Team Green captures the single-GPU crown, again

Today Nvidia pulls the wraps off its $650 GK110-based 700 series flagship card, the GeForce GTX 780. This board slides directly into the yawning chasm that exists between the $500 GK104-based GTX 680 and the $1,000 GK110-based GTX Titan, though despite its price it’s actually much closer in specs and performance to the Titan than it is to the GTX 680.

Specifications

Like the Titan, the GTX 780 is a GK110 board, so it has all 7.1 billion transistors, a 384-bit memory bus care of six 64-bit memory controllers and two less SMX clusters with 12 for a total of 2,304 CUDA cores. Its 3GB of memory runs at the usual 6GHz clock speed, and its 863MHz core clock speed is just a smidge higher than the Titan’s 836MHz clock. Its 900MHz boost clock is also a bit faster than the Titan’s 876MHz. Overall, it would be fair to refer to the GTX 780 as the so-called Titan LE that has been rumored for a while now, as it’s a basically a slightly neutered version of the Titan, at least as far as gamers are concerned. Performance in games is extremely competitive with the Titan, more so than we thought it would be given its price. When it comes to compute performance, however, the GTX 780 is heavily neutered compared to the Titan, and for good reason. The 1.5 Teraflops of double precision performance that was so welcome in the Titan is nowhere to be seen in the GTX 780, as Nvidia is reserving that feature for the pricier Titan, and is also billing the GTX 780 as strictly a gaming card. The GTX 780 still offers respectable single precision performance though, clocking in at 4 Teraflops compared to the Titan’s 4.5 Teraflops. For comparison, the GK104-based GTX 680 can only push 1 Teraflop of single precision, and its double precision performance is just 1/24th of that by Nvidia’s design. It wants these “cheap” cards to be used for gaming, period, and its expensive Tesla cards to be used for Compute. This is why the Titan costs $1,000 and the GTX 780 costs $650.

GTX 780 Specs

Otherwise the card looks, feels, and runs almost exactly like a Titan. It has the same 10.5-inch length, the same six-pin and eight-pin PCIe connectors, and the same HDMI, DisplayPort , and DVI connectors. It supports up to three-way SLI. The minimum power supply required is 600w, and the card’s TDP is 250w, just like the Titan.

New Technology

Nvidia is releasing the GTX 780 with a fair bit of cool new technology that helps round out the package a bit, so let’s look at each of them one by one.

Adaptive Fan Control

Nvidia’s all-new Adaptive Fan Control smooths out the speed at which the fans spin up and down, resulting in less noticeable noise during operation.

Even though the GTX Titan was and is a “quiet” GPU by our standards, Nvidia tells us that it’s not necessarily fan noise that people notice as much as changes in fan speed, and we have to say there’s some truth to that. Once a fan is spinning at a certain RPM we tend to not notice it, especially as the Titan and the GTX 780′s fans never really spin very fast. It is certainly noticeable though when the fans spin up or down suddenly; we all hear that. So on the GTX 780 Nvidia has smoothed out the speed at which the fans spin up and down so you don’t hear the change in fan speed.

ShadowPlay

ShadowPlay is designed to improve the in-game video recording experience over FRAPS by consuming less system resources and compressing the videos. It will work with any Kepler GPU and will be released this summer.

Now this is a cool feature, albeit one we have not tested as it was not available just yet. This is software that only works with Kepler GPUs to record your gaming sessions by using the GPU’s built-in H.264 video encoder. The software will just record everything you do and keep only the portion that you just played, hence it’s name, as it records everything that you just did, like a shadow. You can tell the software to just keep the last five minutes, 10 minutes, or 20 minutes, and it’ll intelligently delete what is not needed, helping keep file sizes down by both deleting unneeded video and also through video compression. Nvidia also claims the peformance hit from turning on Shadowplay is less than five percent, so it’s main advantages over FRAPS are that it only records what just happened instead of everything, and it requires less system resources to do so. This software will be rolling out this summer and will be available via Nvidia’s GeForce Experience software, and will be supported on any Kepler GPU. It should be noted that beginning with this launch driver, the GeForce Experience software will replace the Nvidia Update software in the driver package, and though users can opt out of installing it, by default it will be installed with all Nvidia drivers going forward.

Head on over to page 2 to read about the rest of the new features, see the benchmarks, and our final thoughts.

GPU Boost 2.0

The GTX 780 features GPU Boost 2.0 temperature target settings as well as the ability to overvolt the board.

This feature first appeared on the GTX Titan and it works beautifully. Its main goal is to prevent the GPU from throttling itself due to excessive heat, which results in reduced performance. To prevent this from happening, users can now set the peak temperature for a card, which by default is set to 81C but it can be pushed up to 95C if you like (the card can handle it). Nvidia tells us these cards can go all the way up to 105C before the hardware is damaged, but you’d be lucky to push either a Titan or a GTX 780 past 80C typically since their coolers are so effective. However, the GPU will overclock as high as it can until it reaches that pre-determined temperature, so you can nudge the temperature GPU Boost 2.0 also allows for overvolting a card, so you can overclock it as high as you can get away with by pushing the core clock power target, temperature target, memory, and voltage. Only the Titan and the GTX 780 offers these features at this time. If you try to use them on a GTX 680, for example, they are simply greyed out.

Performance

When we first heard of this card we figured it would land squarely in between the $500 GTX 680 and the $1,000 GTX Titan, both in performance and price. Then Nvidia informed us the card was priced at $650, and we looked at the benchmark charts and thought this was either a very competitively priced product, or the Titan is really overpriced, at least for gamers. Looking at the benchmark chart (below) you can see the GTX 780 is within five to 15 percent of the Titan in all benchmarks, which is seriously impressive given it costs $350 less. In general we run all games at 2560×1600 with 4XAA enabled, which is extremely taxing on even the most hardcore systems, and yet the GTX 780 was just on the cusp of playing all games extremely smoothly, with the exception of Metro: Last Light, which will remain a ball buster for the forseeable future it seems. Even though we’d prefer to have at least 40-50fps, that is just not possible with a single GPU at the resolution we run, and only the Titan can get close to achieving it. That said, the GTX 780 is damn close, and easily puts some distance on both the GTX 680 and the Radeon HD 7970, making it the fastest sub-$1,000 GPU available at this time. It should also be noted that the card ran cool and quiet throughout testing, and we were able to run it at 1,084MHz with no problems at all.

GTX 780 Benchmarks

Final thoughts

This is one wicked fast GPU, and if it was $750 or so like we thought it would be you would see us waffling a bit between this card and the Titan. However, at $650 it is very competitive, though we’ve yet to compare it to dual GTX 660 Ti cards or GTX 670 cards in SLI, but then you have to deal with dual GPUs. Also, add-in board partners will be releasing their own versions of the GTX 780 with custom cooling solutions, but we do not know at this time whether the boards will be overclocked or offer larger frame buffers.

In our opinion, the best news about the GTX 780 is at the resolution we use for testing there was no such thing as an affordable GPU that could handle it. The GTX 680 and the HD 7970 were all stuck around 20fps or so for newer games, though the Titan could handle them much better. With the GTX 780 we have a truly semi-affordable card that can run newer games at these resolutions and AA settings just fine. Nvidia says the GTX 780 is 34 percent faster than a GTX 680 and that seems about right to us, and 70 percent faster than a GTX 580, so people with older cards will see quite a jump in performance by upgrading. There also might be other 700-series cards in the pipeline, so anyone looking to upgrade might want to wait just a bit before pulling the trigger on a new card.

As far as how AMD will respond, that remains to be seen. The company reminded us that the HD 7970 GHz edition is still the fastest card at its price point of roughly $420, and that is certainly true. Whether or not it will respond with a GCN 2.0 board dubbed Radeon 8790 or similar is anyone’s guess, but given the ferocity of the GPU wars as of late we’d be shocked if AMD sat on the sidelines for very long.

PS: Here is a promotional video Nvidia sent us that walks you through the GTX 780′s hardware and software.

3D notebook offers hefty power for a hefty cost

The Eurocom Scorpius lives a dual life. On one hand, it’s a dull-looking workstation; on the other, this highly configurable laptop can also be outfitted with a 3D monitor and killer gaming specs. We opted for the latter.

The Scorpius has a backlit keyboard that can change to seven different colors.

Along with a 120Hz refresh rate, the 17.3-inch monitor features a built-in 3D emitter that syncs up to a pair of bundled Nvidia 3D Vision 2 glasses. To test the 3D experience, we played the Batman: Arkham City and sampled some 3D movie trailers from Nvidia’s stereoscopic technology rivals the best that movie theaters have to offer in terms of depth, and never bothered our eyes, but the experience wasn’t perfect. Even though Nvidia purports to have solved the dimming issue with its Vision 2 glasses, you’re still essentially wearing sunglasses, so the experience is going to be darker than viewing content in 2D without glasses. We also feel that wearing glasses in and of itself is cumbersome.

The most vexing problem is that when 3D is enabled, performance suffers greatly. Our frame rate dropped by half, from 50s to mid-20s, when playing Batman in 3D as opposed to 2D. In our opinion, these various issues detract too much from the experience; we’d rather play in 2D mode. As for the TN panel in 2D mode, while some sing the praises of 120Hz’s ability to improve even 2D image quality, we didn’t see any noticeable improvement over 60Hz performance when surfing the web, watching videos, or playing games.

We had little complaint with our system’s internal specs, however, which included a 2.8GHz Far Cry 3 to give it a real workout. While the Scorpius wasn’t able to run the super-graphically intensive game on “Ultra” (the game’s highest settings), it ran at “Very High” with frame rates in the mid-40s to low-50s range, and often outperformed a 2.8GHz AMD quad-core gaming desktop armed with a GTX 590. It seems unfair to bring up the fact that the far less expensive ($1,500) GT60 zero-point laptop wasn’t even able to muster a solid 30fps on FC3’s “High” settings. The only performance test where the GT60 bested the Scorpius was in battery life, where Eurocom’s laptop lasted a poor 103 minutes. Something’s got to give when you’re powering two 680Ms.

In terms of storage, the laptop comes with two drives: a 128GB mSATA Micron and 512GB SSD. While the two speedy drives are appreciated, the lack of a hefty HDD really limits the overall package. Luckily, the laptop takes up to three 2.5-inch storage drives and installing an additional HDD requires the removal of just one screw from the bottom of the laptop. Users can also access the motherboard from underneath and swap out the RAM. Eurocom has made the main compartment easily removable, with only four screws holding it in place.

While the laptop certainly performs well, it still has issues. Its trackpad is unresponsive and can be an exercise in frustration to use. Perhaps a bigger problem is that this thing is heavy. With a carry weight of more than 13 pounds, the Scorpius is in backbreaker territory. Finally, at almost $4,000, it’s very expensive.

Fortunately, most of these issues can be mitigated. You can save money by skipping out on the 3D monitor and second SSD. And you can overlook its battery, weight, and trackpad issues by understanding that the Scorpius is best used as a desktop replacement. Its performance is simply off the charts, and complaining about its other problems is like complaining about how a Ferrari is expensive, bad with fuel economy, and doesn’t have enough cup holders. If those issues are enough to bother you, you’re looking at the wrong machine. This laptop is all about performance and here the Scorpius is a stinger.

$3,915, www.eurocom.com

USA: M/A-COM Technology Solutions Inc. (MACOM), a leading supplier of high performance analog semiconductor solutions, has extended its strong industry position in diodes with a new family of broadband Shunt PIN Diodes for high-power switching applications.

The MACOM Shunt PIN Diode series is designed for customers who need a versatile, low cost, ultra-small Shunt PIN diode element for land mobile radio, wireless infrastructure and test instrument applications. Unlike the competition, MACOM’s small, 1.5 X 1.2 mm plastic package reduces board space while enabling broadband performance comparable to chip-scale devices.

Typical applications include high-power switching through 6GHz with incident power up to 100W. Boasting excellent performance, low cost and easy implementation, the Shunt PIN Diode series offers a winning combination for customers looking for high-power diode solutions.

“This family of Shunt PIN Diodes offers an excellent combination of broadband performance, ease of use and low cost,” said Paul Wade, Product manager. “These devices are ideal for customers looking for cost and space savings to implement into their solutions.”

The 3-terminal, low-pass filter structure inherent in MACOM’s devices provides superior low- and high signal performance compared to 2-terminal PIN diode devices, making the Shunt PIN Diodes ideal for high-power switching applications through 12 GHz with incident power up to 100W.

When buying parts to build a new system many people underestimate the importance of the power supply. We receive a lot of emails and Facebook questions from readers who are experiencing system stability issues, many of which we find out are related to poor power delivery.

When buying a power supply it is very important to work out exactly what you will need. A little forethought and planning will mean you won’t be spending extra money on a supply you don’t need, or one that won’t meet your demands or expectations.

For this review today we approached OCZ to send us a range of power supplies at various price points. These units are designed to cater to a wide audience. The budget models on test today are big sellers due to the very competitive pricing. Many OEM builders use these units. The high end models are designed for hard core gamers who are running multiple, expensive graphics cards in Crossfire or SLI configurations.

Before we get into the testing, we wanted to go over some points to help our readers when they are shopping for a new power supply. With such a wide range available online there have never been more choices available to the consumer. To the inexperienced user it can be confusing however. What should you be looking for?

If you have a system chassis with a window and are focusing on a tidy system build, then you will more than likely want to buy a modular power supply. There are several kinds of modular power supply available today. A ‘Pure’ modular power supply will have no hardwired cables. This means when you open the box and remove the power supply, there won’t be any dangling from the unit itself. A separate bag will be supplied, with every cable in it.

Other modular power supplies will have some cables hardwired into the chassis. These hardwired cables will often be the main motherboard connector, a couple of PCIe cables and the CPU 8/4 pin connector – these primary cables are often used all the time.

If you don’t get a modular power supply, then you will end up dealing with a plethora of cables you might not be using. All of these will need carefully routed out of sight behind the motherboard tray. Depending on the case you are using and the specific power supply, this can get messy.

Many computer cases don’t have a lot of space behind the motherboard tray, so you will need to get creative. If the cable routing is a key priority for you, we advise you opt for one of the modular supplies currently available online.

Don’t just blindly buy a power supply with a 1000W+ rating. Unless you are running an ultra high end system with two HD7990′s or four high end graphics cards you are highly unlikely to need one. Our recent review of the Sapphire HD7990 dual GPU graphics card running in a Quadfire configuration is a good example of just how much power you need.

The test system we used comprised a liquid cooled, overclocked Core i7 3960X with several hard drives, 16GB of 2,400mhz memory and two AMD HD7990′s running in Quad Crossfire. This is possibly one of the most powerful gaming systems you could build today and under load it demanded 825 watts. For this system we would recommend a 1000W or 1250W power supply, but what about if you are using a more modest configuration?

Many high gaming systems using an overclocked Core i5 3570k or Core i7 3770k with a GTX680 or HD7970 are likely to be requiring between 300-400 watts of power under load.

An example? Our recent review of the Cyberpower Infinity Fusion Titan SE featured information on system wide power demands. This gaming PC was built around a Core i5 3570k which was liquid cooled and overclocked to 4.6ghz. CyberPower also used the class leading £850 Nvidia GTX Titan graphics card. Was this another 800W power drain system? Not even close, this system demanded less than 330 watts under load.

Many people would think just because a system is using a high end graphics card that a 1000W power supply is needed. It isn’t.

A power supply will reach maximum efficiency between 40% and 60% of its maximum rated load. For the Cyberpower Infinity Fusion Titan SE system mentioned directly above, a quality power supply around 750W would be ideal, leaving a little headroom for adding in some additional hardware at a later date. Cyberpower decided to use the budget oriented Corsair CX600M (600W) which is close to 50% demand at load. Not a bad move really on their part but they were clearly trying to balance the monstrous cost of the GTX Titan to achieve the price point of £1999.99 inc vat.

Obviously this is only part of the procedure for buying the right power supply. You can also factor in the 80 Plus certification. This is the certification of rated efficiency. This will certify products that have more than 80% energy efficiency at 20%, 50% and 100% of rated load, and a power factor of 0.9 or greater at 100% load. That is, such PSUs will waste 20% or less electric energy as heat at the specified load levels, thus reducing electricity use and bills compared to less efficient PSUs.

We would always recommend at least 80 Plus Bronze certification for a budget power supply and when you spend more money, this certification will often increase to Silver, Gold and finally Platinum certification. Be aware that some power supplies don’t achieve their claimed efficiency levels. There have been many articles on this in the past and it certainly complicates the issue. Just because you see an 80 Plus Gold badge on the front of the box, it doesn’t mean the power supply actually has been certified.

That is an article for a different day however.

Ripple performance is also an important factor when deciding on a power supply, because if a specific unit falls outside industry tolerance specifications, it can be potentially dangerous. Ripple occurs with all SMPSU’s (switch mode power supply units). These are tiny fluctuations in a power supplies output voltage on the order of tens of millivolts, millions of times a second. Ripple is measured as a peak-to-peak value in milivolts and is measured using an oscilloscope.

The ATX specification sets these limits on computer PSU ripple levels (meaning we don’t want to see results in excess of these figures below):

+12V – 120mV
+5V – 50mV
+3.3V – 50mV
+5VSB – 50mV

Ideally (in my opinion) we want to see figures in this range:
+12V – 80mV
+5V – 35mV
+3.3V – 35mV
+5VSB – 50mV

A power supply which shows ripple outside (higher) than that listed by the ATX specifications can potentially cause hardware damage. There is no way of measuring when a problem might occur, but it is likely something will go wrong, long term. A power supply that shows good ripple suppression will help aid hardware longevity.

Ripple can have a negative impact on electrolytic capacitors. A ripple reading close to a capacitors rated ripple will potentially shorten the lifespan and ripple exceeding its rated level could cause it to blow. It is worth pointing out that electrolytic capacitors are found on a motherboard, a graphics card, the power supply and more – so the potential for damage is high. Using a power supply with good ripple suppression is very important. If you want to learn more about this, there are plenty of indepth technical articles available online.

Additionally, If you are trying to build a quiet system it is important to factor in the fan performance of the unit. A fan that has to work hard to maintain good internal ambient temperatures will likely produce a lot of noise which can be distracting, especially if the rest of your system has been built with silence in mind.

So let’s get on with the testing of these OCZ power supplies. We will be working our way up the power supply prices the further we get into this review and trying to explain what each power supply would be best suited to handling.

USA: Pasternack Enterprises Inc., an industry leading ISO 9001:2008 certified manufacturer and global supplier of RF and microwave products, has added a new line of SMA and N-Type Test Cables, ideal for testing applications requiring low insertion loss and operating frequencies up to 26.5 GHz.

Pasternack Enterprises’ new RF test cables are specifically designed for environments requiring highly durable test cables, but where expensive ultra-stable cables are not required. New coaxial test cables from Pasternack are available in both in-series and between-series configurations.

These cables use male versions of SMA (26.5 GHz) and N (18 GHz) connectors, and Pasternack’s own PE-P141 type coax, a flexible version of 141 semi-rigid coax. Each test cable is constructed with extremely durable, machined, stainless steel connectors and come standard with a double-shielded coax and FEP jacket.

Pasternack’s new SMA and N test cables are used for many applications including prototyping in test labs, production line testing, and antenna range testing. These rugged RF cables are a cost effective solution for various types of production testing, eliminating the need to buy more expensive versions.

Testing cables from Pasternack are available in standard lengths from 12 to 120 inches and can be ordered in 12-inch increments. Metric lengths can be ordered for each series in 100, 150 and 200 centimeter (cm) lengths.