SAPPHIRE PULSE Radeon RX 5700 XT Review: Cool and Quiet
SAPPHIRE PULSE Radeon RX 5700 XT
Don’t fancy a blower cooler? This quiet card may be the answer.
We finally have our hands on a custom AMD Radeon RX 5700 Series graphics card, as we have been testing this new SAPPHIRE PULSE RX 5700 XT over the last few days to see what a custom cooled, factory overclocked option brings to the table. Launching today in both RX 5700 and RX 5700 XT versions, SAPPHIRE lists these features for these new PULSE cards:
- New TriXX Boost Feature delivers up to 15% performance increase in popular games
- Improved PCB, power and cooler designs ensures gamers have superior clock speeds
- PULSE cards deliver cool, quiet and affordable performance
- Components efficiently cooled by finely tuned Dual-X technology
- New SAPPHIRE COOLTECH, DNA and PLUS technologies feature sets
- New extreme AMD RDNA gaming architecture powers PULSE RX 5700 cards to new heights!
- Model: SAPPHIRE PULSE Radeon RX 5700 XT
- GPU: PULSE RX 5700 XT Graphics, 2nd Gen 7nm GPU, RDNA Architecture
- Stream Processors: 2560
- Boost Clock: Up to 1925 MHz
- Game Clock: 1815 MHz
- Base Clock: 1670 MHz
- Memory Size/Bus: 8GB/256 bit GDDR6
- Memory Clock: 14 Gbps Effective
- Displays: Maximum 4 Displays
- Resolution: HDMI: 4096×2160 @60Hz, DisplayPort 1.4: 5120×2880 @60Hz
- Interface: PCI-Express 4.0
- Output: 1x HDMI, 3x DisplayPort
- SAPPHIRE PULSE Radeon RX 5700 XT: $409.99
- SAPPHIRE PULSE Radeon RX 5700: $359.99
“The PULSE Radeon RX 5700 Series is all about providing PC users with an immersive gaming experience and great value for money. Starting with the SAPPHIRE PULSE Radeon RX 5700 XT, this factory-overclocked model is dedicated to customers who want a future proof card that delivers incredible visual fidelity, lightning-fast performance and an advanced technology feature set to power the latest AAA and eSports titles.”
The SAPPHIRE PULSE Radeon RX 5700 XT
For their PULSE design SAPPHIRE has implemented a dual-fan cooler with ball-bearing 95 mm fans along with a heatsink design that targeting gaming load temps under 75 C along with near-silent operation. To this end SAPPHIRE has implemented their own Intelligent Fan Control to balance performance and noise, while keeping component temps as low as possible.
“The industry proven SAPPHIRE Dual-X cooling solution has been finely tuned. Powered by two newly designed silent dual ball-bearing fans and a state-of-the-art radiator design has enabled SAPPHIRE to realise improved performance with cooler and quieter operation. Designed to maintain temperatures under 75°C while gaming and run up to 10 dBA quieter, the streamlined form of SAPPHIRE’s 95mm blades mean greater airflow and superior heat sink coverage at lower noise compared to the RadeonTM RX 5700 reference designs.”
In addition to the cooling capabilities for GPU, memory, and VRMs, SAPPHIRE is providing a custom PCB with these new PULSE cards, offering long-life capacitors, fuse protection, and a 12-layer PCB design. There is also an aluminum backplate that uses a thermal pad to help dissipate heat in addition to serving to keep the card more rigid.
We will examine stock performance first, running most of the same game benchmarks as we did in the recent NVIDIA RTX 2080 SUPER review – and not the the initial Radeon RX 5700 Series review. This is due to the fact that beginning with that RTX 2080 SUPER review we have a brand new test platform, making comparisons to previous results impossible.
|PC Perspective GPU Test Platform|
|Processor||Intel Core i7-9700K|
|Motherboard||GIGABYTE Z390 AORUS PRO|
|Memory||Crucial Ballistix Sport LT DDR4-3200 32GB (16GBx2)|
|Storage||CORSAIR Neutron Series XTi 480GB SSD|
|Power Supply||CORSAIR RM1000x 1000W|
|Operating System||Windows 10 64-bit (1903)|
|Drivers||Radeon Software Adrenalin 19.7.2, 19.7.5 (SAPPHIRE PULSE)
GeForce Game Ready Driver 431.36, 431.56 (RTX 2080 SUPER)
Far Cry 5
First up, in alphabetical order, is the venerable Far Cry 5. This is a DX11 title, run – as with all of the benchmarks to follow – using the “ultra” preset.
This first test sets the tone for the results to follow, wherein the AMD reference 5700 XT is slightly ahead of the factory-overclocked SAPPHIRE PULSE card. Is it possible the SAPPHIRE card is under-performing? Or is it just possible that AMD’s press sample is an above-average performer for reasons that might be obvious, but are simply speculation until you compare clock speeds directly? More on this later.
Final Fantasy XIV Shadowbringers
Another DX11 test, this is actually a new benchmark for the latest expansion of FFXIV. And yes, every card has a lot of frame time variance in this game.
Run in DX12 mode, this becomes the first of two DX12 API benchmarks in the group.
Things are getting closer between the two Radeon RX 5700 XT’s here, with a mere 0.5 FPS separating the two in Metro Exodus.
Shadow of the Tomb Raider
The second, and last, DX12 benchmark of this group is famously unfriendly to AMD cards, but will serve to illustrate another very small disparity between the AMD reference card and this SAPPHIRE PULSE.
World of Tanks enCore
Even at its “ultra” preset the enCore benchmark is not the most demanding test out there. A DX11 title, this will be one of a pair of benchmark results to close this section of the review that feature FPS averages well beyond 100.
World War Z
With the game set to use the Vulkan API rather than DX11 AMD cards fare exceptionally well in WWZ, as you may remember from the RTX 2080 Super review.
With these rather curious benchmark results out of the way let’s examine the difference between these two 5700 XTs a little more closely.
A Tale of Two Graphics Cards
Without a doubt, something strange is going on with the results above. I refused to accept them, and ended up spot-checking performance between the two cards with two driver versions and in different games. The gap was still there, and the AMD reference card was still on top by a very small margin. Yes, the stock version that we received at launch is faster than an AIB partner’s factory overclocked card.
After speaking with SAPPHIRE about my findings and verifying that the PULSE was operating at expected game clock speeds, I decided to do some quick frequency logging. Granted this was hours before launch so a couple of screenshots will have to do.
SAPPHIRE PULSE on the left, AMD reference card on the right
A glance at the base and boost speeds in GPU-Z reveals two curious things: first, what on earth is going on with base clocks as reported here? AMD’s specs for a stock RTX 5700 XT call for a 1605 MHz Base Clock, while SAPPHIRE lists the PULSE RX 5700 XT at 1670 MHz. A GPU-Z anomaly, or perhaps I am seeing the difference between idle clocks and the official Base Clock.
Forgetting the idle clock weirdness for a moment, just look over at the Boost Clocks. 2039 MHz for the SAPPHIRE card, with 2069 MHz for the AMD reference. Both seem way too high, but in any case the reference card has a higher clock than this SAPPHIRE PULSE, which fully explains the slight advantage the press sample enjoyed in the benchmarks above. Here’s a quick example made by running the Metro Exodus standalone benchmark for 10 iterations while logging the GPU frequencies using GPU-Z.First, the SAPPHIRE PULSE:
And now a look at the same series of benchmark runs using the AMD reference card:
There you have it, if you can make out what this quick visualization of a GPU-Z log is attempting to convey. Simply put, the reference card spends more time at higher GPU frequencies, and tops out at higher frequencies, than the SAPPHIRE PULSE.
Expected behavior for the PULSE is to hit about an 1815 MHz Game Clock under sustained load, with a reference card targeting 1755 MHz, according to AMD specs. This is clearly not the case with our sample, which may just mean we hit the silicon lottery to some extent with our reference card.
Power, Temps, and Noise
Having beaten the subject of reference vs. SAPPHIRE PULSE clocks speeds to death (and then beat it some more for good measure), let’s get back on track and look at power consumption.
As expected for a factory-overclocked card the PULSE does draw a bit more power at both idle and load than the reference card. Performance, as we have seen and then discussed at length, disagrees with this chart, but it is what it is. Even if I’m not very happy about it.
Next up we will talk temps, though there isn’t much to say here. With the room at approximately the same ambient temp as our initial Radeon RX 5700 XT review (25C – 26C for both cards) we saw load temps with AMD’s blower cooler hitting 76C with a 94C hot spot in our admittedly not-so-punishing test, with the SAPPHIRE PULSE a couple of degrees below this at 74C/92C hot spot.
Noise levels, on the other hand, were drastically lower with the PULSE card compared to the reference design, with load noise of just 38.3 dBA from the SAPPHIRE card compared to 48.6 dBA from the reference card. SAPPHIRE’s conservative default fan profile might keeping the PULSE from hitting more impressive temps than we saw, but it’s impressive nonetheless that temps are still lower than reference with a card that barely audible under load over average room noise.
TriXX Boost: An Alternative to Overclocking
SAPPHIRE is working on a new version of their TriXX software that we got to play with while testing this new card, and this includes a very interesting new feature called TriXX Boost. This Boost option is designed to provide higher FPS in games by using a lower resolution for rendering and then upscaling the output, and can also leverage AMD’s Radeon Image Sharpening for compatible games (DX9, DX12, and Vulkan).
“TriXX Boost enables gamers to run games at a higher FPS by reducing the rendering resolution and up scaling the final output image by integrating Radeon Image Sharpening. This powerful tool besides being easy to use, sees up to a 15% performance improvement in 3D Mark and 10% performance improvement in popular games the resolution is adjusted from 2560×1440 to 2304×1296; or up to 22% performance improvement in 3D Mark up to a 15% improvement in games when adjusting the resolution from 3840×2160 to 3456×1944.”
If this sounds familiar it is of course reminiscent of NVIDIA’s DLSS, but rather than a fixed lower internal resolution and AI-powered upscaling, this Boost feature simply allows you to choose any percentage from 70% – 99% of the output resolution to actually render the game with, and in combination with Radeon Image Sharpening I must say a modest 90% setting as recommended was very difficult to discern from the natively-rendered version.
Sticking with the 1440/Ultra strategy from the previous benchmarks, and selecting this 90% rendering and thus creating a new 2304×1296 resolution using the pre-release TriXX software I quickly repeated a couple of the game benchmarks from before to see what sort of performance improvement I could get while successfully fooling myself that I was actually seeing 2560×1440.
The best example in my brief experiments in the early hours of this morning was Shadow of the Tomb Raider, where I saw the average go from just under 67.7 FPS to 78 FPS. And a 10% reduction in rendered res is so low that even without Radeon Image Sharpening (which was enabled with the game in DX12 mode) I think it would take screenshots to tell the difference.
The SAPPHIRE PULSE Radeon RX 5700 XT provided performance within 1-2 FPS of our (it turns out) very high-performing AMD reference card, but did so with significantly lower noise levels – almost exactly 10 dBA lower than the reference design under load. Those who have been waiting for a non-blower design for lower noise levels can rejoice at this SAPPHIRE card, which should actually be available as you read this.
While the disparity in performance between the AMD-provided reference card with this first partner card we have been able to test raises questions on both ends, the difference is slight. Even if the Game Clocks we observed from the reference version are totally in keeping with retail cards the loss of 1-2 FPS with the SAPPHIRE PULSE should be more than mitigated by the huge 10 dBA drop in noise.
Silent at idle thanks to a zero RPM fan and ramping up only to 38.3 dBA under regular load (with 48.6 dBA recorded using the reference design), this SAPPHIRE PULSE RX 5700 XT is just what the doctor ordered if you don’t like loud blower coolers. Temps were also a little better than our reference card, and while in our brief testing this meant it was coming in just 2 degrees Celsius below the AMD card under load (longer stress runs might paint a more impressive picture) you have to consider the lack of noise output to achieve this improvement.
Adding to the mix an MSRP of only $10 above a reference card and I think SAPPHIRE has a winner on their hands with the PULSE. Add in the rather ingenious TriXX Boost option in their upcoming software release and SAPPHIRE further (and positively) differentiates themselves.
Time (and more reviews) will tell if AMD’s press cards will end up out-performing more of the partner cards destined to flood the market in short order, but for now I must judge this card against a single press sample, which hardly seems fair. The jury is still out on those benchmark results.
After seeing other reviews from around the web and noticing results in keeping with what I had expected going into this review (slightly higher performance from the SAPPHIRE card than AMD’s reference), I have determined that we simply received a reference sample with above-average performance. The SAPPHIRE PULSE should offer slightly higher performance compared to AMD’s reference RX 5700 XT, and it will unquestionably do so with significantly lower noise and lower temps.
As to the benchmark charts in this review, our reference 5700 XT produces clock speeds when gaming that are consistently higher than even the overclocked PULSE card, explaining the 1 – 2 FPS advantage seen above. Going forward it will be interesting to see if any BIOS difference exists between our press sample and a retail 5700 XT reference card.
Looks great, for some reason I REALLY want to wait for these custom cooler cards of the RX 5700 to drop to $300 before I purchase one myself. I have never purchased a video card before and spending any more than $300 is rather though for me.