Print this article

Super Talent 2x1GB PC2-8000


Date
: 01/25/07 – 12:08:48 AM

Author
:

Category
: Memory

Page 1 : Index

Manufacturer:
Super Talent Technology Corp.

Price:
$420 USD (Newegg.com)

I said 2007 would be the year of memory here at Overclockers Online and I certainly wasn't lying. I return today with a look at yet another 2GB dual-channel memory kit of the DDRII flavour. Today we will be looking at and discussing a fairly high-end pair of modules from the American based memory specialist, Super Talent Technology. Super Talent is a San Jose based memory manufacturer that has been producing memory products for some 20 years in one of the largest and most advanced manufacturing facilities in the United States.

Super Talent Technologies not only produces enthusiast based memory but also provides memory solutions for the server and mobile markets as well as flash memory, Some readers may not be very familiar with the Super Talent brand but as of late, Super Talent has really been turning heads with a number of superior DDRII modules that have left enthusiasts in forums around the internet speechless with their performance. One such memory kit is what we will be looking at today, the T1000UX2G4.

The Super Talent T1000UX2G4 dual channel memory kit consists of a pair of 1GB SDRAM DDRII modules rated for operation at DDR1000 with timings of 4-5-4-15 at a measly 2.2v. These are quite the agressive timings at this high of a frequency and should be one of the better memory kits I have looked at once I start to increase the voltage in the overclocking section. The modules are draped in a cast aluminum heatspreader that is cured to the modules for maximum heat dissipation and also happens to make these modules stand out from crowd. You will see exactly what I mean in a hot minute, let's get to some photos.

Page 2 : Package & Contents

I wasn't kidding when I said that Super Talent has equipped these modules with a heatspreader that will make then hard to miss.

The modules from Super Talent come in a standard plastic blister pack with a cardboard insert that does the job of brand identification. The modules themselves also do a little bit of work in letting us all know they are Super Talent modules by way of those much talked about heatspreaders.

The familiar Super Talent logo graces the top of the insert with their slogon, "the power of memory". This is about the extent of advertising on the front of the insert.

The back-side of the insert does provide a little bit of information about the enthusiast line of Super Talent memory as well as a couple quotes from reviews as well as the address to their web site. All in all, the package and the insert are pretty standard fare.

Unlike the package though, the modules draped in royal blue are far from standard by any stretch of the imagination. As we saw in the first photo, the modules have a sticker on one side and the Super Talent logo embedded in the design on the other.

The sticker on the one side provides all the important specifications of the modules including the model number, their size, rated frequency, and timings that they are specified to run at. I believe the serial number is the number below the barcode near the bottom.

As stated, the raised lettering of the Super Talent logo on one side of the heatspreader provides the same texture of the raised ridges throughout the modules. I am not entirely sure if the ridges are cast into the heatspreader to increase surface area for aiding in heat dissipation or if they are simply there for esthetics. It is probably a little bit of both. Either way, these modules look great and are definitely one of a kind.

I believe I mentioned in the opening that the cast aluminum heatsinks are are cured to the memory modules and in this photo you can see evidence of that. Usually a thick, almost rubber-like, thermal pad is used to adhere heatsinks to memory modules but that clearly is not the case here. It appears that Super Talent has used a very thin and clear adhesive material similar to what Corsair used on the TwinX PC2-8500 C5 modules I reviewed a long time ago. These heatsinks are cured onto the module and hold extremely well to the ICs underneath. Because of this, I will not be removing the heatspreaders to see what ICs are used as it will likely damage the memory.

The top of the modules are open and free to allow hot air to rise out of the modules. In an actively cooled environment with a fan above the modules, like they will be during my testing, the fresh air is free to move right past the ICs and along the PCB to push the hot air away.

Now, despite the unique appearance of the heatspreaders, these modules are not just for looking at or taking pictures of. These modules have some serious horsepower under the hood so let's take a close look at the specifications.

Page 3 : Specifications

The Super Talent web site relies on PDF files to provide specifications and information about their modules. I am not a big fan of this method but that is just me. The PDF Super Talent provides for this memory kit is more than adequate. Here is a brief quote from that PDF describing the T1000UX2G4:

I said earlier that these modules have very aggressive timings at DDR1000. Here is the complete list of specifications for this memory:

In addition to providing the listed specifications and features I have provided above, Super Talent also lists the actual hardware used for testing these modules at the rated specifications.
<div style='margin-left:1cm;color:#FFFFFF;'>Test Specs:
These modules are tested and guaranteed to operate at
these specs.

- DDR2-1000 / PC2-8000 (see p/n T1000UB1G4).
- 4-5-4-15 Latencies (CAS, tRCD, tRP, tRAS)
- Test Voltage: 2.2V
- Tested on Asus P5WD2
- Tested as a matched pair in a dual channel motherboard[/quote]
The Asus P5WD2 is a 975x based motherboard and compares closely to the P965x chipset for memory clocking. I will be doing the overclocking and testing on an Asus P5B-Dlx, which is P965x based, as well as an offering from DFI for testing on the AM2 platform. I did inquire with Super Talent which ICs were used on these modules since I wasn't going to be removing the heatspreaders and was told that it was Micron D9xxx ICs. There is no way to confirm if they are D9GMH or D9GKX but I am assuming it is one of the two. Here are the specifications of both from the Micron web site:
<div style='margin-left:1cm;color:#FFFFFF;'>D9GMH:
Density: 512Mb
RoHS: Yes
Depth: 64Mb
Width: x8
Voltage: 1.8V
Package: FBGA
Pin Count: 60-ball
Clock Rate: 333 MHz
Cycle Time: 3ns
Op. Temp.: 0C to +85C
CL: CL = 5
Data Rate: DDR2-667[/quote]
D9GMH are rated at DDR667 CL5 with 1.8v but obviously run higher frequencies at tighter timings with more voltage.

<div style='margin-left:1cm;color:#FFFFFF;'>D9GKX:
Density: 512Mb
RoHS: Yes
Depth: 64Mb
Width: x8
Voltage: 1.8V
Package: FBGA
Pin Count: 60-ball
Clock Rate: 400 MHz
Cycle Time: 2.5ns
Op. Temp.: 0C to +85C
CL: CL = 5
Data Rate: DDR2-800[/quote]
D9GKX are rated for DDR800 CL5 at 1.8v and are the highest rated ICs out right now from Micron. They technically are a bit better than D9GMH but also tend not to respond to voltage as well so there is a bit of a trade off. Most higher frequency modules that run CL4 such as these Super Talent modules will have D9GKX but the cost of D9GKX ICs is so high that many manufacturers have simply screened D9GMH that can do the job.

Page 4 : Installation

As we have seen on the previous pages, the heaspreaders on these Super Talent modules are not bloated or larger than standard size. This means that installation isn't different from any other standard sized memory module out there. Even still, here are a few photos of the modules installed in the two testing platforms.

Sure the blue heatspreader doesn't match the color scheme of the DFI LanParty motherboard but there won't be many motherboards that compliment the color well. Regardless of color, the modules easily fit the DIMM slots as if they were standard memory modules without heatspreaders.

As I always do, the modules will be actively cooled, as seen here, during overclocking and benchmarking. I cannot stress enough how much I recommend active cooling for memory, especially in an overclocked environment.

Like the DFI SLI-M2R/G, the Asus P5B-Dlx has no clearance issues with memory that has heatspreaders. This will be the case with every motherboard out there, even motherboards that require DIMM slots right next to each other to be used for dual-channel operation. These Super Talent heatspreaders barely even peak over the top of the actual memory PCB so there is never going to be a problem with large CPU coolers that you wouldn't have with a heatspreader-less module.

Once again, I will be using the Hipro-Tech.com Maximizer to supply voltage with these modules so I am not restricted to the VDIMM voltage options available in the BIOS. The above photo is of the system throughout the running of these modules and as always, a 120mm fan will be utilized for cooling the modules.

Page 5 : AMD Overclocking

The photo show is over, time to get down to some work. As always, I will start off by overclocking the memory at a range of voltages and timings. I will then run some benchmarks at various overclocks that were achieved. The photos in the installation section outline the major players for the hardware being used but here is a complete list of components on the AM2 setup:

I will only be posting stable overclocks in this review, like all my memory reviews. There are varrying levels of stability though and because of this, I have outlined exactly what stability testing every overclock has achieved. The recipe of programs I have put together has proved to be a very good little cocktail in stability testing:

• Dual 32M runs of Super Pi Mod 1.5 (ran at the same time)
• 2 hours of dual Prime95 using Orthos Beta on blend mode
• 1 hour of dual MemTest in Windows using 750MB/instance
• Multiple loops of each 3DMark 01 / 03 / 05
  

  The ticket stubs have been torn, playbills are handed out, everyone is in their seats, and the curtain is about to raise. It is time to start the show, let the overclocking begin. I will be testing the Super Talent T1000UX2G4 memory at three standard sets of timings, 3-3-3-8, 4-4-4-8, and 5-5-5-8.

  
  3-3-3-8
  
  

  

  You will notice that i have gone all the way up to 2.5v with this memory but I do not recommend running DDRII modules over 2.3v for daily use. I simply show higher voltage results for those that are into benchmarking and might be interested to see how the memory scales with voltage. Super Talent only warranties up to 2.2v and going over that will void your warranty. With that said, these modules do incredibley well at 3-3-3-8 and only 2.20v running faster than DDR800. Not many 2GB memory kits will run that kind of frequency at those timings with only 2.2v.

  DDR870 3-3-3-8 @ 2.39v

  

  
  4-4-4-8
  
  

  

  The 4-4-4-8 timing set is close to what the rated timings are for these mdoules at DDR1000 but as you can see, with even tighter timings, DDR1000 is easily achieved. In fact, it only required 2.10v to achieve higher than DDR1000. This memory did really well at 4-4-4-8 and climbed to an almost breathtaking DDR1200 with higher voltages. This memory is definitely at the top of its game and shows an impressive amount of headroom from the stock timings and frequency.

  DDR1010 4-4-4-8 @ 2.10v

  

  
  5-5-5-8
  
  

  

  Loosening the timings up to 5-5-5-8, I found that these modules had no problem reaching DDR1200 at a very low 2.20v. My motherboard/CPU combination starts to have problems around DDR1200 so the overclocking stopped there, regardless of whether or not the memory could go further. So far, I think it is pretty easy to see that this Super Talent memory is definitely some top of the line stuff. It should be interesting to see how the Intel platform handles these modules.

  DDR1200 5-5-5-8 @ 2.20v

  

  
  Page 6 : Intel Overclocking
  

  The Intel overclocking section has a tough act to follow. On the previous page we saw some very impressive overclocks on the AM2 platform and in the past, my P5B-Dlx has almost gone clock for clock with the SLI-M2R/G. This has set my expectations at an elevated level and something tells me that I won't be dissapointed. Here is the hardware that will be used for the Intel overclocking and benchmarks:

  CPU: Intel C2D E6300 (L630A864)
  MB: Asus P5B-Dlx Wifi-AP Edition
  GPU: Biostar 7600GS 256MB DDRIII
  PSU: Silverstone Zeus 560W
  HD: Seagate SATAII 80GB 8MB NCQ
  OS: Windows XP SP2 (with all updates)
  Hipro-Tech.com DDR Maximizer v1.2

  Memory:

  Super Talent 2x1GB PC2-8000 4-5-4-15

  Naturally, my stability testing doesn't change going from the AM2 system to my C2D setup. All the same programs are run for identical periods of time. I can't stress enough that during all of stability testing, the modules are actively cooled and running memory over specified voltage voids warranty and does potentially damage the modules. Here again is the list of stability tests that were ran for each over clock listed below:

• Dual 32M runs of Super Pi Mod 1.5 (ran at the same time)
• 2 hours of Prime 95 using Orthos Beta on blend mode
• 1 hour of dual MemTest in Windows using 750MB/instance
• Multiple loops of each 3DMark 01 / 03 / 05
  

  The Intel overclocking will feature the same timing sets with a slight abnormality. On the P5B-Dlx, running memory at 3-3-3-8 is actually slower and performs worse than 3-4-3-8 due to the chipset on this motherboard. Because of this, I see no point in testing the overclockability or benchmarking this memory on this system at 3-3-3-8. So, as you would assume, overclocking and benchmarks will be ran at 3-4-3-8 instead.

  
  3-4-3-8
  
  

  

  It is tough to compare the overclocking at 3-4-3-8 on my C2D platform to the 3-3-3-8 clocks on AM2 but the Intel overclocking slightly improved over the AM2 clocks. Again, DDR900 wasn't a problem with enough voltage and DDR800 was quite easy to surpass with the rated 2.20v. These modules clearly love running tight timings on both Intel and AM2 with or without high voltages.

  DDR860 3-4-3-8 @ 2.30v

  

  
  4-4-4-8
  
  

  

  Moving to 4-4-4-8 on the Intel platform showed almost even clocking ability with AM2 reaching an impressive DDR1160 at 2.5v and the rated DDR1000 with a paultry 2.10v. The next big milestone at 4-4-4-8 is DDR1100 and the screen shot below shows that it took 2.32v to stabilize that.

  DDR1100 4-4-4-8 @ 2.32v

  

  
  5-5-5-8
  
  

  

  The 5-5-5-8 overclocking on the P5B-Dlx was quite interesting. As you can see, I was off to a great start at DDR1140 with only 2.0v but the party ended rather quickly. The memory didn't really scale with voltage that well maxing out at DDR1200 with 2.40v. I guess the motherboard/CPU wasn't to blame for the DDR1200 wall on AM2 as this memory doesn't really like running over 600MHz on either platform.

  DDR1140 5-5-5-8 @ 2.00v

  

  
  Page 7 : AMD Benchmarking
  

  I have picked a good range of memory frequencies and timings to run the benchmarks at with these Super Talent PC2-1000 modules. My 266 divider on the DFI LanParty NF590 SLI-M2R/G decided that it didn't want to work this week so instead of running DDR750 3-3-3-8 2T at a CPU speed of 3GHz, I switched things up and found that I could get DDR900 3-3-3-8 to run 1T at 2700MHz.

  

  The benchmarks will be ran at the stock timings and frequency of DDR1000 4-5-4-15 and at the maximum that I could overclock the memory with stock timings at 2.20v. There will also be results at the higher frequencies that we saw in the overclocking section as well. Let's see how this memory performs at these settings.

  SiSoft Sandra 07 SP1 – Memory Bandwidth

  

  Lavalys Everest Ultimate 06 – Memory Bandwidth

  

  The bandwidth numbers scale appropriately in both Everest and Sandra. That 1T result at DDR900 shows a nice comparison to the DDR1080 result at 2700MHz. With tighter timings and and 1T command rate, DDR900 is able to outpace DDR1080 in Everest and almost SiSoft Sandra. This goes to show that sometimes, outright frequency isn't everything.

  Lavalys Everest Ultimate 06 – Memory Latency

  

  The Everest latency results show the results I was expecting with the higher memory clocks usually bringing better results. The fact that the higher memory frequencies also have higher CPU clocks helps this as well. Again though, the 3-3-3-8 1T result really shows its strength even at 2700MHz for the CPU. It actually outperforms DDR1000 4-4-4-8 at 3000MHz in latency.

  FutureMark PCMark 05 & 3DMark 01 SE

  
  

  There isn't much to discuss from the Futuremark results. The scores scale about what one could expect except I honestly thought the 3-3-3-8 1T score in PCMark 05 would have matched or beaten the DDR1080 4-5-4-15 score being both at the same CPU frequency, but aparently not. In 3DMark 01, it is a different story with the tighter timings showing a slight lead.

  SuperPi Mod v/1.5 – 1M

  

  SuperPi Mod v/1.5 – 8M

  

  SuperPi Mod v/1.5 – 32M

  

  SuperPi shows an even graduated path from top to bottom. This is no surprise as SuperPi tends to like higher memory and CPU frequency over timings. The only exceptions is the 3-3-3-8 1T result again showing a slight gain to the blue lines but they are so close that the gap is almost non existant.

  Gaming Benchmarks – Far Cry / Half-Life 2 / UT 2004

  

  I was interested to see how the gaming benchmark results were going to end up, mostly between the grey and blue results. Interestingly enough, 3-3-3-8 DDR900 outperforms 4-5-4-15 DDR1080 in HL2 and UT04, the two games that rely on system the most of the three benchmarked. Far Cry seems to favour the higher memory clocks. Either way, this memory certianly offers a range of competitive frequencies and timings to match up well with whatever CPU speed you are running.

  
  Page 8 : Intel Benchmarking
  

  We saw a few nice comparisons in the AM2 benchmarks including a 1T set of results thrown in for fun. Unfortunately the P5B-Dlx cannot run a 1T command rate so it will all be 2T results to compare.

  

  The one thing the P5B-Dlx can do that the NF590 SLI-M2R/G can't do is consistantly run all of its memory dividers. This allows a direct comaprison between DDR900 3-4-3-8, DDR1080 4-4-4-8, and DDR1200 5-5-5-8 all at a CPU frequency of 2520MHz.

  SiSoft Sandra 07 SP1 – Memory Bandwidth

  

  Lavalys Everest Ultimate 06 – Memory Bandwidth

  

  Unlike the AM2 testing, there is a natural progression through the settings that I ran the benchmarks at on this C2D setup. This results in an almost even scaling from top to bottom in bandwidth numbers and will likely result in very similar looking charts all the way to the bottom of this page.

  Lavalys Everest Ultimate 06 – Memory Latency

  

  The latency numbers fall right in line with what should be expected at these timings and frequencies. Frequency apparently trumps tight timings in this chart showing the decided advantage going to the higher frequency every step of the way.

  FutureMark PCMark 05 & 3DMark 01 SE

  
  

  The 3DMark 01 scores apparently follow the same pattern as the bandwidth and latency tests giving a slight advantage to higher memory frequency but the PCMark 05 scores seem to favour the DDR1080 4-4-4-8 result giving that the highest result. Despite this numerical win for DDR1080 4-4-4-8, the scores are so similar that it would be tough to call anyone the decided winner in either of these benchmarks.

  Super Pi Mod v/1.5 – 1M

  

  Super Pi Mod v/1.5 – 8M

  

  Super Pi Mod v/1.5 – 32M

  

  SuperPi is a benchmark that produces results heavily based on memory bandwidth and latency. The above results show that there is very little performance difference between the three sets of results at the same CPU frequency of 2520MHz. These numbers are pretty much exactly what all the other benchmarks have shown us so far. I expect the same from the last set of benchmarks that deal with gaming performance.

  Gaming Benchmarks – Far Cry / Half-Life 2 / UT 2004

  

  The numbers we see here from the gaming benchmarks are very similar to the PCMark 05 and 3DMark 01 results. The higher memory frequencies seem to have a slight advantage in gaming performance for the most part, but the difference is minimal at best. In fact, in Far Cry, DDR1080 4-4-4-8 actually out performs the DDR1200 5-5-5-8 result. If I were to make a judgement as to which settings were best for performance, I would have to say that DDR1080 at 4-4-4-8 seems to be the best overall between the bottom three set of settings.

  
  Page 9 : Conclusion
  

  We have taken a very good look at the Super Talent T1000UX2G4 kit of memory and there isn't enough experlatives that can be said. My main interest with memory is the overclocking and how well the memory runs at a range of timings. The T1000UX2G4 showed a distinct ability to run tight timings ridiculously well. The overclocking results at the timings of 3-3-3-8/3-4-3-8 and 4-4-4-8 are the best I have seen, without a doubt.

  

  At 5-5-5-8 the overclocking ran out of gas around DDR1200 which is normally very impressive but with the performance at the tighter timings, I was hoping for more the 600MHz. Either way, it is hard to complain when memory runs DDR1200 at only 2.20v and 2.40v on the platforms tested.

  The overclocking is great, the stock frequencies and timings are great, and the modules themselves look fantastic. The blue heatspreaders make these modules esthetically pleasing and seem to provide ample cooling because I was able to easily run at higher voltages. There isn't much else to say except that the Super Talent T1000UX2G4 are sure to continue to turn heads everywhere they go, as they have here today.

  Advantages

  Extremely aggressive stock timings at PC2-8000
  Exceptional overclocking headroom at tight itmings
  Cast aluminum heatspreaders work well and look great

  Disadvantages

  Loose timings didn't provide as much overclocking as I wanted
  Cannot find a set for sale in Canada, we want Super Talent up here!

  Overclockers Online would like to thank Super Talent Technology for the review opportunity.