Two memory module companies, G.Skill and ZADAK, have developed a new memory format with double capacity DDR4 memory. These new modules put the equivalent of two standard modules onto one PCB, making one memory module act like two. Both companies have introduced 2x32 GB kits. As a result, these memory modules are double the capacity over regular DDR4 memory, with the downside that they are also double the height. We have tested both of the available kits using this technique, both coming in at DDR4-3200: the G.Skill TridentZ RGB DC DDR4-3200 and the ZADAK Shield RGB DC DDR4-3200.

Double Height Memory?

One of the disadvantages of small form factor motherboards is that they only feature two memory slots. Most consumer DDR4 modules can have a maximum capacity of 16 GB per module, which for mini-ITX would mean a maximum capacity of 32 GB. For ATX motherboards, this usually means a maximum capacity of 64 GB.  Users looking to build those small form factor systems have historically been capped to 32 GB - these new modules doubles that capacity to 64 GB.

There are currently three motherboards on the market validated for this new double height memory:

1. ASUS ROG Maximus XI Apex (Z390)
2. ASUS ROG Maximus XI Gene (Z390)
3. ASUS ROG Strix Z390-I Gaming

These are all Z390 chipset motherboards, and thus applies to Intel's 8th Gen and 9th Gen processors. In this review we used an i7-8700K, and it didn't have any issues. We tested on some other motherboards from MSI and ASRock, however those systems did not post. It would appear that the memory has been developed in conjuction with concurrent validation with ASUS.

Left to Right: ZADAK Shield RGB DC, G.Skill TridentZ RGB DC,
G.Skill TridentZ RGB & Corsair Vengeance LPX

Each memory company is only offering a select number of kits, presumably due to the low catchment rate of the motherboards they are validated for, but also likely due to cost.

G.Skill is offering three different 'TridentZ RGB DC' memory kits, each 2x32 GB:

1. G.Skill TridentZ RGB DC DDR4-3000 CL14
2. G.Skill TridentZ RGB DC DDR4-3200 CL14
3. G.Skill TridentZ RGB DC DDR4-3200 CL14 (looser secondary latencies)

ZADAK by contrast is offering five 'Shield RGB DC' memory kits, each 2x32 GB:

1. ZADAK Shield RGB DC DDR4-2666 CL16
2. ZADAK Shield RGB DC DDR4-3000 CL16
3. ZADAK Shield RGB DC DDR4-3200 CL16
4. ZADAK Shield RGB DC DDR4-3200 CL14
5. ZADAK Shield RGB DC DDR4-3600 CL16

All of these kits (except ZADAK's DDR4-2666) have a rated voltage of 1.35 V. G.Skill's kits do not have associated MSRPs due to the way G.Skill operates, however ZADAK has shared that their memory kits will start from an MSRP of $800.

The benefit of dense(r) memory is thatmotherboards can double up on maximum capacity without needing four memory slots. All three motherboards listed, the Apex, Gene, and the Z390-I Gaming, only have two memory slots which makes them ideal targets, even though only one is mini-ITX and the other two could have had four memory slots in place. The reason why only these ASUS motherboards are supported at this time is due to the collaboration between the companies: ZADAK says the modules were 'co-developed', but failed to go into detail about who designed what. Nonetheless, these modules are ASUS exclusive for the time being.

Double Height: Cooler Selection

Using memory that is double the height of a normal module, certain other restrictions come into play: namely the CPU cooler being used has less room to exist.

The almost universal intention in this environment, unless a low powered CPU is used, is for a liquid cooler to be in play. 

Under The Heatsinks

We'll go into more detail on the specific kits over the page, but both memory modules work in a similar fashion.

Using 8 Gb Samsung B-die chips, a normal 16 GB module would have 16 of them to make it up to capacity. For these modules, to reach 32 GB per module, there are 32 x 8 gigabit chips. These are split into two ranks of sixteen chips, and act as if there are two memory modules on the same channel. Modern mainstream processors support 'two DIMMs per channel', meaning two memory modules per channel, which is why we see motherboards for dual channel processors have a total of four slots. By putting two modules onto one PCB, only one slot is needed to hit 'two DIMMs per channel'.

One would assume the signalling would be different, however within the same channel, a normal set of memory modules would use the same traces and either (a) daisy chain, or (b) split near the end in order to support both simulateously. There are stability advantages to the (b) method, known as T-Topology, however it is often more difficult to do. Either way, because two memory modules on the same channel do not need separate motherboard traces, that is what makes these modules possible. There are some additional fine tuning elements to the system as well, as with all memory.

Ultimately any memory vendor's chips could have been used, but both G.Skill and ZADAK have gone with Samsung B-die, which are known for being good overclocking-focused memory. It likely also helps that the optimizations for one company's kit also helped with the second, as different ICs would have different requirements in the firmware.

Ultimately with this review, we want to answer the following questions:

1. Is there any performance difference against normal 16GB modules, and
2. How is the power draw affected?

The goal for both answers should be negative: we shouldn't expect any performance difference and no power difference. We also do some overclocking to see if they can be pushed harder even in this form factor.

Test Bed and Methodology

Test Setup
Processor	Intel i7-8700K, 95W, $359 / 1ku 6 Cores, 12 Threads, 3.7 GHz (4.7 GHz Turbo)
Motherboard	ASUS ROG Strix Z390-I Gaming (BIOS Version 1003)
Cooling	Corsair H100i V2
Power Supply	Thermaltake Toughpower Grand 1200W Gold PSU
Memory Tested	2x32GB G.Skill TridentZ DC RGB DDR4-3200 CL14 (1.35 V) 2x32GB ZADAK Shield DC RGB DDR4-3200 CL14 (1.35 V) 2x16GB G.Skill TridentZ RGB DDR4-3200 CL14 (1.35 V) 2x16GB Corsair Vengeance LPX DDR4-2400 CL14 (1.2 V) 2x8GB TeamGroup Night Hawk RGB DDR4-3000 CL16 (1.35 V)
Video Card	ASUS GTX 980 STRIX (1178/1279 Boost)
Hard Drive	Crucial MX300 1TB
Case	Open Test Bed
Operating System	Windows 10 RS3 inc. Spectre/Meltdown Patches

 

New Test Suite: Spectre and Meltdown Hardened

Our motherboard benchmarking suite which includes our short form CPU performance tests and gaming tests were selected in conjunction with our new 2019 bench suite. Our test bench OS has been updated with drivers, newer software and as with our CPU testing updates, also includes Spectre and Meltdown patches. 

Pages In This Review

1. Double Capacity, Double Height RAM [this page]
2. G.Skill TridentZ RGB DC Overview: Analysis of the Board Components
3. ZADAK Shield RGB DC Aura2 Overview : Looking that the non-hardware portion
4. CPU Performance
5. Gaming Performance
6. Power Analysis
7. Overclocking Performance
8. Conclusion

G.Skill TridentZ RGB DC DDR4-3200 

The new G.Skill TridentZ RGB DC DDR4 memory features double the number of memory chips of a conventional design, with doubled EPROM control and doubled power management. As mentioned in the introduction, this is essentially two modules on a single PCB. In order to make this system work, the 288-pin layout is slightly different from the regular DDR4 memory pin layout, which is why these modules require per-motherboard firmware updates and validation.

The G.Skill TridentZ RGB DC UDIMMS use Samsung B-Die ICs which are considered by many to be the best around, but not just in performance, but in offering overclocking headroom. Standard DDR4 UDIMMs comes in two main varieties, 1Rx8 (single sided) which means one row with eight chips, or 2Rx8 (double-sided) with two rows of eight chips. The new DC memory is, in essence, a 'fake' 4Rx8 which is four rows of eight chips. The way the DC memory is laid out within the ASUS UEFI BIOS is as if it's operating with two sticks of memory per one physical module.

Comparing the smaller standard DDR4 to the newer DDR4 DC

The G.Skill TridentZ DC RGB memory doesn't differ stylistically from the 'standard' G.Skill TridentZ RGB memory. The main and only difference is in its height with the DC RAM measuring at a height of 60 mm, whereas the regular G.Skill TridentZ RGB sticks have a maximum height of 45 mm. This means the G.Skill DC modules are 33% bigger in size than the regular model. The heatsinks are made from aluminium and have a brushed finish, with one side grey and the other side black.

Along the top of the G.Skill TridentZ RGB DC memory is a plastic strip which diffuses the integrated addressable RGB LEDs. The RGB can be controlled by the ASUS AURA software or via G.Skill's own software. 

ASUS and G.Skill both recommend users use liquid cooling methods on the CPU when this new memory is used due to potential issues when installing tower style air coolers that won't likely fit. This is due to the height clearance and the obstruction between the memory and either the side fans on a tower or the tower itself. For users angling the airflow top-to-bottom, the new memory can get in the way of heatpipes that like to stick out.

Although there is no information in regards to availability as of yet, the G.Skill TridentZ RGB DC memory is available in three different frequencies configurations. The first two are DDR4-3000 and DDR4-3200 kits with CL14-14-14-34 latency timings. The other kit is DDR4-3200 with slightly looser timings. All these kits have an operation voltage of 1.35 V and fully supports XMP 2.0 SPD. The kits are available in 64 GB kits (2x32 GB) only, are currently compatibility limited to three ASUS Z390 motherboards, and they operate in dual-channel.

G.Skill's Trident Z RGB DC Kits for ASUS Z390 Platform
Speed	CL Timing	Voltage	Kit Config.	PN
DDR4-3000	CL14-14-14 34	1.35 V	2×32 GB	F4-3000C14D-64GTZDC
DDR4-3200	CL14-14-14 34			F4-3200C14D-64GTZDC
	CL14-15-15 35			F4-3200C14D-64GTZDCB

No pricing information has been provided by G.Skill or ASUS, but we expect pricing to be at the high end of three figures.

Comparing 2x32GB DDR4-2400 to 2x16GB DDR4-2400

G.Skill TridentZ RGB DC (Left) and TridentZ RGB (Right) XMP 2.0 Latency Timings

When comparing the G.Skill DDR4-3200 14-14-14-34 TridentZ RGB DC UDIMMs (2x32 GB) to its DDR4-3200 14-14-14-34 TridentZ RGB UDIMMs (2x16GB), not much is different in terms of latencies. The only difference is that the DC TridentZ has a slightly higher tRRD_S which represents the number of clock cycles between activated commands from a different bank group.

ZADAK Shield RGB DC DDR4-3200 Overview

We first learned about the new ZADAK 32 GB modules back in September. Since then, ZADAK has announced a total of five different 2x32GB Shield RGB DC kits in four different clock speeds, from DDR4-2666 through to DDR4-3600. There are two kits available at DDR4-3200 with different latencies; at CL14-14-14-34 and with slightly looser secondary and tertiary timings of CL14-16-16-36. All of these kits are compatible with three specific ASUS ROG Z390 models: the STRIX Z390-I Gaming, MAXIMUS XI APEX and MAXIMUS XI Gene.

For this review, we received a kit of the DDR4-3200 CL14-16-16-36. This dual channel 64 GB (2x32 GB) kit has an operating voltage of 1.35 V and features Samsung B-Die ICs. By contrast to ZADAKs other kits on the market, this is quadruple the capacity of its 8 GB UDIMMS.

ZADAK Shield RGB DC Memory in an ASUS Z390 Motherboard

While we know the ZADAK Shield RGB DC memory has a total of 32 memory ICs on its PCB (16 on each side), and it uses a similar pin-out implementation to G.Skill, leading to the specialization that they only work on select ASUS motherboards that have been validated. Unfortunately the ZADAK headspreader is firmly bonded to the chips, meaning removing the headspreader has a chance that it could remove a chip; we were unable to secure any images from ZADAK for underneath the heatspreaders, so at this point we are assuming that they work similarly to the G.Skill modules with double EPROMs onboard and act as two DIMMs on one PCB.

Analyzing at the mechanical design of the ZADAK Shield RGB DC DDR4-3200 memory, they are 73 mm in height which isn't much taller than ZADAKs own 8 GB Shield RGB memory which have a height of 60 mm. This means the heatsinks extend quite a bit further than the PCB with the majority of the height coming as a result of the integrated RGB LED lightbar. The heatsinks feature a three-tone design with black, grey and silver aluminium heat spreaders. The effect is not as pronounced as the G.Skill RGB light bar.

The RGB LEDs built-into the top of the heat spreader is addressable and can be customized with ASUS AURA software.

ZADAK in itself is an arm of Apacer Technology who formed back in 1997 as a DRAM supplier. ZADAK formed in 2015, and would prefer to be identified as a separate entity from Apacer, similar to how HyperX is to Kingston, or Crucial is to Micron.

Using the Taiphoon Burner DRAM software, we can see the JEDEC DIMM Label is listed as 16 GB of 2Rx8. This means the sensor is showing that each module is organised by the firmware into two lots of 16 GB and not one of 32 GB. The latency timings of the XMP 2.0 profile are 14-16-16-36.

ZADAK has better market variation with a total of five different kits.

The baseline model is a DDR4-2666 kit which is the cheapest of the five models at $799 and has timings of 16-16-16-36. This is relatively slow in terms of latency timings for a kit of this speed overall, but so far its the only DC kit from any vendor to operate at just 1.2 V on its XMP 2.0 profile.

Stepping up to the next mark is a DDR4-3000 kit with slightly slower timings of 16-16-18-38 for $899 and operates at 1.35 V. There are two DDR4-3200 kits available with both kits having the tightest timings, and a final DDR4-3600 kit. This DDR4-36000 kit has an MSRP of $1299 USD which is $500 premium over the DDR4-2666 base model which is a considerable chunk of budget for a desktop platform.

ZADAK Shield RGB DC Kits for ASUS Z390 Platform
Speed	CL Timing	Voltage	Kit Config.	PN	MSRP (USD)
DDR4-2666	CL16 16-16-36	1.20 V	2×32 GB	-	$799
DDR4-3000	CL16 16-18-38	1.35 V		-	$899
DDR4-3200	CL14 14-14-34			-	$1199
	CL14 16-16-36			ZD4-SHC3200C14-32GCS	$999
DDR4-3600	CL16 18-18-38			-	$1299

The DDR4-3200 kit which we received for testing has slightly looser timings of CL14 16-16-36 and has a price of $999.

The current availability in retail channels of the ZADAK Shield RGB DC 2x32 GB kits is limited to Japan, China and Australia. ZADAK has stated that its DC kits will be available in Vietnam and Thailand during January, while the kits are planned for launch in the US and EU during Q1 2019.

CPU Performance, Short Form

For our motherboard reviews, we use our short form testing method. These tests usually focus on if a motherboard is using MultiCore Turbo (the feature used to have maximum turbo on at all times, giving a frequency advantage), or if there are slight gains to be had from tweaking the firmware. We put the memory settings at the CPU manufacturers suggested frequency, making it very easy to see which motherboards have MCT enabled by default.

Handbrake 1.1.0: Streaming and Archival Video Transcoding

A popular open source tool, Handbrake is the anything-to-anything video conversion software that a number of people use as a reference point. The danger is always on version numbers and optimization, for example the latest versions of the software can take advantage of AVX-512 and OpenCL to accelerate certain types of transcoding and algorithms. The version we use here is a pure CPU play, with common transcoding variations.

We have split Handbrake up into several tests, using a Logitech C920 1080p60 native webcam recording (essentially a streamer recording), and convert them into two types of streaming formats and one for archival. The output settings used are:

• 720p60 at 6000 kbps constant bit rate, fast setting, high profile
• 1080p60 at 3500 kbps constant bit rate, faster setting, main profile
• 1080p60 HEVC at 3500 kbps variable bit rate, fast setting, main profile

In Handbrake both the G.Skill TridentZ RGB DC and ZADAK Shield RGB DC memory show a noticable benefit.  The results are beyond the margin of error associated as well as both kits performing very similar across all three formats.

Blender 2.79b: 3D Creation Suite

A high profile rendering tool, Blender is open-source allowing for massive amounts of configurability, and is used by a number of high-profile animation studios worldwide. The organization recently released a Blender benchmark package, a couple of weeks after we had narrowed our Blender test for our new suite, however their test can take over an hour. For our results, we run one of the sub-tests in that suite through the command line - a standard ‘bmw27’ scene in CPU only mode, and measure the time to complete the render.

Blender can be downloaded at https://www.blender.org/download/

The extra capacity and design proves ineffective within Blender and actually performs similar to other normal capacity kits at similar frequencies.

Rendering - Cinebench R15: link

Cinebench is a benchmark based around Cinema 4D, and is fairly well known among enthusiasts for stressing the CPU for a provided workload. Results are given as a score, where higher is better. The benchmark was created by MAXON and integrates workloads suitable for applications such as graphic design, VFX, game development and render engines. The testing is split into single thread and multi-threaded performance and is primarily a CPU and graphics benchmark.

Performance in Cinebench R15 proved similar to other kits on test with the G.Skill TridentZ RGB DC DDR4-3200 proving the best of the pack in the multi-threaded test due to sub-timings.

POV-Ray 3.7.1: Ray Tracing

The Persistence of Vision ray tracing engine is another well-known benchmarking tool, which was in a state of relative hibernation until AMD released its Zen processors, to which suddenly both Intel and AMD were submitting code to the main branch of the open source project. For our test, we use the built-in benchmark for all-cores, called from the command line.

POV-Ray can be downloaded from http://www.povray.org/

POV-Ray performance is essentially the same in this test.

WinRAR 5.60b3: Archiving Tool

My compression tool of choice is often WinRAR, having been one of the first tools a number of my generation used over two decades ago. The interface has not changed much, although the integration with Windows right click commands is always a plus. It has no in-built test, so we run a compression over a set directory containing over thirty 60-second video files and 2000 small web-based files at a normal compression rate.

WinRAR is variable threaded but also susceptible to caching, so in our test we run it 10 times and take the average of the last five, leaving the test purely for raw CPU compute performance.

WinRAR is one of the more memory sensitive benchmarks and encoding can gain from having higher performance memory installed. Both kits of DC memory show benefit here and outperform the standard kits consistently.

7-zip v1805: Popular Open-Source Encoding Engine

Out of our compression/decompression tool tests, 7-zip is the most requested and comes with a built-in benchmark. For our test suite, we’ve pulled the latest version of the software and we run the benchmark from the command line, reporting the compression, decompression, and a combined score.

It is noted in this benchmark that the latest multi-die processors have very bi-modal performance between compression and decompression, performing well in one and badly in the other. There are also discussions around how the Windows Scheduler is implementing every thread. As we get more results, it will be interesting to see how this plays out.

The DC memory also displays consistently better performance over standard capacity memory in 7-Zip.

3D Particle Movement v2.1: Brownian Motion

Our 3DPM test is a custom built benchmark designed to simulate six different particle movement algorithms of points in a 3D space. The algorithms were developed as part of my PhD., and while ultimately perform best on a GPU, provide a good idea on how instruction streams are interpreted by different microarchitectures.

A key part of the algorithms is the random number generation – we use relatively fast generation which ends up implementing dependency chains in the code. The upgrade over the naïve first version of this code solved for false sharing in the caches, a major bottleneck. We are also looking at AVX2 and AVX512 versions of this benchmark for future reviews.

For this test, we run a stock particle set over the six algorithms for 20 seconds apiece, with 10 second pauses, and report the total rate of particle movement, in millions of operations (movements) per second. We use a non-AVX version here.

3DPM v2.1 can be downloaded from our server: 3DPMv2.1.rar (13.0 MB)

Results in our 3DPM benchmark are all within a percent from top to bottom and there isn't much benefit.

DigiCortex 1.20: Sea Slug Brain Simulation

This benchmark was originally designed for simulation and visualization of neuron and synapse activity, as is commonly found in the brain. The software comes with a variety of benchmark modes, and we take the small benchmark which runs a 32k neuron / 1.8B synapse simulation, equivalent to a Sea Slug.

Example of a 2.1B neuron simulation

We report the results as the ability to simulate the data as a fraction of real-time, so anything above a ‘one’ is suitable for real-time work. Out of the two modes, a ‘non-firing’ mode which is DRAM heavy and a ‘firing’ mode which has CPU work, we choose the latter. Despite this, the benchmark is still affected by DRAM speed a fair amount.

DigiCortex can be downloaded from http://www.digicortex.net/

Similarly, there is no real benefit to the new memory. The G.Skill is higher than the ZADAK again however.

Gaming Performance

Civilization 6 (DX12)

Originally penned by Sid Meier and his team, the Civ series of turn-based strategy games are a cult classic, and many an excuse for an all-nighter trying to get Gandhi to declare war on you due to an integer overflow. Truth be told I never actually played the first version, but every edition from the second to the sixth, including the fourth as voiced by the late Leonard Nimoy, it a game that is easy to pick up, but hard to master.

Benchmarking Civilization has always been somewhat of an oxymoron – for a turn based strategy game, the frame rate is not necessarily the important thing here and even in the right mood, something as low as 5 frames per second can be enough. With Civilization 6 however, Firaxis went hardcore on visual fidelity, trying to pull you into the game. As a result, Civilization can taxing on graphics and CPUs as we crank up the details, especially in DirectX 12.

Perhaps a more poignant benchmark would be during the late game, when in the older versions of Civilization it could take 20 minutes to cycle around the AI players before the human regained control. The new version of Civilization has an integrated ‘AI Benchmark’, although it is not currently part of our benchmark portfolio yet, due to technical reasons which we are trying to solve. Instead, we run the graphics test, which provides an example of a mid-game setup at our settings.

1080p Ultra

4K Ultra

World of Tanks enCore

Albeit different to most of the other commonly played MMO or massively multiplayer online games, World of Tanks is set in the mid-20th century and allows players to take control of a range of military based armored vehicles. World of Tanks (WoT) is developed and published by Wargaming who are based in Belarus, with the game’s soundtrack being primarily composed by Belarusian composer Sergey Khmelevsky. The game offers multiple entry points including a free-to-play element as well as allowing players to pay a fee to open up more features. One of the most interesting things about this tank based MMO is that it achieved eSports status when it debuted at the World Cyber Games back in 2012.

World of Tanks enCore is a demo application for a new and unreleased graphics engine penned by the Wargaming development team. Over time the new core engine will implemented into the full game upgrading the games visuals with key elements such as improved water, flora, shadows, lighting as well as other objects such as buildings. The World of Tanks enCore demo app not only offers up insight into the impending game engine changes, but allows users to check system performance to see if the new engine run optimally on their system.

Strange Brigade

Strange Brigade is based in 1903’s Egypt and follows a story which is very similar to that of the Mummy film franchise. This particular third-person shooter is developed by Rebellion Developments which is more widely known for games such as the Sniper Elite and Alien vs Predator series. The game follows the hunt for Seteki the Witch Queen who has arose once again and the only ‘troop’ who can ultimately stop her. Gameplay is cooperative centric with a wide variety of different levels and many puzzles which need solving by the British colonial Secret Service agents sent to put an end to her reign of barbaric and brutality.

Power Analysis

One of the interesting aspects of the double height memory is how it affects power consumption. It would be natural to assume that double the number of chips and EPROMs would result in double the power draw of a standard capacity module, and the power per GB should be similar.

To measure the power consumption, we ran Intel's Power Gadget 3.5.0 utility during benchmark runs in our POV-Ray 3.7.1 test and in our Memory Latency Checker. POV-Ray 3.7 is a rendering based benchmark which stresses a system - It's a good indicator of memory stability and overall performance so it made it a natural choice for a power point of view. Our second test involves our MLC2 memory benchmark which is purely memory focused and loads the memory with high workloads as well as testing latency. 

POV-Ray 3.7: link

The Persistence of Vision Ray Tracer, or POV-Ray, is a freeware package for as the name suggests, ray tracing. It is a pure renderer, rather than modeling software, but the latest beta version contains a handy benchmark for stressing all processing threads on a platform. We have been using this test in motherboard reviews to test memory stability at various CPU speeds to good effect – if it passes the test, the IMC in the CPU is stable for a given CPU speed. As a CPU test, it runs for approximately 1-2 minutes on high-end platforms.

On average when directly comparing the G.Skill TridentZ and TridentZ DC RAM, the power consumption on average in POV-Ray was 285% higher. This is a noticeable jump over two sticks and more than double in terms of overall power used. If we convert this down to average energy per gigabyte:

There is still an additional penalty in energy for using the new modules per GB.

Memory Latency Checker: link

Intel's Memory Latency checker is a tool designed to measure memory latency and bandwidth. MLC measures multiple aspects of DRAM with idle and load latencies, cache to cache data transfer latencies and peak memory bandwidth. The benchmark focuses purely on the memory and is influenced by higher clock speeds and latency timings.

Over a longer duration and in a high memory weighted benchmark such as MLC2, the power variation from the double capacity to the standard was more consistent with what was initially expected; double the power consumption for double capacity RAM. For what it's worth, the Corsair Vengeance LPX kit at 1.2 V was no better off from a power consumption standpoint than the 1.35 V kits tested.

If we compare energy per gigabyte, it is actually very competitive compared to the smaller kits. Here, the 2x8GB kit is actually consuming the most energy per GB, which suggests that the static power is a significant proportion of this analysis.

 

DC RAM Overclocking Performance

Memory overclocking can usually be approached in multiple ways. DDR4 Memory kits come supplied with integrated 'XMP 2.0' profiles which means users don't really need to do anything other than clicking a button inside the BIOS in order to achieve the memory kit's rated performance. However manual tweaking is required to go further.

The fine art of memory overclocking is a long and windy road for minimal performance gains. It almost becomes a performance piece, needing a seasoned expert to get the best out of it. However most users can at least adjust the frequency the primary latencies. The frequency which can be changed via the strap/multiplier and through the base clock generator is more common, while the timings are usually done as the second step once a stable frequency has been found. 

Both of the memory modules we have are equipped with Samsung B-die chips. These chips are often considered some of the best overclocking parts, however there is some natural speculation as to how they will perform in this double height form factor.

To measure the overclockability of the DC RAM, the stock performance with the XMP 2.0 profile applied on both the G.Skill TridentZ RGB DC and ZADAK Shield RGB DC 2x32 GB kits are used as a baseline. We exokired both pushing the frequency as high as could be managed, and then with latencies as tight as much as possible. As WinRAR 5.40 proved one of the more memory sensitive benchmarks in our list, we used this to highlight the performance, if any, from the DC memory.

Highest 24/7 Frequency at 16-16-16: DDR4-3500

The maximum overclock we managed to achieve with a voltage of 1.5 V on the DRAM for each kit is as follows:

G.Skill TridentZ RGB DC - DDR4-3500 16-16-16-38
ZADAK Shield RGB DC - DDR4-3500 16-16-16-38

We were unable to go beyond DDR4-3500 even with 18-18-18 sub-timings.

Best 24/7 Latencies at DDR4-3200: 12-12-12

The tightest timings while keeping the frequency at their XMP rating of DDR4-3200 with a voltage of 1.5 V on the DRAM is as follows:

G.Skill TridentZ RGB DC - DDR4-3200 12-12-12-32
ZADAK Shield RGB DC - DDR4-3200 12-12-12-28

Overall Reasonable Overclock: DDR4-3400 14-14-14 at 1.5 V

The highest achieved DRAM frequency achieved with tighter timing adjustments with a voltage of 1.5 V on each kit is as follows:

G.Skill TridentZ RGB DC - DDR4-3400 14-14-14-34
ZADAK Shield RGB DC - DDR4-3400 14-14-14-32

As our maximum achieved overclocks show, there isn't much additional headroom available to push the dual capacity DRAM much further than the XMP 2.0 profiles already allow. Both the G.Skill and ZADAK kits allowed us to overclock up to DDR4-3500 with 16-16-16-38 at 1.5 V. Any attempt to go past DDR4-3500 would result in failed POST even with timings slackened as loose as up to CL18. When it came to tightening up the latency timings, both kits experienced similar timings with the G.Skill TridentZ RGB DC allowing for DDR4-3200 14-14-14-32, with the ZADAK Shield RGB DC giving slightly tighter tRAS timing with DDR4-3200 14-14-14-28.

Users thinking that overclocking will be as prosperous as others have achieved on DRAM in 1Rx8 or 2Rx8 featuring Samsung B-die ICs aren't going to be as happy with the headroom on these modules. The headroom available is limited and the performance displayed in our WinRAR test was mirrored here. On the G.Skill TridentZ RGB DC kit a total of 0.9 seconds with a mixture of frequency and timing adjustments equalling DDR4-3400 14-14-14-34 applied. Overclocking the ZADAK Shield RGB DC memory to DDR4-3400 14-14-14-32 proved the best settings that we tried with a reduction of 1.3 seconds. Answering any questions about the overclocking ability of the memory, our testing with the samples we received proved dreary in comparison to what's usually expected from Samsung B-die ICs.

Double Height DDR4 Conclusion

In this review we have tested the new 32 GB DDR4 modules from both G.Skill and Zadak. These modules are 'double the height' of traditional DDR4 memory, as they essentially use two modules worth of DRAM chips to achieve the double capacity. The modules do this through adjusting the pins on the DDR4 interface as if one memory slot were responsible for two memory sticks. This change means that these modules need to be qualified for motherboards on a per-motherboard basis. To that extent, at the time of the review, these modules are only validated on three ASUS ROG motherboards. These are enthusiast grade motherboards, to go along with enthusiast grade memory.

The benefit of these 32 GB DDR4 modules means that a motherboard with only two memory slots can achieve 64 GB of memory by using 2x32 GB, rather than the current limit of 2x16 GB. The penalty for using the memory comes in the form of physical space on the motherboard - the G.Skill kit we tested is substantially taller than its regular counterpart, while the Zadak kit we tested is marginally taller. In both circumstances, the manufacturers recommend that liquid cooling is used on the processor on the motherboard, given the difficulty that air coolers can have with tall memory.

On the modules we get a doubling of almost everything: there are 32 Samsung B-die chips on each module, leading to 32 GB of capacity per module. There is also double EPROMs, and when installed in the system, each module is seen as two. What we essentially have is the 'two DIMMs per channel' technique exploited by having two DIMMs on one PCB.

The purpose of this review was to examing the hardware, and test to see if there are any performance differences. Also, the presence of Samsung B-die gave us an interesting application - these high-performance chips are usually an overclockers' favorite, so we did a bit of overclocking as well. We also looked into power, which is likely to be a sticking point for these modules.

Performance

Performance across both the G.Skill and ZADAK kits was consistent in our bench suite, and on par with G.Skill's TridentZ RGB DDR4-3200 standard 2x16 GB kit. Before we started, one could have assumed that the double height memory would hinder performance due to its larger size longer traces from the ICs, EPROM and pins - however having a unified design actually seemed to benefit in a few of the scenarios.

For power, we tested using Intel's Power Gadget tool, which has the hooks to test DDR4. The results show have two interesting scenarios. For power consumed per gigabyte, the new memory is fairly competitive:

However, when we look at power per performance, in a benchmark where all the memory kits performed the same, then the value only comes from having the larger memory:

Compatibility

As mentioned, only three ASUS motherboards currently support this new double height DDR4 memory. It means these modules are going after a small market, or it will be up to system integrators to get the best deals for end-users when combining the right motherboard with the right memory. ASUS is keeping the details of how they enabled the new memory under wraps for now, likely seeing it as a competitive advantage in the aggressive motherboard market. 

In the ASUS BIOSes, there's nothing much that gives them away, except that the BIOS shows that four memory slots are populated, despite the motherboard only having two available, confirming that each of these 32 GB memory modules are seen internally as 2x16GB modules. 

Double Capacity, Double The Price

On the pricing front, this is down to the memory companies. G.Skill doesn't typically give official pricing for any of its kits, and the company states they like to be competitive per region on a weekly basis, which means they don't give 'global MSRPs' as they can change. The company did state however that the 2x32GB kits would be priced similar to its 4x16 GB kits. If this is right, the price for the G.Skill DC kits is expected to start from around $650 based on a 64 GB DDR4-3000 kit and closer to $1000 for DDR4-3200.

ZADAK unveiled its full MSRP pricing list to us. The baseline DDR4-2666 CL16 kit starts at $799, and the highest end DDR4-3600 kit finishes $1299.

Obviously, the benefits of the double height memory are focused on one key direction only: the need to have a total of 64 GB of memory in a particular form factor. Out of the three motherboards where this memory is actually qualified, only the Z390-I Gaming is a small form factor motherboard, so for users looking to optimize an SFF system, this would be the one to get. Larger systems rarely have this two memory slot limitation, and so the options for this memory are a lot larger, unless the user absolutely needs to use the Gene or the Apex motherboards from ASUS.