Address:   Home Page
Motherboards Memory Storage Cases/Cooling/PSUs IT Computing Displays Mobile Mac CPUs & Chipsets Video Digital Cameras Linux Gadgets Systems Trade Shows Guides Home Increase Font Size Decrease Font Size Change Page Size
AnandTech.com Blogs


  December 31, 2008

Innovation and Anticipation
blog post by Derek Wilson
What did you think was the most interesting, innovative and exciting developments we talked about in 2008?

If I had to pick something that actually shipped this year, it would be Intel's SSD drives. Those things are the new hotness. Still a bit pricy, and still a bit small, but two of these in RAID 0 can saturate a SATA controller. Consistent latency even in random accesses, no need to defrag, and significantly reduced fear of mechanical failure are great things indeed. The improved responsiveness of the system is quite nice and definitely noticeable when moving back to mechanical drives.

If we don't restrict it to things that are available, but only that we have seen ... well there are some really cool things out there. Having watched stereoscopic movies at IDF and seen stereoscopic games at NVISION, I suspect we all might start wearing glasses more often. Hmm ... I wonder if polarized contact lenses are a good idea. Probably not.

One of the innovations I really want to get my hands on is Lucid's Hydra technology. They claim near linear scaling with multiple GPUs regardless of the game. We sort of have our doubts, as do AMD and NVIDIA. But wouldn't it be cool if they could actually pull it off? A vendor independent multiGPU motherboard that scales better than both CrossFire and SLI in all cases? That'd be very cool.

2008 was quite a year in graphics with new architectures from both AMD and NVIDIA. The unexpected success by AMD with RV770 was quite impressive and really threw NVIDIA's pricing for a loop. Which is great for the consumer. With high performance graphics cards available at good prices, pervasive DX10 hardware, and OpenGL 3.0, OpenCL 1.0, and DX11 in the pipe for the next year or so, it is a good time to be into computer graphics. It's taken a while but we are seeing PC games that surpass console games, and we expect the gap to widen quite a bit in 2009. 

Here's to technology, and happy new year. Let us know what you think is the coolest stuff that happened in 2008 and what you are looking forward to in 2009 

December 31, 2008, 27 comments
  December 29, 2008

Personal Experiences With The Need For NAS and SSDs
blog post by Derek Wilson
In to every geek's life a little rain must fall. This month, that rain fell on my wife, Laura. 

Laura is a sort of right brained artist. She has a bachelors degree in applied mathematics from NCSU, and she's also a musician and an artist. Before our daughter Lorien was born, Laura drew a web comic. Most recently, she was getting into flash. She's totally awesome and I'm the luckiest guy in the world. But that's not what this blog post is about, so I'll get back to the story.

Since laura does a lot of content creation, she tends to spend a lot of time on her computer, which happens to be a MacBook a bunch of her friends and family pitched in to get her for her birthday last year. Since then she has worked on a few new comic ideas as well as a bunch of different flash projects. And getting things done while raising a baby and being pregnant with another child (due in March), it's understandable that backing up data (especially on such a relatively new computer) might not be a top priority. I'm sure you can see where this is going ...

So one day, Laura's computer just locks up while she's playing Civ 4 on the windows partition she installed. Of course, this isn't abnormal -- it is windows after all. But the computer won't boot up again this time. It spends some time trying to boot, but we get the blinking question mark on a folder indicating there is nothing to boot from, and the hard drive is making a very disheartening squeaking and clicking noise. 

That's right, mechanical drive failure. The worst possible thing that could happen to a disk, especially when none of the data has been backed up. For a year and 3 months. Oh yeah, did I mention it's just out of warrantee and we didn't pick up the Apple Care plan? That might not have been the best way to go. I'm more used to building PCs from components and warrantees on individual parts aren't that important to me, but I've decided that when buying whole systems (and especially notebooks) I'll be picking up what ever extended warrantee I can get my hands on.

So... yeah. That really sucked. Laura lost a bunch of data, and there isn't really anything I can do about it (we can't afford physical data recovery). We are going to hang on to the disk in case some day we are in a position to recover whatever data wasn't destroyed from physical damage. But for now, Laura has been quite upset about having lost the data and is a little hesitant about doing anything on the computer any more. Which is a shame, as she's very good and I hate to see her not able to enjoy the things she loves. But I know how upset I get if I accidentally lose something as trivial as a few hours of whatever final fantasy game I happen to be playing when my dog trips over the power cord.

Anyway, I'm trying to rebuild her trust in computers to the point where she doesn't feel nervous doing anything involved and creative. I'm going with a two fold solution. First, I'm trying to set up a cheap NAS device for both explicit back ups of her art and for her to use with Time Machine. So I picked up a 1Tb LaCie Ethernet Disk mini. 

I started with a WD My Book World Edition, but I had a bunch of trouble setting up Time Machine to work with SMB. I actually got it working, but it just wasn't as transparent or easy to use as I wanted, and I seemed to have some speed issues even over gigabit ethernet. The LaCie solution will work much more easily out of the box, and it also supports USB 2.0 which will offer easier recovery in the event of future catastrophic failure. 

I debated going with Apple's Time Capsule, but we've already got gigabit switches and an 802.11n access point. We also wanted to be able to access the drive from our windows based HTPC so we can store some video on it as well. The LaCie shares the drive using multiple protocols, so we can use it with both PCs and Macs. Which was nice.

Anyway, that's the first prong of the attack. After explaining it all to Laura, she asked: but the drive can still fail -- is there anything we can do to keep it from dying? And my reply was that she would want to use SSDs. And thus, some time next quarter we are going to pick up an SSD. The Intel drive is great, but even it is a little pricy and doesn't offer a whole lot of space. We are hoping that by holding out a bit longer we'll see better quality MLC drives from other manufacturers  that can offer some competition to Intel and help bring prices down a bit lower. 

I'll throw this out there: how many of you guys would have gone with an SSD first and a NAS back up disk later? Let me know in the comments.

December 29, 2008, 32 comments
  December 26, 2008

ASUS P6T - Not the Deluxe Version but equally impressive...
blog post by Gary Key

The competition in the X58 market is finally heating up and that is good news for enthusiasts and buyers alike.  The retail introduction of the Foxconn BloodRage and DFI LP X58-T3eHS at the $299.99 price range is significant in this market sector as both boards were expected to be priced near the $400 mark.  This pricing strategy should drive down the cost of the current $300 boards shortly, provided their performance and stability are on par or better than current products.  This is something we should find out shortly as Raja has been busy testing these boards against the EVGA X58 SLI with the sub-zero setup.

In the meantime, the lower end (not really considering the price point is $210~$250) of the X58 market is starting to heat up, finally.  Gigabyte has introduced their GA-EX58-UD3R product at $210 that is a well rounded board but with CrossFire support only.  Gigabyte just introduced their UD4P board with CrossFire and SLI support in the $270 range.  The MSI X58 Platinum (CrossFire only) is available for $220 and after the latest BIOS, this board is really starting to look an excellent performer for the money.  Biostar's TPower X58 is another mid-level X58 product selling for $270 that supports CrossFire and SLI.  We just purchased this board and the Gigabyte UD4P for our retail customer experience program discussed in our X58 roundup and should have results shortly.

All that said, we had a very interesting delivery a couple of days ago.  ASUS dropped off their new P6T board that offers CrossFire and SLI support along with a smorgasbord of features for $250.  Our first experiences with this board have been excellent with both performance and stability matching that of the $300 boards. To be honest, we think it is probably a better upgrade option than the $300 boards for users looking to move to the i7 platform that are not interested in extreme overclocks or rarely utilized features.  Is it better than the Gigabyte and Biostar $270 boards? We have not had enough test time with the other boards yet to make a decision, but it appears they are basically equal when overclocking with an air cooler while application performance is just slightly better on the P6T.

Those wondering what the differences between the P6T and P6T Deluxe can refer to this chart.  The quick run down is a change from the 16+2 phase setup to an 8+2 setup that we find just as stable so far, a single Realtek Gigabit LAN controller instead of dual Marvell controllers, Marvell SAS support is dropped but the two additional SATA ports are retained via a JMicron JMB322, and on-board audio switches from the ADI AD2000B to the Realtek ALC 1200 chipset.  The P6T features true dual-slot card three-way CrossFire or SLI support compared to the two-way dual-slot card support on the Deluxe board.  Also, the P6T features both an LGA 775 and 1366 mounting hole pattern that allows current LGA 775 liquid cooling users to retain their setup.  We still recommend that air cooler users utilize an LGA 1366 setup (native or mount kit update) although you could get by with a LGA 775 design in a pinch.

We will have a first look up shortly featuring the BIOS layout and initial overclock results against our other mid-range boards, but in the meantime we think ASUS has a potential winner here.

Gallery: ASUS P6T

December 26, 2008, 14 comments
  December 19, 2008

New Video about Positive Air Pressure in PC-Chassis
blog post by Christoph Katzer

Silverstone has just released a new video about the positive air pressure they are using in the previously tested Fortress FT01 case. Silverstone's engineers used smoke to show the air circulation within the chassis which looks very impressive. We have tested this feature before with turning the large upper fan around that it exhausts the air. The temperatures were worse which shows that possitive air pressure indeed helps this case with good cooling.

 

Wouldn't this be a nice addition to our chassis-reviews? We will see what Santa Claus is bringing this year.

December 19, 2008, 42 comments
  December 16, 2008

Intel Xeon 5570: Smashing SAP records (scoop!)
blog post by Johan De Gelas
We have emphasized it more than once: the Nehalem architecture is all about regaining the performance crown in servers and HPC, desktop and mobile use were sometimes a bonus, sometimes an afterthought. Today it becomes almost painfully obvious. Just read Anand's thoughts about the Core i7:
 
"The Core i7's general purpose performance is solid, you're looking at a 5 - 10% increase in general application performance at the same clock speeds as Penryn"
and now look at the graph below.

 
Intel has apparantely allowed HP and Fujitsu-Siemens to break the NDA on the Xeon 5570 processor for PR reasons as both companies have published SAP numbers on a Dual Xeon 5570. The Xeon 5570 is based on the same architecture as the Core i7. It is a 2.93 GHz quadcore CPU with 4 times a 256 KB L2-cache and one huge shared 8 MB L3. 
 
 
SAP Sales & Distribution 2 Tier benchmark
 
The SAP numbers are absolutely astonishing, as Intel's dual socket is able to outperform quad socket opteron machines. Based on the scaling of Barcelona, we speculate that a quad Shanghai at 2.7 GHz would obtain the performance of the Dual Xeon 5570 w/o HT.The new Xeon 5570 outperforms the "old" 5450 by 119%!!!
 
These numbers are so high, that we checked and checked again. The database used is the same (SQL Server 2005), so unless there is some incredible tuning parameter that HP and FS have discovered and that we have yet to hear about, that is not it.
 
At this point we have no idea how it is possible that a 3 GHz Nehalem outperforms the latest Opteron by a margin as high as 80% and more. But we can give it a try. In a previous server oriented article, we summed up a rough profile of SAP S&D:

• Very parallel resulting in excellent scaling
• Low to medium IPC, mostly due to “branchy” code
• Not really limited by memory bandwidth
• Likes large caches
• Sensitive to Sync (“cache coherency”) latency
 
One of the biggest bottlenecks for Intel has been the sync latency. It is possible that once the "sync" bottleneck was removed, the intel architecture is able to show it's real integer crunching power thanks to the out of order loads (memory disambiguation) and better branch prediction.Those are two areas where the opteron architecture is still weak.
 
The slightly lower latency of the L3-cache of Nehalem helps too. This kind of software also makes the buffers fill up due to the long dependency chains. Those OOO buffers have been increased and the depencency chains have been shortened by a very low latency L2 cache and relatively fast L3.
 
Still we are absolutely amazed that the difference is this large. We would have expected Nehalem to outperform Shanghai by lower margins. Although we still are a bit skeptical that the difference is this large ("too good to be true" syndrome), we do not see how you could artificially inflate a SAP benchmark. It sure is not as easy as SPECJBB or SPECfp/int. 
 
 
Update (a few hours later): It seems that the SAP page was wrong about HT. It reported 8 threads on 8 cores on the Fujitsu Siemens Primergy Server. The certification page says otherwise: 16 threads on 8 cores. So hyperthreading (SMT) plays probably an important role in this benchmark as the SAP application has very low IPC and is very parallel. So this completely annihilating performance comes from combining a wide superscalar CPU with an excellent Simultaneous Multithreading implementation. Hats off to the Intel engineers...
 
 
 

December 16, 2008, 29 comments
  December 12, 2008

ABS Canyon 695 - Hummer H2 meets Anodized Aluminum
blog post by Gary Key

After our first receipt of Core i7 processors and boards, we realized quickly that 24/7 stable overclocks past 4GHz were going to require excellent air cooling products, but more than likely liquid cooling or TEC systems.  Considering the upper end boards are capable of 3x SLI or Crossfire and have at least 8 SATA ports, it became obvious that we were going to need a different case for enthusiast level motherboard testing.

We still perform all our motherboard tests in an actual case to ensure ease of installation, peripheral locations, and to gage thermal performance under real use conditions among other items of interest; something you typically will not experience on an open test platform.  Another consideration to consider is that we load up most of the storage ports during testing to ensure the various controllers work together properly so we need plenty of drive ports available. The reason being is that compatibility in this area has just fallen apart as of late as more and more controller chipsets are added to the high-end motherboards.

The additional drives, multiple video cards, and other peripherals being installed require a lot of power, so we utilize 1000W+ power supplies.  These power supplies usually end up being slightly longer than the units we test in the midrange systems and require a case design with additional room.  That said, we started looking at new cases for our test beds and found several that we really like. In fact, there are almost too many choices and we really had a difficult time in choosing cases for each our product sectors. 

We have new uATX cases from Zalman and SilverStone, mid-range cases from Antec and Thermaltake, along with high-end cases from Cooler Master and Lian Li.  One case caught our eye as we browsed through the various sites and it comes from ABS.  The ABS Canyon 695 is one of the more interesting cases to make it into our labs in a long time.  It might not have the same kind of Wow factor that one experiences when first seeing the Antec Skeleton, a custom painted Cooler Master case, or one of the esoteric Thermaltake designs, but it is a case that will definitely start a conversation.

We will be taking an in-depth look at this case shortly.  In the meantime, if you were wondering what to get the computer enthusiast who already has just about everything then we suggest you seriously consider this case as a gift this holiday season.  At an $599 MSRP (various deals are available to reduce the price), it certainly would be a luxury gift and for a lot of us the price of this case exceeds the cost of our systems already.  However, ABS is not marketing this as a case for everyone, instead the target market is the very high-end enthusiast who utilizes their system in a variety of ways ranging from serious gaming to a home based server.  Ignoring the marketing speak, it turns out that it makes for a perfect test bed platform.

This case addressed all of our needs for high-end component testing by providing a pull-down motherboard tray for quick installation or removal, three separate cooling/component compartments, plenty of airflow, easy setup for our base water cooling or TEC products, and hot-swap SATA capabilities along with a bevy of drive openings.  Nice touches include a front air-filtering system, three speed fan controller for the four 140mm fans, sound insulation on the side covers, pre-drilled holes for liquid tubes, anti-vibration technology throughout the case, tool-less peripheral installation, optical drives can be configured for right or left side operation, and media ports on top of the case.

The case is handcrafted from black anodized aluminum and although it is a massive case, it is surprisingly light weight for the most part without giving up any structural integrity.  We have a had a blast using this case for our X58 testing and think ABS just about nailed the design and usability aspects.  We were worried about airflow with the front three 140mm fans pushing more air than the single rear 140mm fan, dual 80mm fans, and power supply could exhaust, but it turns out that by running the rear 140mm fan at a speed step higher than the front fans that all was well in the thermal and acoustic departments.  Overall, considering all of the components installed during 3x SLI/CF testing, we had no problem with noise considering the case was sitting next to us during testing.

About the only things we could nitpick after using the case for the past few weeks was the lack of wheels for a case this size, lack of fan control for the 80mm fans, and better vibration dampening for the hard drive bays.  We never really had a problem with acoustics and loading out six hard drives, but we could imagine the plastic rails loosening up just a little over time, especially if you hot swapped drives on a regular basis.

Oh yes, the price is certainly out of reach for many but we have seen $100 rebates or $250 off price schemes when purchased with certain i7 processors.  However, we would not be surprised if this case ended up in the Neiman Marcus holiday season catalog.  But if you can afford it, we strongly recommend this case.  Just consider it an investment for the next few years or as a gift from a very generous Santa Claus.


December 12, 2008, 19 comments
  December 8, 2008

Mirror's Edge PhysX Side-by-Side Video
blog post by Derek Wilson

A little while back, NVIDIA brought us the news that Mirror's Edge for the PC would feature PhysX support and include some neat effects physics. Effects physics, as you may recall, is the physical simulation of things that don't impact gameplay but simply enhance the visual impact of a game. This can range from particle systems to persistent debris enhanced destructibility or more accurate simulation of fluids, smoke or other volumetric effects. The impact is in immersiveness but it doesn't bring game changing aspects of hardware accelerated physics to the table quite yet.

And we haven't seen anything, until Mirror's Edge, that looked promising in terms of adding anything really compelling to a game. The previous video we posted showed some nice potential, but we still haven't gotten the opportunity to play with it ourselves and really feel the difference. We requested a side-by-side video hoping to get a better handle on what, exactly, is improved in Mirror's Edge. NVIDIA delivered.

Here's the original video of Mirror's Edge we posted.

Here is the side by side video showing better what DICE has added to Mirror's Edge for the PC with PhysX. Please note that the makers of the video (not us) slowed down the game during some effects to better show them off. The slow downs are not performance related issues. Also, the video is best viewed in full screen mode (the button in the bottom right corner).

The effects in there can be simulated on either a CPU or an NVIDIA GPU. The advantage to the GPU is performance and NVIDIA indicates that even an Intel Core i7 processor will have a tough time without GPU support. So these effects aren't anything we've never seen before, but it certainly looks like there is just a lot more of it in Mirror's Edge (and not in that really bad too many particles/too much debris sort of way). The glass breaking itself honestly looks the same (or close enough) to us, but the persistent particles are where it's at. Having a little debris stick around and be affected by the character is a nice touch. The cloth, plastic and tarp effects are what look like the real icing on the cake in the game though. The complete absence of the cloth objects when physics is disabled makes an already sparse looking world look pretty empty by comparison.

We still want to really get our hands on the game to see if it feels worth it, but from this video, we can at least say that there is more positive visual impact in Mirror's Edge than any major title that has used PhysX to date. NVIDIA is really trying to get developers to build something compelling out of PhysX, and Mirror's Edge has potential. We are anxious to see if the follow through is there.

Extending this story is the fact that today NVIDIA is announcing that EA and 2K games have both licensed PhysX and will be working with NVIDIA to include the technology in future titles they publish. All EA and 2K development studios will now have license to develop with PhysX for all platforms. This means Mirror's Edge may not be the only EA title going forward to get the PhysX treatment, and 2K will bring PhysX to the table with Borderlands (which is being developed by Gearbox).

It's no secret that NVIDIA wants effects physics and PhysX specifically to become the next big thing. The fact that this game enables all the effects to be run on any hardware at whatever performance it can manage is a very good move. Only enabling the effects with PhysX hardware present isn't the way to get more developers to adopt the technology. If other publishers and developers start to pick up and extend this technique of including effects physics, we could start seeing physics hardware start to live up to its potential. It may be until we have a physics API that is hardware accelerated on all platforms before we really see ubiquitous use in games, but at least NVIDIA and some game developers our there are doing what they can to move the industry forward in the meantime. That doesn't mean we'll blindly be happy with the way developers use the technology, or that we'll talk about PhysX as a must have feature until there are games that make it true. But moving forward is always a chicken and egg problem and we are happy to see NVIDIA staying behind hardware accelerated physics DICE actually trying to do something interesting with it.

December 8, 2008, 82 comments
  December 1, 2008

LINPACK: Nehalem vs Shanghai part 2
blog post by Johan De Gelas
The last post generated some very interesting comments and questions, which I wanted to address. Unfortunately, some people misinterpreted the post as a "the best scores Nehalem and Shanghai can get in Linpack" review.
 
So let me make this very clear: this and the previous blogpost are not meant to be a "buyer's guide". The Nehalem desktop system and AMD "Shanghai" server are completely different machines, targeted at totally different markets. Normally, we should wait for the Xeon 5500 to run these kind of benchmarks, but consider this a preview out of curiosity.
 
Secondly, we were not trying to get the highest possible LINPACK scores on both architectures. We wanted to use one binary which has good optimizations for both AMD's and Intel CPU's. Fully optimized binaries won't even run on the other CPU. Our only goal is to get an idea how the Nehalem and Shanghai architectures compare when running a "LINPACK" alike binary which is optimized to run on all machines.
 
Thirdly, this is not our review of course. This is a blogpost which talks about some of the tests we are doing for the review.
 
MKL on AMD?
Using the Intel Math Kernel Libraries on an AMD CPU is of course a good way to start some heavy debates. As I pointed out in the last blogpost however, in some cases, the slightly older MKL versions still do a very good job on AMD CPUs when you benchmark with low matrix sizes. You don't have to take my word for it of course.
 
Compare the Intel Linpack 9.0 (available mid 2007) with the binary that AMD produced at the end of 2007. AMD made a K10 only version using the ACML version 4.0.0, and compiling Linpack with the PGI 7.0.7 compiler (with following flags: pgcc -O3 -fast -tp=barcelona-64).
 
All the benchmarks below are done on one CPU with 4 GB (AMD, Intel Xeon) or 3 GB (Intel Core i7). Speedstep, Powernow! and Turbo mode were disabled. 
 
LINPACK version 2007
 
As predicted, the ACML binary which was compiled with 2007 compiler is slower than the MKL "2007" version also compiled in 2007. The MKL version runs on any CPU that has support for (S)SSE-3, so it continues to be a very interesting one for us to test. As you can clearly see from the Xeon 5472 (3 GHz) score, it is not fully optimized for the latest 45 nm Intel CPUs with SSE-4. It is a good "not too optimized" version which can be used on both Intel and AMD CPUs.  You can clearly see this as the 3 GHz Xeon 5472 is behind the AMD Opteron 8384. If this Intel Binary was giving the AMD CPUs a badly optimized code path, this would not be possible.
 
As we move forward to 2008,  we have to create a new binary as both AMD and Intel's fully optimized Linpack versions will not run on the competitor's CPU. Intel released the Linpack benchmark version 10.1, which is not fully optimized for the "Nehalem" architecture, but for 45 nm "Harpertown" family.
 
AMD has created a new Linpack binary using ACML 4.2 and the PGI 7.2-4 compiler.  Below you see how the two CPUs compare.
 
LINPACK version late 2008
 
Bottom line is that these LINPACK benchmarks are moving targets like the SPEC CPU benchmarks, as the compilers and libraries used are just as important as the CPUs.When the Xeon 5500 will materialize, LINPACK performance will probably be higher as the binary is built for the "Penryn/Harpertown" family.
 
While it is useful for the HPC people to see which CPU + compiler can offer the best performance, it is also interesting to understand what kind of performance you get when you compile binaries that have to run on all current CPUs. It is pretty hard to compare CPU architectures if you are using totally different binaries.
 
In the next post we'll delve a bit deeper on what is happening with Hyperthreading, Linpack and the new architectures.

December 1, 2008, 35 comments
  November 28, 2008

LINPACK: Intel's Nehalem versus AMD Shanghai
blog post by Johan De Gelas
A "beta BIOS update" broke compatibility with ESX, so we had to postpone our virtualization testing on our quad CPU AMD 8384 System.
 
So we started an in depth comparison of the 45 nm Opterons, Xeons and Core i7 CPUs. One of our benchmarks, the famous LINPACK (you can read all about it here) painted a pretty interesting performance picture. We had to test with a matrix size of 18000 (2.5 GB of RAM necessary), as we only had 3 GB of DDR-3 on the Core i7 platform. That should not be a huge problem as we tested with only one CPU. We normally need about 4 GB for each quadcore CPU to reach the best performance.
 
We also used the 9.1 version of Intel's LINPACK, as we wanted the same binary on both platforms. As we have show before, this version of LINPACK performs best on both AMD and Intel platforms when the matrix size is low. The current 10.1 version does not work on AMD CPUs unfortunately.
 
We don't pretend that the comparison is completely fair: the Nehalem platform uses unbuffered RAM which has slightly lower latency and higher bandwidth than the Xeon "Nehalem" will get. But we had to satisfy our curiousity: how does the new "Shanghai" core  compare to "Nehalem"?
 

 
 LINPACK

 
Quite interesting, don't you think? Hyperthreading (SMT) gives the Nehalem core a significant advantage in most multi-threaded applications, but not in Linpack: it slows the CPU down by 10%. May we have found the first multi-threaded application that is slowed down by Hyperthreading on Nehalem? That should not spoil the fun for Intel though, as many other HPC benchmarks show a larger gap. AMD has the advantage of being first to the market, Nehalem based Xeons are still a few months away.
 
Also, the impact of the memory subsystem is limited, as a 50% increase in memory speed results in a meager 6% performance increase. The Math Kernel Libraries are so well optimized that the effect of memory speed is minimized. This in great contrast to other HPC applications where the tripple channel DDR-3 memory system of Nehalem really pays off. More later...
 
 

November 28, 2008, 61 comments
  November 27, 2008

My Dream Notebook, SSDs and what they mean for Notebooks, Touch Interfaces and more...
blog post by Anand Lal Shimpi
I've been working on a side project with ASUS called WePC.com. The idea is pretty cool: ASUS is tapping the community for ideas on what they'd like to see from its users in future notebook designs. 

ASUS brought in several authors from around the web to talk about various aspects of computing. I wrote a handful of posts about everything from my dream notebook to the future of input devices on PCs

I'd really like your feedback on some of these things I posted over there. They are very short blog posts atypical of the type of writing I normally do here at AT, but I'd like to know your thoughts. 

If you want to comment on the content specifically, head over there and take a look at them. If you want to let me know if you'd like to see similar styled articles over here, respond to this post. For those of you in the US, Happy Thanksgiving :)

November 27, 2008, 15 comments
  November 25, 2008

New Pictures of Cooler Master's Black Edition Bundle
blog post by Christoph Katzer

Cooler Master gave us some more insides on their upcoming Black Edition of the Cosmos chassis. As it seems the changes to the case are rather minor, besides the black paint. There will not be any change to the hard drive bays as we have hoped for. It's a matter of fact that the case seems not changed at all. Even the large flat front door will come as it was. If you are interested in the further outcome and more information about this bundle be sure to hit Cooler Master's special website for the Black Edition.


November 25, 2008, 8 comments
  November 25, 2008

Lab Notes - Gigabyte GA-E7AUM-DS2H - GF9400
blog post by Gary Key

Remember the NVIDIA GeForce 9300/9400 launch a few weeks ago, yeah we forgot about it also, until NVIDIA called last week to remind us this product was no longer on life support. All joking aside, this product introduction turned out to be a paper launch for the most part with ASUS being the only partner that delivered boards into the retail channel, and barely at that. Over the course of the last week, supply is finally catching up to demand with a variety of boards being available through most channels now.

In the meantime, we had this small thing called the i7 launch and it has consumed us for the past few weeks, too much so to be honest. Instead of completing our IGP Chronicles Part 4 with a final look at the GF9300 and AMD 790GX product lines, we ended up waiting and waiting and waiting on the GF9300 product to show up for review. In hindsight, this was the wrong direction to take after it became obvious that the product was being delayed without "officially" being delayed. Apologies are in order for the wait, but fortunately we received our retail review samples from MSI, EVGA, and Gigabyte today. We should have the Zotac GF9300 here by Friday. Our i7/X58 motherboard coverage starts in a couple of days and with those boards tested and out of the way, we can get back to covering product that most of us afford.

We really liked the GF9300 in our launch coverage and considered it to be the ideal chipset for the HTPC and SOHO markets. We had a few problems (growing pains) with the chipset, BIOS, and drivers, but for the most part our concerns have been addressed with the latest driver and BIOS releases. We are still experiencing problems getting CAS4 stable but based on initial testing with the new boards today, it appears the product has matured quickly. This development, along with decent supply, has us strongly recommending the GF9300 product now.

Of course the question now is which motherboard to recommend. We still have significant testing to complete but the Gigabyte GA-E7AUM-DS2H has caught our eye in early testing.

The GA-E7AUM-DS2H features the speedier GF9400 chipset, four DDR2 DIMM slots with 16GB support, Realtek RTL-8211CL Gigabit LAN, Realtek ALC 889a HD audio codec, 5 3Gb/s SATA ports and 1 3Gb/s eSATA port (RAID 0, 1, 0+1, 5, and JBOD), IEEE 1394a via a T.I. TSB43AB23 chipset, an x16 PCIe 2.0 slot, one x1 PCIe slot, and two PCI slots. Gigabyte covers most of the video standards with VGA, HDMI, and DVI-D outputs. All in all, a full featured uATX board that appears to also be a decent overclocker. Our E8400 has reached a stable 450FSB in early testing and the performance of this board is already at the top compared to other boards in this category.

That is it for now, we will be back as soon as possible to wrap up our IGP series. In case this makes any difference in your purchasing decision before Black Friday, we will be recommending the GA-E7AUM-DS2H in our upcoming buyer's guide.


November 25, 2008, 25 comments
  November 21, 2008

Cooler Master Reveals New Luxury Limited Edition Bundle
blog post by Christoph Katzer

Cooler Master announced the pending arrival of a luxury edition of... yes, of what exactly? Cooler Master couldn't tell us exactly what the full package will entail, but the manufacturer at least revealed a picture of a modified Cosmos chassis and some specs and information for a "luxurious limited edition bundle". Within the package users will get a modified Cosmos chassis decked out in black on the outside and inside. The bundle is to provide "the best products Cooler Master has to offer", so we expect it will also include a power supply, and possibly a water-cooling system or air-coolers for the CPU and GPU. There might be a few other items as well.

Cooler Master also indicates the chassis will have some upgrades in the form of two 140mm fans and a new hard drive bay that includes a 120mm fan. The only picture available right now is the above image, but we do know the chassis is "a special version of the popular Cosmos", so it should be similar to the Cosmos 1000 that has been in the market for a while. Let's look at the Cosmos 1000 a bit more to get an idea of what to expect.

Looking at the inside we see two top-mounted 120mm fans, which are now being changed to two 140mm fans. Since the existing Cosmos 1000 cases already have 140mm mounting holes, this change won't present any problems.

The drive bays will needs some changes to accommodate the new "HDD cooling module with 120mm fan". The number and arrangement of 3.5" bays in the module is not yet known; it could remain six like in the existing Cosmos with a fan in front, or the fan could take up two of the existing HDD spots.

We have seen some black painted interiors in the past, for example from Antec, and these were popular among a certain segment of users. Of course, it costs extra to paint the chassis interior and exterior, especially if you use high-quality paint, and you need to account for additional defect rates. Not long ago, Cooler Master offered CSX brand custom paintjobs, so the idea of a case with high-gloss black paint inside and out might have started there. It's unlikely that the CSX artists have anything to do with the painting of the limited Black Edition, however. Here are some additional pictures of the Cosmos 1000 for reference.

Besides the products, Cooler Master also offers all customers buying the Black Edition bundle exclusive membership in Cooler Master's VIP Club. Details of what this entails are scarce, but this membership provides benefits such as "exclusive articles, downloads, forum access, VIP invites, and much more". Cooler Master is keeping the remaining details secret until December 11, at which time we should also hear about pricing (though we'd hazard "not cheap" is a safe bet).

Until then, we're left with more questions than answers, but Cooler Master expects these bundled cases to fly off the shelves, so if this sounds like something interesting you might want to reserve the date. There will only be 500 of these custom cases/bundles available worldwide, each individually numbered, with no plans for further production. Hopefully the additional components will be as special as the case, or users might need to think twice about whether or not the bundle is worth the price of admission.


November 21, 2008, 12 comments
  November 17, 2008

Lab Notes - EVGA X58 SLI
blog post by Gary Key

As of today, the Intel Core i7 processor series and X58 chipset based motherboards are "officially" on sale. With that bit of news out of the way, EVGA is officially launching their X58 SLI motherboard today.  We received a retail board on Saturday and have been running it through the benchmark gauntlet for the past 37 (and counting) hours.  It's amazing what one can try to accomplish with a case of Rockstar and a few AC/DC albums.  However, even unlimited amounts of caffeine and Angus Young failed to get us to the point of having a full preview ready tonight.  That will come later today but in the meantime we would like to provide a recap of our experiences this weekend.

First off, the EVGA X58 SLI motherboard is a solid piece of workmanship, feature laden, and offers excellent performance when compared to its competition from ASUS, Gigabyte, Intel, and MSI.  Aside from some early BIOS problems and a couple of questionable layout decisions, we have throughly enjoyed our time so far with the board.  A few hours ago, we were singing a different tune but have since worked through a few problems with EVGA.

EVGA is prepping a new BIOS at this moment to address some items of interest.  Like most initial BIOS releases, there are usually a few items that need to be corrected.  Most of our items are minor or just recommendations, but we had a couple of issues that had us shaking our heads.  Fortunately, EVGA normally provides excellent customer support and that is the case with this board.

The current BIOS release does not support overclocking profiles, so those enthusiasts who like to overclock will find themselves resetting numerous BIOS options if an overclock setting goes awry.  Not a major item, but one we think should have been addressed before retail release for an enthusiast level board. Especially one that will save users significant keyboard time as they go through the process of figuring out what overclocked settings work or not.

We also discovered during testing that setting the ICH10R to AHCI/RAID operation causes a bit of a problem for us storage hounds.  After switching from IDE mode to AHCI/RAID on the ICH10R, if the user tries to install a drive on ports four or five, then hard drives on ports zero through three will not be recognized.  This also holds true if you attach a hard drive to the JMicron SATA/eSATA ports on the board or install a RAID card on of the available PCIe slots.  This problem will be addressed very shortly. (Update - EVGA delivered a new BIOS this morning to address this problem.)

The only other item of note is that the user will need to disable EIST (speedstep) and CxE for the system to overclock properly. We found out the hard way that disabling all the power management functions but leaving CxE set to Auto will create some very interesting and frustrating overclock sessions. EVGA recommends the Load Optimal settings routine after the first POST and then again if the board does not recover properly from an errant BIOS setting as it could remain in safe mode. Several power management items will be turned off as a result of this action, but CxE is left at Auto with the BIOS setting enabled instead of disabled in this situation. This holds true with the available auto overclocking feature and the results will more than likely end in an BSOD or several of them. CxE will be disabled in these situations in an upcoming BIOS release. (Update - EVGA delivered a new BIOS this morning to address this problem.)

Once we had the board setup properly, we were rewarded with some excellent results from our i920 setup. Our "safe" overclock on air cooling ended up being a 21x196 configuration with Turbo enabled. This resulted in a final core speed of 4115MHz on 1.40V. Actual voltage was 1.411V as the board will slightly overvolt at this point.

Our Corsair 6GB DDR3 Tri-Channel kit afforded us a 8-8-8-19 1T setting at DDR3-1568 on 1.65V. EVGA is still tuning memory performance, so we expect this board to match our settings on the ASUS P6T-Deluxe board that ran at 7-8-7-20 1T. Currently CL5 and TRAS 15 is not available for users who want to push timings at 1066/1333. (Update - BIOS update this morning opens up CL5 and TRAS 15 settings)  Even so, this board offers excellent performance at stock settings or when overclocked.

We will be back shortly with a quick look at the board, BIOS features, overclocking, and general performance results against the like priced ASUS P6T-Deluxe OC Palm Edition. Our full X58 roundup will be published later this week with boards from ASUS, Gigabyte, MSI, Intel, EVGA, and maybe even DFI. In the meantime, we would like to give a shout out to TankGuys for getting us setup with some retail i7 processors this weekend.

 


November 17, 2008, 9 comments
  November 13, 2008

Lab Notes - J&W MINIX 780G-SP128MB
blog post by Gary Key

We recently received several mini-ITX form factor boards for an upcoming HTPC guide. These boards represent a wide variety of chipsets ranging from the Intel G45 to the NVIDIA GF8200. However, none of them prepared us for the motherboard that we received from J&W Technology.

Based on the AMD 780G chipset, the MINIX 780G-SP128MB is one of the most impressive motherboard designs we have seen in the labs recently.  Impressive not only from an aesthetic and layout viewpoint, but the quality of components and available ports put this board ahead of our other candidates.  However, since it is based on the 780G chipset, the board carries a penalty for HTPC users as multi-channel LPCM via HDMI output is not available. If you can live without that feature then we highly recommend this board.   

J&W managed to shoehorn four SATA 3Gbps ports, IDE connectivity, a physical x16 PCIe 2.0 slot, two SO-DIMM slots, a USB 2.0 header providing four additional ports, COM header, IR connector, and an iPartner Control Center header. The PCIe x16 slot actually operates at x4 electronically due to the limitation of available trace counts on the mini-ITX board design. This could be seen as a limitation, but the primary purpose of the board was never to be a high-end gaming system.

That said, we have not noticed any real performance differences when utilizing our HD 4350, HD 4550, or HD 4670 video cards on this board compared to a uATX 780G board from Gigabyte. In fact, an HD 4350 card on this board will offer improved casual gaming performance while also providing multi-channel LPCM audio for another $30 if a user wanted to go that direction. The board supports Hybrid graphics although your money is better spent on a HD 4550 card instead of the previous generation HD 3450. J&W also included 128MB of 800MHz DDR2 memory for side-port functionality.

This board supports current 45W/65W/95W TDP AM2/AM2+ processors including the Phenom 8750 and 9350e series. J&W explicitly states that 125W or higher processors are not supported due to the three phase power delivery design. An additional warning is also provided that 95W TDP processors should only be utilized if proper heat dissipation is available for both the CPU and motherboard.

A single heatsink covers the 780G Northbridge and SB700 Southbridge. Unless your case is well ventilated, the chipset heatsink will require additional airflow to keep thermals under control. Fortunately, J&W is shipping a fairly quiet AVC Digital Home Series 4010 DC fan with the kit. Three fan connectors are included along with decent BIOS support for monitoring and control.

The board features a loaded I/O panel that consists of a PS2 keyboard port, six USB 2.0 ports, Optical and RCA S/PDIF ports, VGA/HDMI/DVI output, an eSATA port run off the SB700, an RJ45 Gigabit Ethernet port, and a audio panel featuring six 3.5mm jacks for the Realtek ALC 885 7.1 channel HD codec.

Installing SATA cables requires some forethought as the area between the heatsink, SATA/IDE ports, and the 24-pin ATX power connector is cramped. We installed the SATA cables on the board before case installation to alleviate any potential problems.

The MINIX 780G-SP128MB features 4GB memory support that is provided by two DDR2 200-pin SO-DIMM slots. This board supports DDR2-533, DDR2-667, DDR2-800, and DDR2-1066 memory speeds. DDR2-1066 support requires a Phenom based processor but without . We purchased 4GB of GSkill DDR2-800 with timings of 5-5-5-15 at 1.8V for $49, so there is not a significant cost penalty for using SO-DIMMs unless you are trying to locate the DDR2-1066 variety.

In the end, J&W has designed and produced a well engineered motherboard that has us excited about the mini-ITX market again.

 

November 13, 2008, 20 comments
  November 12, 2008

Lab Notes - Core i7 920 OC with Patriot Memory DDR3
blog post by Gary Key

We have been busily preparing an X58 motherboard roundup and overclocking guide that will be published shortly.  During the course of testing over the past few weeks we have encountered more than our fair share of problems.  Problems ranged from poor GPU driver support to BIOS releases that had difficulty booting 12GB memory configuration much less offering a decent performance or overclocking experience.

We are still not pleased with multi-GPU performance on the X58 platform and even less pleased that we do not expect to see several problems fixed until December.  That is a story for another time, but had we known ahead of time the number of problems we would encounter, our graphics card of choice for motherboard testing would not have been from AMD.

As for the motherboards, well we are finally at the point with the latest BIOS releases that we would not hesitate to purchase an X58 platform now.  Our latest BIOS releases from ASUS, MSI, Gigabyte, and Intel are finally up to speed, not that they are perfect, but certainly worthy of stable 24/7 operation.   Memory compatibility and performance is greatly improved, on-board peripherals are working properly, and overclocking capabilities are up to speed now. We are utilizing the following BIOS spins for our two roundups, ASUS P6T-Deluxe (0901), ASUS Rampage II Extreme (0602), Intel DX58SO (2724), MSI Eclipse (1.121), and the Gigabyte GA-EX58-UD5 (F3D).  One note of interest is that we had several problems with our Rampage II Extreme board in regards to memory compatibility and overclocking.  The 0602 BIOS certainly improved the situation but ASUS has updated the board and retail units will incorporate the latest component updates.

In the meantime, we have burned through our credit lines at Newegg purchasing various DDR3 memory kits that are specified for use with the X58 platform.  Our latest purchase was 12GB of Patriot Memory PVT36G1600ELK featuring 9-9-9-24 timings at DDR3-1600 on 1.6V or so.  Patriot also offers the PVT36G1600LLK kit that features timings of 8-8-8-24 at DDR3-1600 speeds on the same voltages along with additional overclocking headroom.


One of our new benchmark tests is Microsoft's Flight Simulator X with the Acceleration expansion pack. FSX is multi-core aware and the i7 is a perfect match for this normally CPU intensive game.  Utilizing the ASUS P6T-Deluxe, i7 920, CoolIT System's Freezone Elite, Sapphire HD 4870, and 12GB of Patriot's latest memory we were able to push the system to a 24/7 benchmarking friendly 4009MHz (21x191) with memory speed at DDR3-1528 on 7-8-7-20 1N settings.  Our settings consisted of 1.3975V VCore, 1.64V VDimm, and the secret sauce setting of QPI voltage at 1.3750V.  Our FSX settings were run at Ultra High with AA/AF/DX10 enabled in order to stress the CPU and GPU during the benchmark sessions.

Keeping VCore, VDimm, and VQPI in proper sync is the key to a stable overclock at voltages that will not shorten your processor's lifespan, well not by that much from what we can tell at this point.  Intel's guidance continues to be not setting VCore past 1.55V, VDimm past 1.65V and VQPI (uncore to us) past 1.315V when overclocking.  Kris will go over this in detail shortly, but in my experiences so far, getting these three voltages too far out of sync will quickly cause problems as one of our i965 processors can attest to now. Keeping them in sync and tweaking a few other settings will result in a finely tuned system that is capable of running settings higher than Intel's guidance although it is a warranty buster.  At this point we like to keep VCore and VQPI within 0.0250~0.0375V of each other as you clock up, generally speaking, once you exceed about 1.3V on VCore it is time to start syncing these settings and others in the BIOS.  

For us, VQPI is one of the most sensitive settings in the BIOS and one that will reward the user with higher bclk and mclk capabilities quicker than most other settings (except VCore/VDimm, although clock skews and amplitude are critical tuning settings also).  When tweaking the BIOS, raising VQPI will generally allow memory speeds to either increase or memory timings to decrease without raising VDimm at certain gates.  Depending on the memory utilized, you can even trade off higher VQPI for lower VDimm and improved timings, it's all a balancing act. In fact, our Corsair DDR3-1600 6GB kit is capable of running at DDR3-1066 on 5-5-5-15 timings with 1.58V VDimm and VQPI raised slightly to 1.2125V.

The same holds true for running VCore slightly lower when you are looking for the highest possible overclock without exceeding the capabilities of your cooling solution. While the cascade and LN2 users will see different results and can push settings a lot further, for those of us on air or water, these settings are critical to getting the most out of this platform in a 24/7 environment.

We will be back shortly with additional results. In the meantime, we had excellent results with our first X58 specific memory kits from Patriot Memory.

 

Gallery: