set to ultra and the game surely does look beautiful. But how does it perform?

You still can’t tell much difference here. One thing about measuring FPS in a game, like I mentioned earlier is the struggle to get consistent data. As for the 1.5X DSR for example, the score was 153.6 on standard SLI bridge and 153.3 on the HB SLI Bridge, making the data look weird. At UHD+, you can see a slight improvement on the HB SLI Bridge because at this point, the frame rates became a lot more consistent as the GPU was really pushing it hard to deliver the frames. So I thought, hey, why not take it a step further?

I set the Overwatch to render 200% resolution of whatever is being projected. Which means, at 5K, the frames were doubled, and then squeezed into my now puny sized 1440p display. This is how you bring the Titan X on SLI, down to its knees! The pixels are so densely merged, the output starts looking blurry like it has a soft focus touch. Beautiful for movies, but ugly AF for games. lol. Wanna see the results?

As I ran these tests, the two Titan X were screaming at the top of their lungs, spinning the fans at their maximum speed, begging me for mercy as I, with my evil laugh capture every blood-clad frame squirting out of these once dominating monsters. At this stage, ever single frame counts and you want to squeeze every single bit out of it. Between 19 fps and 29 fps, there’s a whole painful 10 fps difference that can turn the experience into a matter of life and death. In the game that is.

Tom Clancy’s The Division

Finally, on our third test, we put Tom Clancy’s The Division to test. Division is a very graphics intensive game and has some very cool features that can put the GPU to work aggressively. Once again, everything is set to ultra on this game. Let’s have a look at the results.

 

So, what do we have here. The Division shows the most significant differences in terms of the results. Even at 1.5X DSR, the difference is quite prominent with a gap of 17 fps. At UHD+, we are already approaching 60fps which, while is still absolutely perfect, is now the borderline. The HB SLI Bridge does show its prowess here afterall.

User Experience

Well, I can’t just end the review without mentioning its experience, right?

The first thing that you need to know before purchasing the SLI HB Bridge, is its compatibility. Obviously, you would need an Nvidia card because AMD is now going bridge-less for their Crossfires. You can view the full list here, but so far, at the time of writing, these cards are proven to work:

  • Titan X
  • ASUS ROG Strix GeForce GTX 1080
  • ASUS Turbo GeForce GTX 1080
  • ASUS GeForce GTX 1080 Founders Edition
  • ASUS ROG Strix GeForce GTX 1070
  • ASUS Expedition GeForce GTX 1070
  • ASUS Dual GeForce GTX 1070
  • ASUS Turbo GeForce GTX 1070
  • ASUS GeForce GTX 1070 Founders Edition

It should also work with other similar GPUs from different manufacturers, if that’s what you want to do.

This is different from what ASUS had with their previous ROG Enthusiast SLI Bridge that had support for up to 4 GPUs.

The second thing that you need to consider before purchasing the SLI HB Bridge, is its size. There are two variants available; 3-slot and 4-slot. Depending on your motherboard, you would require only one of them and it’s important to know which one. As in the case of my rig running on ASUS Rampage 5 Edition 10, I needed the 4-slot version.

Of course, we definitely have to talk about the RGB LED on the ASUS ROG SLI HB Bridge. Compared to the standard SLI Bridge which would only lit in red LED, the ASUS ROG SLI HB Bridge has an RGB glowing logo on it. It has touch sensitive panel or you can use it with AURA Sync. Yep, you can only use ONE of the methods. If you want to use the touch sensitive panel, you can use it out of the box. But if you want to use it with AURA Sync, you will have to plug in the RGB cable to the SLI HB Bridge, and connect the other end to the RGB header on your motherboard. Worry not, the cable has an extra RGB port in case you need to plug in another AURA Sync compatible hardware. Another thing you need to keep in mind is that the current RGB plugs are notchless. It means, you might end up plugging it the other way around. So if the RGB on your ASUS ROG SLI HB Bridge is not working, invert the plug and it should work. Speaking from experience here. lol.

ASUS ROG SLI HB Bridge’s RGB effect on touch. #asus #asusrog #sli #hb

A post shared by Pokde.net (@pokdedotnet) on

Conclusion

There you go! I hope you got some insights from my review of the ASUS ROG SLI HB Bridge. Does it work? Definitely! Do you need it? If you’re not gaming on anything above 4K, then its pointless. Then again, if you are running either 1070 or 1080 or Titan X in SLI and you’re not gaming on 4K and above, you’re in the wrong league. That’s not what you do with such powerful components. And if you do have an SLI setup of either of those cards, another extra RM200 wouldn’t hurt your pocket and is definitely something you need for all the raw power under the hood. I love the performance, but sorry ASUS, looks wise, I still prefer the design of the older standard bridge 😛