|by Misoprostol | August 20, 2007|
Thermalright XWB-1 CPU Water Block Review
Thermalright XWB-1 CPU Block Review
(A Hardware Canucks Exclusive)
Price: Estimated to be ~$72USD upon releaseAvailability: *VERY SOON*
Warranty: 1 Year
Table of Contents
3- What's Included
4- Block Appearance
6- Installation & Instructions
7- Test Bed Hardware
8- Testing Methodology
For years now, Thermalright has been an innovator, and a leader in the high end air cooling market. Today marks the day that Thermalright enters uncharted waters (snicker). Thermalright has released their very first water block, and we will be the first to report the good, the bad, and the ugly. Now that I think about it, there's nothing ugly here. Check out the eye candy.
For manufacturer information about this CPU block, including specifications and compatibility, please refer to this link.
The Compatibility of the Thermalright XWB-1 may seem somewhat limited compared to some of the blocks we've recently reviewed, such as the Koolance CPU-330, but honestly, how many enthusiasts are using a platform other than AM2, S939, or LGA775 at this point? In this reviewer's opinion, the most important bases have been covered.
SPECIAL NOTE: Even though the Thermalright site lists AM2 as compatible, we were not able to install the block on our test AM2 motherboard. More on this in the AMD installation section.
In typical Thermalright fashion, the XWB-1 comes in a plain brown box with the word Thermalright in big, bold letters across the top. The model name "XWB-1" is listed on the side. This company has always let their performance speak for itself and has never relied on flashy packaging to sell their products.
The XWB-1 was extremely well packaged, and came in what was by far the biggest box I've yet seen for a water cooling block. The APOGEE GT's retail packaging is actually larger in length and width, but does not contain nearly the amount of packing material that Thermalright's packaging does. It's clear that they're serious about this piece of engineering reaching the end-user in one piece. That won't stop us from ripping it apart though. More on that later.
Everything was well protected, and our block arrived in tip-top shape.
3. What's Included
Thermalright includes everything you see in the above picture. The CPU water block itself, a simple yet effective instruction page, the hold down plate, spring-loaded screws, a back plate for LGA775, Thermalright-branded paste, and a Thermalright sticker.
The CPU block is wrapped in plastic but has no sticker covering the bottom of it to protect the finish. This is rather unusual, and every other block that's arrived in our labs has had a sticker on the bottom. It could be on account of our XWB-1 being a sample, so I wouldn't be too surprised to see your retail block come with a sticker over the base.
A neat thing that I've never seen before is the way that Thermalright colour-codes their mounting screws. The all silver ones are for LGA775, the Silver/Black ones are for AM2, and the all black ones are for S939. This really takes the guess work out of figuring out which screws are properly threaded for your motherboard/CPU combo.
One emerging trend is the inclusion of an LGA775 backplate with many CPU water blocks. It's good to see Thermalright continue this trend by including a top notch backplate that is designed to work perfectly with their hold down plate and mounting screws. Way to go, Thermalright! One negative thing was that Thermalright neglected to include an AM2/S939 backplate. The one included on our motherboard ended up not being compatible with the hold-down screws provided by Thermalright due to the raised nipples where the screws go in (more on this in installation).
It would be possible to mod this backplate so the block would fit on this AM2 motherboard, but you would be voiding your warranty. Thermalright has advised us that a separate AM2 backplate will be available to customers for use with this water block.
The included Thermalright thermal paste should be familiar enough to most enthusiasts who have used one of their high end air coolers in the past. I won't comment on the performance of it though, because I am not familiar with the OEM who produces it.
Thermalright has taken a very "no-nonsense" approach to their instructions. They were easy to read, and to-the-point. While not being full colour like the Swiftech instructions, they got the job done.
It took me a while to figure out if the included barbs were supposed to be large 3/8" compatible barbs, or if they were very poorly designed 1/2" compatible barbs. Thermalright informed us that they are supposed to be large 3/8" barbs. That means you'll be stretching your tubing quite a bit to get it over the barb. That's ok. That means less chance for leaks.
If you do choose to install 1/2" barbs though, the XWB-1 has enough room for 3/4" OD tubing, but there's no way you'd be able to use plastic host clamps. I'd recommend metal worm drive clamps or a zip ties in this case.
4. Block Appearance
Thermalright, although new to the water cooling scene, is no stranger to producing fine-looking shiny pieces of metal. This block will look great on your CPU no matter how you slice it. It's not as shiny as the Koolance block, but it's still very beautifully finished.
Here's a couple more photos, but these ones have the hold down plate installed on the block.
The base is not the shiniest I've ever seen, but as we saw in our last review, the shiniest base finish is not always the best for performance. The pictures below should give you a good idea of how shiny the base is, and how flat it is. The first photo shows the base reflection of a toonie, and the following three reveal a little bit of distortion and bending in the straight line around where the screws in the base plate are. Hopefully this won't impact performance because it's outside of where a CPU IHS would normally contact the block.
In an effort to provide the most detailed and accurate reviews possible, we try to provide as much detail as possible on the inner workings of every water block we review. If you'd like to see any of these photos at a larger size, please PM me with your email address and I would be happy to email you 2.1 MP images.
First a little bit of background on water block design:
The first thing that came into my mind when I heard that Thermalright was going to release a water block was the fact that they are inexperienced with the differences between water cooling and air cooling. Water block design is not like heatsink design. A water block the size of a heatpipe tower would perform incredibly poorly.
It doesn't take a PhD in physical sciences to tell you that more surface area means faster heat transfer, so why is it that a water block can't be as big as a heatpipe tower cooler?
Many modern high end air coolers use copper heatpipes to transfer heat away from the base of the heat sink and up to a large aluminum fin array. This would be ineffective for water cooling because it would not remove the heatpipes as a bottleneck. The water would be able to remove as much heat as you want from the aluminum fins, but the heatpipes would end up being a bottleneck in such a design.
Water blocks are about having as much surface area as possible, but with the caveat that the area has to be as close to the die of the CPU (where the heat is generated, directly in the center of the CPU IHS - integrated heat spreader) as possible.
There are also other factors to consider. You could design a water block that has TONS of surface area, but restricts the flow of the water so much that it slows to a trickle. This would cause the water up against the copper of the block to have very little turbulence. That means the water contacting the block is not the coldest water in the loop. Warm water = warm CPU. That's not effective either.
That's not to say that restriction is all bad either. Many older style blocks, such as the D-TEK White Water, Danger Den TDX, and Swiftech STORM used a method called "impingement" to intentionally restrict the water, turning it into a jet that was then directed at the area directly above the CPU die. Impingement increases the turbulence and velocity of the water inside the block, which both improve performance, but sacrifices the velocity of the water through the rest of the loop (so your GPU block performance may suffer if you are using a very restrictive CPU block and you may need to use a stronger pump).
The recent trend among the leading CPU block manufacturers has been to abandon impingement in favor of more "free-flowing" designs that are optimized to provide good system performance without high powered (loud) pumps.
With all that background information behind us, it's time to tackle the internals of the XWB-1.
Here's a teaser
And now the meat. Once we took apart the XWB-1, it became more clear how the Thermalright engineers have worked their magic. The center inlet is located almost directly above the CPU die. This design has been quite popular since the release of the White Water some years ago, and is still in use by many manufactuers. Unlike the White Water or the Storm, the impingement used in this block is not actually done on the water coming down the center onto the die. Rather, it occurs throughout and between the pins of the baseplate. The water flows down into the center of the pins, where it is forced to squeeze out between the pins due to the soft material that pushes down in a circle on the micro pin pattern.
The pins on the XWB-1 are by far the smallest I've seen, and although it's an impressive display of manufacturing prowess, it is not necessarily a good thing. My initial evaluation of the design leads me to believe that it would be very restrictive, meaning that with an XWB-1 in your loop, water would flow at a significantly reduced rate.
One somewhat confusing aspect of the design was the lack of a seal between the top of the block and the mid-plate. It seems to me that with a seal to keep the water from the center barb directed at the pins below, better performance may be achieved. It's possible, however, that this compromise was made in order to keep the restrictiveness of the block under control.
At this point in my analysis, it is unclear to me exactly how well the block will perform, but I can tell that it will be heavily dependent on the amount of flow I am able to supply to it.
6. Installation & Instructions
Installation will be shown on spare mATX LGA775 and AM2 motherboards that were lying around in the lab because it would have been a pain to dismantle my test bed for the sake of showing the stock mounting hardware when I don't intend to use it for my testing. The installation procedure shown here is done according to the instructions included with the block and is NOT what is used in testing the blocks (see more on our standardized testing methodology in the appropriate section).
In an odd departure from the usual procedure for installing aftermarket coolers, Thermalright indicates that they would like users to spread thermal compound on the base of the water block, as well as on the CPU. We used a finger wrapped in plastic wrap to do this.
The next step is to do the same thing on the CPU. This step is common between the AMD and Intel platforms, so we will split the instructions into two columns after these two steps.
The next step is to put the backplate on the back of the motherboard. The four nipples should protrude through the 4 mounting holes in the PCB.
Afterwards, you're ready to put the CPU block down on your processor.
Place the hold down plate on top of the block, and use the all stainless steel spring-loaded screws.
At the end it should look like this:
Installing the XWB-1 was not quite as simple on our AM2 motherboard. First we had to remove the stock mounting bracket
Then the block was placed on top of the CPU and the hold down plate on top of the block.
As indicated in the instructions, we used the black screws with the stainless steel springs. Unfortunately they did not reach because the raised plastic nipples on the AM2 backplate included with our motherboard block the screws from reaching the threads in the backplate. It's possible that the backplate could be modded to make this work, but rather than void the warranty on your motherboard by altering it physically, we'd recommend you pick up the Thermalright AM2 backplate that is available separately. It's disappointing that Thermalright was not able to design a mounting system that would work on ALL AM2 motherboards, especially with the advertised "AM2 compatibility"...
Here's the block mostly assembled, but the screws were not able to reach the threads.
7. Test Bed Hardware
Gigabyte GA-965P-DS3 rev.1.0
Intel Core 2 Duo E6400 @ 3.6GHz 1.5Vcore (in BIOS) 8x450
*one test run was completed at 3.2GHz 1.4Vcore (in BIOS) 8x400*
2x1GB OCZ PC2-6400 @ 2.0V DDR2-900 5-5-5-15
eVGA 7900GS KO
Silent Green Modular 600W power supply
Samsung 80GB SATA2 HDD
Lite-On IDE DVD-ROM
Microsoft Windows XP MCE 2005 (ok technically not hardware but where else was it going to go?)
Core Temp (what are ya gonna do...)
Water Cooling Hardware:
APOGEE GTX (bowed)
APOGEE GT (bowed)
Magicool CPU Block
STORM G4 (one test run at each setting)
Swiftech MCP655 variable (set to 5 for all performance tests and 1 for bleeding)
Clearflex 1/2" ID 3/4" OD tubing
Thermochill PA120.2 radiator with 2x Yate Loon D12SL-12 @ 12V
D-TEK High Flow 1/2" ID Barbs were used on all components for consistency
D-TEK FuZion GFX for "higher restriction" testing
Arctic Silver Ceramique thermal compound
this photo demonstrates my Ceramique application method. It's thinner than it looks in the picture
8. Testing Methodology:
The XWB-1 will be tested against the other high end CPU blocks we have in the lab as well as the Koolance CPU-330 and Magicool CPU block that were featured in our last two reviews. In order to provide a reference point for those unfortunate souls still using air cooling, the Tuniq Tower's results on this test bed can be found in our Koolance CPU-330 review, but due to the unfairness of the comparison between air and water cooling, it will no longer be included on our charts.
Top view of the Thermalright XWB-1 on the GA-965P-DS3 Test Bed
For the sake of generating consistent results, all blocks are mounted using the same hardware. A Zalman plastic back plate is used to prevent motherboard warping and other than that, the D-TEK FuZion screws, springs, and thumb nuts were used. Once the block is mounted, mounting pressures are tuned by running my stress program (dual Prime small FFT) and observing changes in CPU temperature through Core Temp while tightening or loosening any of the four corners. The only results included are the best ones we were able to obtain.
Side view of XWB-1 Mounted on the GA-965P-DS3 Test Bed. You can see that 1/2" ID 3/4" OD tubing is a VERY tight fit. You'd be forced to use zip ties or worm drives rather than plastic hose clamps.
In order to ensure that the mount is not an anomaly, each block is mounted a second time. Both results are reported in the performance charts.
These two results are presented as "Mount 1 Idle", "Mount 1 Load", "Mount 2 Idle", "Mount 2 Load" in our performance charts.
Room temperature was held as close as possible to 25C.
The *UPDATED* Test Bed in all its Glory
The system is allowed to sit at the Windows XP desktop for 15 minutes to take an idle reading. Dual Prime95 small FFT is used to load the CPU for 15 minutes, and then a load reading is taken. The CPU is then allowed to sit idle at the desktop for 15 minutes, and idle temperatures are then verified.
Due to the high voltages being run through this CPU, it was necessary to place a 92mm fan over the northbridge, and pointed at the MOSFETs.
Arctic Silver Ceramique was used in all tests to ensure consistent performance and to avoid the curing times associated with Arctic Silver 5 and other such high performance thermal compounds.
At Hardware Canucks we are always striving to obtain the most accurate results. As requested by some of our members, our water block numbers have been painstakingly RE-MEASURED with a Thermochill PA120.2 radiator instead of the Magicool dual radiator used in our previous test bed. For this reason, results in this review do vary from those obtained in the past.
I believe at this time that the changes at the top of the charts is due to the increased flow in the next test setup. The third graph in the results section illustrates what I mean by this. With a D-TEK FuZion GFX in series with the CPU block being tested, the FuZion pulls within one degree of the APOGEE GTX under load, and the Koolance CPU-330's results fall back closer to what they were when we were using the Magicool radiator. To me this means that users planning a low flow loop (weak, silent pump, or several blocks together) should look at the 3rd graph, while the other two will represent the kinds of results our reader with a dedicated CPU loop and a high powered pump can expect.
The XWB-1 puts up a valiant fight against the best that D-TEK and Swiftech can offer. The first graph is with the CPU at 3600MHz 1.5Vcore.
As you can see, the Thermalright XWB-1 is a top performer, coming within 1C (margin of error of the on-die Intel temp sensor) of the top performing block, the (bowed) Swiftech APOGEE GTX. This is phenomenal for a first attempt from Thermalright, but this doesn't tell the whole story. The Thermalright XWB-1 is incredibly flow dependent. What that means is that once the flow rate decreases, performance drops off significantly. The third graph shows what happens to performance with a D-TEK FuZion GFX in the loop. This should give users a good idea of how they can expect these blocks to perform with an under-volted pump, or with a GPU or chipset block in their loop.
For those running a more moderate overclock and who are wondering if they should upgrade their current water cooling solution, the graph below that was made up for this review may be of interest. Each block was tested (one mount, but it was one of the mounts used for the high heat load test above, which has been verified with another mount) once, but it should tell an interesting story about what kind of voltages you have to put through your CPU to actually see a big difference between the top water blocks.
Clearly at 3.2GHz with 1.4Vcore, you will be happy with any current tier 1 water block. The differences in efficiency do not show up until more extreme voltages are applied to the CPU in pursuit of that extra little bit of speed.
As promised, here are the results for these blocks with a D-TEK FuZion GFX in-line. You can see that the Thermalright XWB-1 loses its lead over the D-TEK FuZion and ends up as much as 3 degrees higher than the APOGEE GTX depending on the mount you look at for each. With a full cover block, or a chipset block in the loop, I would expect this trend to continue. It's clear that of the tested blocks, the FuZion is the least flow-dependent with its CPU core temps going up by a trifling one degree with a GPU block in-line, the APOGEE GTX is almost as efficient, gaining only 2-3C, but the XWB-1 loses 3-4 degrees by adding a CPU block.
A limited flow rate test was performed with all blocks available to me at the time of publication. The cooler pictured below was filled with 4 liters of water. The pump pictured below was then turned on at the same time as the stop watch was started, and pumped water into the rubbermaid container.
**I found my notes on the flow rate testing** Here's the ranking from least restrictive to most restrictive of all the blocks I have available to me:
Magicool CPU Block
Swiftech APOGEE vanilla
D-TEK FuZion ~= Swiftech APOGEE GT
Danger Den TDX
Swiftech APOGEE GTX
Alphacool NexXxos XP ~= Thermalright XWB-1
I am fully aware that this is not perfect, but the margin of error is high enough that this is all I'm willing to report. I ran this test only to satisfy my own curiosity about the restrictiveness of the XWB-1.
Everything from the CPU-330 up was able to drain the cooler in less than a minute.
Everything from the STORM down took more than two minutes.
This is NOT the pump I used in my temperature testing, and the only reason I used it was for its 110V power hook up.
I was surprised. Thermalright's first attempt at a CPU water block performs well on our test bed, looks great, and has a good bundle included with it, especially the backplate, and colour-coded mounting screws. Little touches like that go a long way. One thing I'd like to see added to the bundle would be an AM2 backplate, but realistically, how many water cooling enthusiasts are running AM2 CPUs right now...
Unfortunately that's not the end of the story. It became clear to me during testing that our test bed represents an UNFAIR advantage for this block. It's an extremely low restriction setup with a very good quality pump running at full speed. For some users, this is representative of what they are using, but most users are either running their D5 (or equivalent) pump at a lower speed for less noise, or are running other blocks in their loop OR BOTH. What that means for the Thermalright XWB-1 is that its performance is no longer chart-topping once you take away these advantages.
At the end of the day, it comes down to the individual user. This block looks great, and provides the best out of the box* performance in a high flow loop. If you're running a low restriction PA120.X radiator with a high flow pump (D5, DDC w/Petra top, or Iwaki) and nothing else in your loop, the XWB-1 is the block to beat.
*it should be noted that the APOGEE GTX needed to be "bowed" with a thicker O-ring in order to beat the XWB-1
If you're running your pump at half speed for less noise, or you have a GPU (particularly with a full-coverage block) or chipset in your loop with your CPU, then you may find the XWB-1 less than ideal.
- Best out of the box performance in a high end CPU only loop
- Great aesthetics
- LGA775 backplate included to prevent motherboard flexing
- Includes everything you need out of the box for LGA775
- Instruction manual was simple and clear.
- Price seems a little high considering the latest Swiftech price drop on the APOGEE GTX
- Included barbs, although of excellent quality, are 3/8" OD, which is not what most enthusiast users have in their systems right now. Many people will need to replace them.
- Proper AM2 mounting hardware was not included. Perhaps it would work with other style backplates, but not with ours.
- Higher performance drop with a GPU block in the loop than either of its closest competitors.
- The most restrictive CPU block we've ever tested.
The bottom line: Recommended
Review by: Misoprostol