The Corsair Team managed to make the Dominator Platinum SE heat spreader even more spectacular than before, trying to replicate the aesthetics of high performance car engines was their mission and they have succeeded big time. The black brushed aluminum heat spreader in combination with the heat treated top bar and them orange cooling fins make you immediately think of hot exhaust manifolds of an overcharged powerhouse.

 

 

The package itself is quite discrete and reminds me of the DDR3 Dominator Platinum versions a long time ago. We received the Quad Channel kit as it would be more practical for our testing purposes on high end desktop platforms, as lower density modules are more versatile to conduct any tweaking or overclocking tests on than the larger variants.

 

 

Both Dominator Platinum Special Edition Torque memory kits are 32GB versions; the CMD32GX4M23200C14T model is the Dual channel consisting of two 16GB memory modules running at 3200MHz with CL14-16-16-36 timings. The CMD32GX4M43200C14T is the Quad Channel version sporting the same timings and module voltage of 1.35 volts. The 10-layer PCB and the handpicked Samsung B-Die ICs warrant easy overclockability to at least DDR4-3600 speeds.

 

 

Both kits support XMP version 2.0 so your motherboard will use the appropriate timings to warrant compatibility and performance. While this all sounds nice the price is the only thing that might a bit overwhelming. Snooping around at the Corsair website it seems this brand has some high price tags all around for 32GB kits, no matter the heat spreader or version type. Either Corsair is a strong believer users will buy their premium products even with a  100 euro higher price tag or the webadmin made a few too many copy paste errors.

 

Each module has a removable foil to protect the brushed heatspreader from any scratches.

It took us a while to get access to an X299 platform, hence why the long delay to start properly testing the Corsair SE Torque quad-channel kit.

Our Madshrimps Intel bench platform comprises the following parts:

• Intel i9-7900X 10 core processor
• EK Waterblocks L360 kit
• Gigabyte X299 Aorus Gaming 9 motherboard (F4 bios version)
• MSI R290X Lightning videocard
• Western Digital 2TB Caviar Green Edition
• Seasonic 1200W Platinum power supply
• Win 8.1 64bit professional Edition
• Catalyst 17.7.1 Crimson Edition graphics driver

 

 

 

The Madshrimps AMD Ryzen testbed is made up of:

• AMD Ryzen 1800X processor
• ASUS Crosshair VI Hero motherboard (1402 bios version)
• MSI R290X Lightning videocard
• Western Digital 2TB Caviar Green Edition
• Seasonic 1200W Platinum power supply
• Win 8.1 64bit professional Edition
• Catalyst 17.7.1 Crimson Edition graphics driver

 

 

The memory kits we used to compare the performance of the Dominator Platinum Special Edition Torque to:

• 16GB G.Skill Ripjaws 4 3000C15 (link)
• 32GB G.Skill TridentZ RGB 3600C16 (link)
• 32GB G.Skill TridentZ 3600C17 (link)
• 32GB G.Skill TridentZ 4133C19 downclocked to 4000MHz (link)

• Corsair 32GB Vengeance 2400C14 LPX  (link)
• Corsair 32GB Vengeance 2666C16 LPX  (link)
• Corsair 32GB Dominator Platinum SE Torque  (link)

 

 

 

We always go through a multitude of benchmarks to test the performance of the installed memory:

 

• AIDA64 Engineer Memory Bandwidth
• Super PI 32M 1.5XS Edition 2D benchmark
• Maxon Cinebench 64bit Release 15 2D benchmark
• HWBOT RealBench 2D benchmark
• HWBOT X265 render 2D benchmark
• Futuremark FireStrike Ultra 3D benchmark
• Tomb Raider build in benchmark
• Bioshock Infinite build in benchmark

First up is the out of the box performance of the Corsair Dominator Platinum Special Edition Torque series versus the rest of the available Corsair and G.Skill kits. Besides the 16GB G.Skill 3000C15 all the other contestants are 32GB kits either made up of two dual channel kits or full blown quad channel versions.

 

AIDA64 Engineer has a hardware detection engine unrivaled in its class. It provides detailed information about installed software and offers diagnostic functions and support for overclocking. As it is monitoring sensors in real time, it can gather accurate voltage, temperature and fan speed readings, while its diagnostic functions help detect and prevent hardware issues. It also offers a couple of benchmarks for measuring the performance of individual hardware components or the whole system. It is compatible with all 32-bit and 64-bit Windows editions, including Windows 10 and Windows Server 2016.

 

 

The AIDA64 Bandwidth test is a synthetic benchmark that adores memory clock speed and lower timings. Compared to lower end offerings (read sub 3000 speeds) the Torque series provide a good balance between memory speed and timings, though it can't keep up with the faster running G.Skill siblings, though how does this bandwidth reflect into more real life scenarios like rendering and gaming?

 

MAXON Cinebench R15

Anyone who needs to evaluate hardware performance should add MAXON CINEBENCH to the test tool arsenal. System administrators can use CINEBENCH to help make purchase decisions, journalists can use the results in reviewing hardware, hardware manufacturers may utilize the feedback in optimizing their latest products. Any computer owner can evaluate his or her individual system. Unlike abstract benchmarks, which only test specific functions of CPUs or GPUs, CINEBENCH offers a real-world benchmark that incorporates a user's common tasks within Cinema 4D to measure a system's performance. For those who have to do a serious amount of testing CINEBENCH also provides a command line option, allowing users to run automated test procedures.

 

 

 

Again decent performance of the Torque kit, the memory speed versus timings are well balanced as it is almost besting the higher specced G.Skill 3600C17 TridentZ kit in this test. We have to admit though that the differences between all the tested kits is nowhere near mindboggling. In fact further manual tweaking is required to max out these Samsung B-die memory kits.

 

Developed by Czech developer and HWBOT member Havli, HWBOT x265 Benchmark is based on the open source x265 encoder. It can take advantage of modern CPUs instructions and scales well with multi-core processors. With two presets available, 1080p and 4k, the main workload involves converting H264 source video to H265/HEVC and measure average fps. HWBOT x265 Benchmark v2.0.0 is based on improved x265 encoder, build 2.0.0.24 (compiled by GCC). Compared to previous version (1.7.0.374 GCC), the new one greatly improves encoding speed on virtually all processor architectures. Also multi threading seems to be better - the bigger your CPU is, the better relative fps gain you can expect.

 

 

 

And that same pattern appears with each benchmark, the Torque series deliver versus the clock speeds and timings they are running at, but there has to be more under the hood than this right ?

 

 

 

Even though Corsair mentions that they warrant clock speeds of up to 3600MHz thanks to their IC binning, we know there's a lot more in these Samsung B-Die ICs. Take into account however we are working initially in a quad channel environment. The motherboard's layout & affiliated BIOS, plus the Processor's Integrated Memory Controller (IMC) do play a big part in the final outcome.

To focus on the raw clocking of this memory kit we loosened the timings from CAS14-16-16-35 to 16-16-16-32 2T Command Rate.

We achieved the following stable memory clocks on the AORUS Gigabyte X299 Gaming 9:

• 3466MHz at 1.35 volt
• 3733MHz at 1.4 volt
• 3800MHz at 1.45 volt
  
• 4000MHz required C19 timings and was still not stable for the entire benchmark suite, not even at 1.65Vdimm

 

   

 

This memory divider test is also interesting to see how big the performance improvement can be when one tends to opt for higher speed memory, hence why we also ran down clocked tests with the same main timings set at C16-16-16-32 2T. We really compare apples with apples at MadShrimps!

 

 

As mentioned on the previous page AIDA64 scales massively with increased memory speed and tight memory timings. Interesting to see that the 4000MHz results are lower than the 3800MHz: this is mainly due to the fact we had to loosen the main timings from CAS16-16-16-32 to CAS19-21-21-35. Even with these looser timings we could not complete the entire benchmark suite. Initially we suspected it to be the X299 combination, though with a quick retest on the Z270 with an Intel Core i7-7700K we ran into similar problems once surpassing 3866MHz mark.

In the other tests reflect the same principle, more speed in general is better, though the exotic speeds of 3800MHz and beyond are better avoided for 24/7 systems.  Running these high frequencies daily put a huge strain on the motherboard and the integrated memory controller of your processor. Also the gain above 3800MHz is not as big as one expects, mainly explained due to the very loose timings of the XMP memory profile.

 

 

The Corsair Dominator Platinum Special Edition Torque show that they still got plenty of oomph under the hood. This without selecting any crazy voltages in the bios voltage tabs. How about 3800MHz running perfectly fine in quad channel through the entire test suite multiple times without any hick-up at 1.45Vdimm on X299?

We wouldn't be MadShrimps if we didn't take the time to unleash some extra power for free! We will be doing this on the Intel X299 platform and also on the next page for the AMD Ryzen platform. Too bad no Threadripper setup was available.

Take note these are daily stable speeds we are showing here, not benchmark only stable for extreme overclocking!!

 

Without any voltage increase for the memory we achieved the below settings to run our entire test suite multiple times and also be 200% HCI memtest stable

 

• 3200MHz C13-13-13-24 2T at 1.35Vdimm vs. 3200C14-16-16-36 2T

 

With a voltage bump to 1.55Vdimm 3600MHZ with nearly identical timings was possible C13-13-13-26 2T

 

• 3600MHz C13-13-13-26 2T at 1.55Vdimm

 

 

click image to enlarge

 

Higher speeds were possible but only in Dual Channel mode on the X299 Gigabyte board. With looser timings as you could read on the  previous page 3800MHz was no biggie, but at CAS13 we would only have solid performing Quad Channel performance at maximum 3600MHz speeds.

How does either the tighter timings on the B-die memory aid your system performance or the combo of higher memory speeds and tighter timings?

 

Starting off with AIDA64 Engineer Edition: Versus the stock XMP settings we spot after dialing in the timings a huge bandwidth increase in the Write and Copy tests. If we add 400MHz on top we achieve an additional boost of around 7000-10000MB/s on all tests versus the 3200 tweaked setting.

 

 

SuperPi 32M seems less impressive in gains, though a 10 sec difference between the XMP and the tweaked settings is quite spectacular. Increasing the memspeed to 3600MHz shaved off another second.

 

 

 

Cinebench, HWBOT's Realbench and X265 Encoding tests tell the same tale, a nice performance increase versus XMP and that little extra boost on top with increased memory speeds.

 

 

Our test on the AMD Ryzen platform is pretty fast and straightforward. We tested XMP again versus the highest possible 4 DIMM Dual channel speed we could run the test suite on.

Too bad for the Torque memory, but our Ryzen 1800X stopped already at 3333MHz. Booting with 4 DIMMs is heavy on the integrated memory controller; with only two DIMMs we had no issues at 3466MHz.

 

 

Running 3200MHz XMP with 4 dimms is no problem on the right motherboard

 

 

3333MHz run at 14-13-13-32 at 1.45Vdimm gives a slighty higher bandwidth, better encoding performance and Cinebench score, Though far from spectacular and not at the same pace of the Intel tests.

 

 

 

The Corsair Dominator Platinum Special Edition Torque series are one of the prettiest memory modules money can buy. The Heat-treated effect on the top bar and the black brushed heat spreader are a real feast to the eye. While some might want to debate the lack of RGB I find it daring that Corsair brings out a high end memory kit sporting just a solid white LED light.

 

 

Regarding speeds Corsair chose not to go straight into the higher frequency regions.  A wise choice as for most platforms 3200MHz is plenty for great all-round performance. Anyway, running 3600MHz is warranted thanks to the B-Die IC binning done by Corsair. Nevertheless how high you can push these will depend a lot on the opted platform you tend to run these sticks on. Running very high memory frequencies is not just a matter of plug in and play. The memory controller of your processor and the motherboard must be able to handle the provided bandwidth. For Intel's Z170/270 I see no problem, neither for X299. The older Intel X99 and any DDR4 AMD platform might be another cup of tea. Hence why Corsair labels these as optimized/compatible for Intel 100/200 platforms.

Secondly, while one would expect loose timings of CAS Latency 16, Corsair opted to go a bit tighter than most vendors and chose CAS 14 for their XMP profile. On paper this looks extremely good, but this lower latency will not make up for the raw frequency of slightly higher specced kits. Look at the tweaked results, plenty more power is up for grabs.

The tweaking potential is good for a quad channel kit, with speeds of up to 3800MHz on the X299 platform. Yet B-Die based IC memory kits are more than just raw frequencies. One can tune and tweak these to get the best performance possible. The 3600MHz CAS 13 clearly shows there is a lot more unexplored potential. Take note that these frequencies and timings could be used for daily usage and not solely for benchmarking. For the die-hard overclockers on Z270 we could run frequencies of 3866MHz at CAS 12. These settings are not for the faint-hearted as it required 1.85V DIMM. 4000MHz C12 was sadly not possible, with neither of the four sticks. If you are a die-hard OCer, better look elsewhere, this kit will not be your next World Record breaker.

 

 

Who will buy these Torque kits? Possibly users that like to have the most sexy memory kit on the planet, or the ones that prefer to use a limited Edition memory kit, that sports them looks to kill. Now the average Joe might be held back by the steep retail prize, even though they might make everybody drool. One has to remain honest in the fact that Corsair is charging quite a premium for the limited edition looks these Torque memories offer. No arguing possible though on the out of the box performance. These Torques are pretty spot on and a tad better than other offerings when comparing the same memory frequencies. Though if we start to compare that price tag again, it is clear one can buy faster siblings at other vendors. Remember though speed is not everything, a fast memory kit is a combination of good clocks with appropriate XMP timings, a rule that the Corsair engineers followed nicely.

 

Corsair has upped the level of gorgeous looking memories, the Fremont team have proven that there is no real need for RGB. Performance wise the 3200C14 Torques offer a good combination of speeds and timings. Tweakers will love to explore new performance levels with these stunning Samsung B-Die based kits. These sick looking Torque kits are however only available through the Corsair website, a logic move as they are produced in limited numbers, Sadly this leaves no room for any possible debate on the prize.

 

If you want a great-looking memory kit that has plenty of performance under the hood and that will blow your friends minds than these Dominator Platinum Special Edition Torque kits are highly recommended. Get them while they are hot!!

 

 

 

PROS:

• High build quality
• Performant heatspreaders
• Stunning looks
• XMP Speed versus timings ratio
• Good overclocking potential

 

 

CONS:

• Price premium

 

 

 

We wish to thank Pascal and Harry from Corsair for the Torque review sample. Bernice from Gigabyte for the AORUS X299 Gaming motherboard.