Gigabyte Radeon RX Vega 56 Gaming OC 8G Review

Cooling & Noise

Cooling Solution & Backplate

There’s a direct relationship between power consumption and waste heat, and it's the thermal solution's job to cope with the latter. This is exactly where Gigabyte's card reveals the compromises made to keep costs low, even if the cooler still mostly works well.

Since the backplate is made of relatively thin aluminum, material is embossed in certain places to improve stiffness and add a bit of visual flair. A flattened heat pipe is glued to the backplate as well. Its purpose is to dissipate waste heat from the aforementioned MOSFETs across the plate more effectively. Other components on the PCB's back side are cooled directly using thermal pads.

Gigabyte coats the inside of the backplate in black, so thermal energy from components on the PCB is absorbed more easily. As you can see, pads help cool the area opposite AMD's Vega 10 GPU, along with nearby capacitors.

Cooling System Overview
Type
Air cooling
GPU CoolingHeat pipe direct touch
Aluminum sink
Cooling FinsAluminum, vertical orientation
Narrow configuration; partially inclined
Heat Pipes2x 8mm + 3x 6mm
Copper composite material
VRM CoolingGPU and memory VRM via cooling frame
RAM CoolingMemory cooling of HBM2 modules via heat pipe
Fans2x 9.5cm fans (10cm opening), 11 blades
Semi-passive control
BackplateAluminum
Cooling function with heat pipe and thermal pads

Two 8mm and three 6mm heat pipes made of composite material are responsible for transporting heat from the GPU and distributing it through finned areas of the sink. Heat sinks over some of the voltage converters also help prevent problematic hot-spots from developing.


Although we've seen many negative examples of heat pipe direct touch cooling, this approach can certainly be effective if it's implemented properly.

Our overlay shows that, in this case, Gigabyte's solution is a good fit. The heat pipes are flattened no more than necessary to achieve complete and functional coverage. As a result, the temperature range difference for the GPU, memory, and hot-spot are only 2°C above what we measured from Sapphire's vapor chamber-based cooler. That's extraordinary for a normal heat sink.

Fan Curves &  Noise

Semi-passive operation is implemented though an on-board controller, meaning that software like WattMan still reports a rotational speed even after this controller deactivates the fans. Fortunately, the truth reveals itself to a tachometer with a laser sensor. Using this hardware, we're able to map out the fan curve, which appears tuned to keep AMD's GPU from exceeding a 75°C temperature target.

After a period of heavy cooling during warm-up, the fans slow down and stabilize. However, Gigabyte tried a little too hard to keep noise down with its small cooler, resulting in fan speeds that have to speed up and slow back down under load. It certainly would have been possible to specify a faster, more constant speed setting. This would have evened out the curve and helped on-board components run a little cooler, too.

Little changes during our stress test. In short, the card's two fans perform quite well, but would definitely benefit from slightly higher rotational speeds.

As a consequence, the Radeon RX Vega 56 Gaming OC 8G has no margin left for lower fan speeds. This setup is too heavily optimized for noise, as the following table shows:

Fan RPM & Noise Measurements
Fan RPM, Open Test Bench, Maximum1731 RPM
Fan RPM, Open Test Bench, Average1134 RPM
Fan RPM, Closed Case, Maximum1730 RPM
Fan RPM, Closed Case, Average1264 RPM
Noise (Air) Range
33.4 (Minimum) to 40.8 dB(A)
Noise (Air) Average36.1 dB(A) (Warmed up)
Noise (Air) Idle0 dB(A)
Noise characteristics / Subjective ImpressionsLow-frequency bearing noise
Some motor noises below 1 Hz
Moderate air and turbulence noises
Hardly any voltage converter noise

This snapshot illustrates the entire frequency range of our laboratory measurements, adding some data to our subjective observations. The alternating fan speeds we mentioned previously are clearly visible.

An average of 36.1 dB (A) is more than acceptable for such a powerful card. In fact, the outcome is almost too good. We would have tolerated a bit more noise to get a stable fan curve. Fortunately, you could solve this on your own with a bit of manual adjustment.

MORE: Best Graphics Cards

MORE: Desktop GPU Performance Hierarchy Table

MORE: All Graphics Content

This thread is closed for comments
24 comments
    Your comment
  • marcelo_vidal
    With the pricey from those gpus :) I will get an 2400g and play 720P. maybe with a little tweaking I can boost to 1920x1080
  • Sakkura
    This thing about board partners only getting a few thousand Vega 10 GPUs goes back many months now. Has AMD just not been making any more? What the heck is going on?

    Seems like Gigabyte did a really nice job making an affordable yet effective cooling solution for Vega 56, it's really a shame it goes to waste because there just aren't any chips available.
  • P1nky
    I think you forgot to add the overclocking and undervolting content.
  • FormatC
    Read the review. And you will find the answer ;)
  • CaptainTom
    To those complaining about the low supply (and resulting high prices) of AIB cards:

    It's because the reference cards are still selling very well (at least for their supply). If vendors can sell the $500 Vega 64 for $600 and sell out, why would they bother wasting time on any other model?
  • g-unit1111
    741589 said:
    To those complaining about the low supply (and resulting high prices) of AIB cards: It's because the reference cards are still selling very well (at least for their supply). If vendors can sell the $500 Vega 64 for $600 and sell out, why would they bother wasting time on any other model?


    That's because miners are the ones buying the cards as fast as they come in stock. It's us gamers and enthusiasts that are waiting for the high performance models. Bad thing is, we don't matter to the bottom line. All they see and want is our precious money, and they don't care what model they sell to us.
  • aelazadne
    Because, the Vendor's making money doesn't equal AMD making money. AMD is losing market share in the GPU scene. With Vega unable to keep up with demand AMD is losing customers who would have bought Radeon's but instead go with Nvidia due to availability. The lack of Availability stemming from August and the fact that even now in early 2018 the Vegas are over priced and hard to find ruins customer confidence. In fact, this situation is so bad that the only people benefitting are the people gouging both Nvidia cards and Radeon cars because at this point there is NO COMPETITION.

    Also, just because you are gouging doesn't mean you are making money. AMD has to make money and they need to sell these things in a certain volume. In their contracts with Vendors, they will require their vendors to sell a certain amount of vegas in order to order more. Due to scarcity the only companies making money are Retailers. AMD is going to have to address this issue otherwise their investors will begin to come after them for bungling so bad that their market share dropped so bag. Literally, the intel screw up plus Ryzen being good has been a godsend for AMD, they do not need a declining GPU market share sparking a debate with investors over whether AMD should get out and play the Intel game.
  • bit_user
    482859 said:
    ...

    I really appreciate the thorough review.

    The super-imposed heatpipes vs. GPU picture was a very nice touch. For any of you who missed it, check out page 6 (Cooling & Noise) about 1/3 or 1/2 of the way down.
  • bit_user
    741589 said:
    It's because the reference cards are still selling very well (at least for their supply). If vendors can sell the $500 Vega 64 for $600 and sell out, why would they bother wasting time on any other model?

    I think you're too cynical. It's an ASIC supply problem. The AIB partners would probably spend the time if they could get enough GPUs to sell custom boards in enough volume to offset the overhead of doing the extra design work.

    The only real way out of this is for AMD to design a more cost-effective chip with the graphics units removed. That will divert miners' interest away from their graphics products.
  • bit_user
    Almost as surprising to me as how much more oomph they got out of Vega 56 is how well the stock Vega 64 is holding up against stock GTX 1080. Is it just me, or did AMD really gain some ground since launch?
  • aelazadne
    You know I was saying the same thing? I really want a Vega 56, especially because their is a nice freesync monitor that I want. Also, I am surprised that neither AMD nor Nvidia have made cryptocurrencies specific cards. They would make a ton of money and they would free up their gaming lines. They could make mining specific cards and make those people pay through the nose.
  • Blytz
    Need to try overclocking the ram and flashing a 64 bios onto it for more headroom in the overclock and the ram overclock.

    Makes a BIG difference.
  • FormatC
    OC'ing a Vega (on the right way) isn't so easy. It's simple to flash a Vega64 LC BIOS, but with exception of an explosion of your power consumption you get only a few percent more gaming performance. Undervolting, to break the power limitation, is the better way. You can keep the performance and save a lot of ressources. But exactly this thing is very difficult to realize. WattMan alone isn't able to realize such projects. And with a new driver this game starts again and again.

    I tried it so often with now more than 10 cards - each card is so different and I was not able to found a global setting for all cards to write a recommendation. It makes no sense for me to write about an undervolting sensation for one card, if the next card was a potatoe chip. I have here both extreme, good and bad chips. And if you see this big bandwidth, you can understand me, why I dislike such OC or UV parts in a review.

    Spoiler:
    I'm just testing the Gigabyte RX Vega64 Gaming OC and it is interesting to see, that Gigabyte is very well knowing the limitations of their cooling solution. The power draw is very similar and I really like this bigger card. But it is nothing for a big OC. If you need significant more performance (f.e. in Ultra-HD), Vega is the wrong horse. But for WQHD and Freesync it is a very good offer.

    I've played Wolfenstein 2 with a water cooled Vega64 (full-cover water block) in Ultra-HD and was satisified - but to finish now Witcher 3 for a 100% savegame (story), it was totally unusable. I also tried to OC this Vega64 and got stable 1,7 GHz - but this was also not enough for a good immersion. The advantage of an open loop cooling system and quick connectors is, that I can change the cards within five minutes. The 1080 Ti was is this case the significant better solution. This is, what I meant above with the horse. It differs from case to case :)
  • bit_user
    2634802 said:
    I am surprised that neither AMD nor Nvidia have made cryptocurrencies specific cards. They would make a ton of money and they would free up their gaming lines.

    Remember, the latest round of GPU-based cryptomining only blew up in about June 2017. It takes much longer than that to design a chip like these and get it to market. On top of that, they were probably uncertain about how long the crypto craze would last. So, it would still be a while before such chips might hit the market.

    2634802 said:
    They could make mining specific cards and make those people pay through the nose.

    AMD's official line is that there's not much they can do to cripple gaming cards for crypto purposes that wouldn't also compromise gaming performance. That means the only way to divert cryptocurrency miners from their gaming products is to make a chip with better price/performance at crypto. They could accomplish that simply by stripping off the graphics-specific units from their GPUs, which would shrink the die and make it cheaper (i.e. more cost-effective). As long as they could make enough of them (a big "if", I know), they would continue to be more attractive to miners than their graphics chips.
  • Ravnholt
    So from a noise perspective, this is a better card than the Sapphire Nitro?
  • Ravnholt
    So noise wise, this is a better card than the Sapphire Nitro?
  • FormatC
    I was suprised - but yes. First time, that I saw such a quiet Gigabyte card out of the box.
    But don't forget - each MHz more needs a significant higher power consumption.
    Sapphire made an excellent cooler - to cool down al this waste of energy. Hearable waste, only to be 1-2 FPS faster... :D
  • Blytz
    Guess I struck silicon lottery to a degree then (especially in light of the boost from the flash)
    Granted my 56 overclock gpu limited to around 1600 (1610-1630 was seen in 3d mark)
    But I got a perfectly stable ram oc from 800 to 1100 day in day out, under the 56 bios it was limited to 950 (960 on a good day)

    edit - I just realised I was too retarded to quote your post formatc (my bad)
  • FormatC
    The problem of all this custom 56 is/was:
    AMD sent now to the AIB a new source BIOS kernel for the custom designs. It gave me 50 MHz more clock speed for free with the same power consumption and also an OC to 1050 MHz mem clock without flashing to Vega64. Gigabyte also improved today their BIOS to get a tick more fan speed to keep the fan speed stable now. They followed 1:1 my suggestions and it is in average not noisier than before.
  • bit_user
    482859 said:
    AMD sent now to the AIB a new source BIOS kernel for the custom designs. It gave me 50 MHz more clock speed for free with the same power consumption and also an OC to 1050 MHz mem clock without flashing to Vega64.

    So, are you going to update the benchmarks?

    482859 said:
    Gigabyte also improved today their BIOS to get a tick more fan speed to keep the fan speed stable now. They followed 1:1 my suggestions and it is in average not noisier than before.

    Nice!
  • FormatC
    No re-bench, because I got it earlier as Beta BIOS for the vega56 and benched it with the new power table inside. Compare it with the Sapphire Nitro review, where we had performance issues in a few games. All this is solved with this new kernel. After heating on, I measure now a little bit more than 36 dB(A). Also not noisy.

    Just sitting on the RX Vega64 Gaming OC. I got the new BIOS after benchmarking and I had to re-bench all. The Vega56 BIOS was earlier available and so I changed also the order for my reviews.
  • bit_user
    482859 said:
    No re-bench, because I got it earlier as Beta BIOS for the vega56 and benched it with the new power table inside. Compare it with the Sapphire Nitro review, where we had performance issues in a few games. All this is solved with this new kernel.

    Okay, so these benchmarks reflect the new BIOS?

    Sorry if you stated this, but I skimmed parts of your article due to having so much CES coverage to read.
  • FormatC
    This Vega56 is retail (self-imported from Asia) and came with the newer BIOS. The Vega64 was updated yesterday and I started it again :(
    Today in the morning came the next vega64 BIOS with 100 rpm more fan speed. I feel me like an AMD alpha and beta tester :D
  • P1nky
    Thanks for keeping us updated. :)