Overclocking Performance, Benchmarks, Temperatures and Power Consumption

You always want to get more than your payed for, so that is what this section is for, squeezing that little extra performance.

Overclocking is not easy and what works for me, may not for you and visa versa but it’s always good to have a point of reference.  These overclocks are by far not the best or furthest you can push the card but it’s what I could do in the time I had with this card.

To start with, let’s compare the readings in GPU-Z of the stock settings and the overclock settings.

Above, I took the liberty of outlining in red the changes the overclock has changed.  With the overclock, we increased the GPU clock by 61Mhz and the Memory by 140Mhz.  Doing this we increased the bandwidth by 57.9 GigaBytes, Texture Fillrate by 15.6 Gigatexels and the Pixel Fillrate by 3.9 GigaPixels.

Aside from raising the GPU and Memory speeds, we also raised the fan speed and power limit if the card.  Normally on a Sapphire card, I would have used TRIXX but it seems like they do not have a version of TRIXX for the VEGA64, so I had to look towards MSI for their Afterburner.  Here are the settings I used.

It took me a while to get to this point with the overclock, higher that that I would have had to unlock the core voltage and spend a few more hours or days on this.  Again, this is to give you an idea of how the overclocking is done and how it performs with a little push, surely you can push it more than I did.

So let’ start this off with 3DMark.

With the overclock, we can see the complete score was 18,982 while the Graphics score was 26,613, Graphics test 1 was 121.01 FPS, and test 2 was 95.92 FPS.  The overall Physics score was 14,061 and the test was 44.64 FPS.  Finally, the combined score was 8,666 with a combined test score of 40.31.

So let’s compare the stock and the overclock settings to see how much the settings affected the outcome.

The 2 most notable bars here between the 2 are the Graphics Score and the 3DMark overall score.  In the overclock bench, we can see the overclocked coming at 7.74% and 1,834 points higher than the stock speeds.  On the Graphics score, we can see the overclock came in 5.04% and 933 points higher than the stock speeds.  All of the results here came up ahead though the rest maybe not as notable.

This bump in performance also brought up the results in 3DMarks online performance index from 93% better than all the rest to 95% better, that’s pretty nice.

With an overclock though, higher wattage and temperatures are expected and we were not let down.  The overclocked temperature was 5° over stock and while the max wattage consumed 12.94% more power than stock, the average power drain was actually lower than the stock card.  A bit suprising the average was slightly lower but a welcome surprise that the temperature was only 5° higher, Sapphire did a great job introducing 3 fans instead of 2.

One might think that adding the 3rd fan would me more noise, but AMD introducing Radeon Chill and Sapphire design help to keep the card cool and quiet and when a game is not being played, the fans actually stop, making this card 100% quiet.

So, let’s check out Time Spy, to see what sort of improvement we see there.

Just like in the previous test sadly, it is still 80% better than all the other results.  I mean, it still is good where it is, but we all like it to be as best as it can be.

Ok, in these results we can see the overall Time Spy result was 7,299, Graphics score was 7,756, Graphics test 1 was 53.51 FPS and Graphics test 2 was 42.41.  Along with the graphics results, the CPU score was 5,475 and the CPU test was 18.40 FPS.

Now let’s see how the two compares.

Things heat up here, literally.  On the overclock, on the Graphics score, the overclock performed 7.67% better than the stock card, a decent increase.  On the Graphic Test 1 and 2, we also see a 7.92% and 7.34% increase in performance respectfully.  The overall Time Spy score increased by 6.08%.  Now, the CPU tests tell a different story that might confuse you.  Since we have improved the performance on the GPU, the necessity to depend on the CPU decreases ever so slightly.

With this increase in performance, we also have an increase in wattage used and heat generated.  The heat generated was partially my fault though, I set the fan speed to low so that it would not generate too much noise, not that it matters to me because I have my headphones on.  At stock, the hottest the GPU reached was 69°C and the overclocked reach 82°C though both only for a second or 2.

During these times of greater heat, the power increases as well because it is using more of the GPU and resources all together.  On average during these tests the stock card only hit 383Watts while the overclock hit 460Watts, an increase of 18.27%.  Even though we are not going by maximums here, I do want to mention that at one point the system did hit 533Watts.

Each of the game benchmarks will be broken down into resolutions.

Let’s move on to Metro.

1920 x 1080

Ouch, it looks like the overclock actually performed lower, though I can see that the clocks were all correct while it was benchmarking and the temperatures did not cause any throttling.  The stock actually performed 0.08% better than the overclock, though the overclock consumed on average 10.69% more wattage.  The max wattage on the overclock peaked at 519 Watts, 20.15% higher than the peak wattage on the stock.  The saving grace here is that the overclock was 1°C lower.  Let’s see what 2560 x 1440 does.

2560 x 1440

Things are making a little more sense now.  Here we can see on average the overclock performed 6.62% better and on average consumed 17.27% more power than the stock card.  The temperature did increase by 6.37% though, but it could have easily been adjusted the fans a little higher.  Not sure what happened on the 1920 x 1080 results, but things look a little better on here, making 2K much more playable.

Now let’s check out 3840 x 2160.

3840 x 2160

A 4K, there is a slight bump in performance by 17.70%, an increase on the overclock.  There was a huge increase on performance on the max frame rate by 64.77% but you can’t judge perfomance by a spike.  On this, the average wattage consumed on the overclock was 11.63% higher than the stock consumption and 13.16% higher in temperature.

During this overclock, 2560 x 1440 became much more playable but oddly enough performance dropped at 1920 x 1080, though the performance was still very playable.  There was no saving 3840 x 2160 though.  Hopefully Laura can help save this.

1920 x 1080

On the overall score, we find that the overclock performed 1.34% higher than the stock speeds.   Across the board we can see there was an improvement, though a minuscule one.  With that increase, even though small the average power consumption increased by 10.99% and temperature increased as well by 8.92%.  Seeing as Metro’s performance decreased at this resolution, it is a welcomed surprise to see that performance increased, but now what will the next test reveal?

2560 x 1440

Alright, a little better here, we can see that the overclock performed 3.12% better than the stock card.  With that, the temperature did increase by 10.81% coming in a 78°C and Laura demands some power with the average consumption coming in at 18.27% above stock.  While both were very playable, extra performance is always welcome.  Let’s see what 384- x 2160 provides.

3840 x 2160

At 4K, things are a bit better, but not where you would want them.  Overall performance was increased by 4.69% and with that the average wattage was also raised 15.27% watts.  The temperature only increased by 8.11%, a tiny 6° and while the power draw was increased, the fans took care of their jobs very well.

Performance overall at 4K was increased but still left about 11 frames per second before the magic number would be reach.  Do you think this is all due to drivers, could drivers get you that extra fps to make 4K 100% playable?

While I let you ponder on this, let’s check out Tom Clancy’s Ghost Recon Wildlands.

1920 x 1080

Wildlands, while already playable at this resolution, the overclock barely made an impact.  We can see the FPS count rose .35%, and thankfully the wattage only increased by 3.71% and a temperature increase of only 1°.  Oddly enough at this low resolution, the CPU usage actually decreased where at other games at this resolution would normally increase and GPU usage increased.  A little backward, nothing really to be concerned with but worth mentioning.

2560 x 1440

On stock we were already at 78.31, well above 60FPS but at the overclock we punched it up 1.78%, not amazing but every single fps count.  The wattage went up by 7.61% but the temperature dropped 2°s which is always welcome.  Can we hit the same at 4K?

3840 x 2160

So close here, we jumped from 45.79 frames per second to 47.72 frames per second here, a bigger jump than before coming in a 4.13% increase in performance, but still slightly short of 60 frames per second.  The boost however drew on average 10.94% more power but the temperature dropped 3°C, so some good and some bad.  Let’s see what kind of performance we can see in The Division.

Well, this resolution does not really need a boost, but what it does take I will gladly take.  On the overclock we gained 6.97% over the stocks 113.6 frames per second.  This increase in performance did bring about a 6.37% increase in thermals and a 12.90% increase in power consumption.  On to 2560 x 1440.

2560 x 1440

The score here was not drastic, but the power draw was.  On average the overclock consumed on power was 21.81% over the stock speeds, a pretty bit hit though the thermals only went up a slight 7.69% over stock speeds as well.  The performance increase was a slight 2.46%, nothing big here, but it was already playable as it was.  So let’s check out the last of the resolutions and the most aggressive, 3840 x 2160.

3840 x 2160

Sadly, it looks like 4K has suffered once again, the overclock performing only 4.95% better than stock.  Like we have already seen before, the power draw increased in the overclock by 24.70%, a pretty significant increase and the temperature has increased as well but only by 8.81%.  Both of course could have been lowered depending on the overclock settings.

So now that we have completed the overclock versus stock comparison, the overclock did not make a huge difference in games, though it did make a pretty significant improvement on a synthetic benchmark… which is more important to you?

So then next treat, something that we don’t get to do often is that there will be a bit of a comparison between cards, not just same chipsets though, this will be an AMD versus NVIDIA type of thing.  I will preface it by stating, a like comparison between card would have been this card versus a GTX1080 but I don’t have one, so we will go up one level and compare this card to a GTX1080 TI, namely the EVGA Geforce GTX1080 TI FTW3 Gaming Edition.

Let’s check out the differences between the two, though it is a bit unfair since they are not the same category of cards.  The direct competition for the Vega64 is the GTX1080, not the GTX1080 TI, but still a nice comparison.  We will be comparing stock versus overclocked for both cards.

Iggy Castillo

Iggy Castillo

Senior Editor an Reviewer at Dragonblogger.com
I have spent many years in the PC boutique name space as Product Development Engineer for Alienware and later Dell through Alienware's acquisition and finally Velocity Micro. During these years I spent my time developing new configurations, products and technologies with companies such as AMD, Asus, Intel, Microsoft, NVIDIA and more. The Arts, Gaming, New & Old technologies drive my interests and passion. Now as my day job, I am an IT Manager but doing reviews on my time and my dime.
Iggy Castillo

@zangza

I love the arts and technology. IT Manager by day at Jewett Automation and Reviewer by night at https://t.co/eYqx0uJofz and my own at @ThisBytesForYou
RT @NellyFRyz: Save $122 on AMD Ryzen 7 2700X plus get a free game choice. WQL. https://t.co/hstyE3okgq https://t.co/8sTBDkL4g9 - 2 weeks ago
Iggy Castillo