t-break: Tech @ Its Fastest » Graphics Cards

MSI R7850 Power Edition Review

Taimoor Hafeez — Wed, 23 May 2012 13:29:33 +0000

Features

While MSI already launched their R7850 graphics card a few months back, there’s nothing like a new refresh with added features to entice customers for an upgrade. Today I’ll be looking at the MSI R7850 Power Edition graphics card.

The new ‘Power Edition’ brings a good number of features with it. First and foremost is the new Twin Frozr IV heatsink, which until now was featured only on the R7970 card from MSI. The other thing you’ll notice is how impressively small the R7850 Power Edition is, measuring just 8.5-inches; less than the width of a standard ATX motherboard.

Design

The custom PCB which measures 8-inches uses MSI’s top end Military Class III components, known to be used in their high-end motherboards and graphics cards. Coming back to the heatsink, we see aluminum fins that sit directly on top of the GPU and memory modules, with three heatpipes expanding outside.

The heatsink is covered by two 80mm fans that push air down to cool the components. The whole area is then covered by a cool black, metal casing on top. It’s interesting to note that MSI have used complete metal components as much as possible, and except for the fans, nothing on the heatsink is plastic.

Interestingly the two fans on the Twin Frozr IV heatsink adopt the same ‘Dust Removal’ technology that debuted for the first time in the MSI GTX 580 Extreme Lightning Edition. What basically happens here is that during the first 30 seconds of startup, the fans will run in reverse direction which should help pull off dust particles stuck in the heatsink fins.

The defining factor of the R7850 Power Edition comes from the improved PWM design which allows for massive power efficiencies, which further allows you to control voltage settings for GPU, memory and VDDCI. Of course, what all this, the improved heatsink and PWM, translates into is simply better overclocking. You should be able to take the R7850 Power Edition to the limits.

Nvidia & Zotac GTX 670 (AMP! Edition) Review

Taimoor Hafeez — Thu, 10 May 2012 12:56:35 +0000

Nvidia’s GTX 680 launched 2 months back with much hype, and not only did it live up to the expectations, but also set a new benchmark on how much performance can be squeezed out with such little power.

GTX 680 DNA

What Nvidia has built is a slightly detuned version of the fastest single GPU based graphics card in the world. Looking at the spec sheet below we see a small reduction in the CUDA Cores and core clock speeds, but pretty much everything else remains the same.

We have the same 28nm GK104 chip inside as the GTX 680, the same four 64-bit memory controllers, giving the 2GB GDDR5 memory 256-bit bandwidth and the same number of transistors. Coming in at 9.5-inches, the GTX 670 requires a 500W PSU powered with two 6-pin connectors.

Interestingly the power supply was moved towards the left side, rather than the complete right side as is usually the case. The fan, which is the same one used in the GTX 680, is thus moved to the right end. Sucking in the air, the hot air is then blown out through the left side of the GTX 670, i.e. outside the case. Nvidia have said that thanks to the power efficiencies of the Kepler architecture and the high overclockability of the GTX 670, which partners can make cars with customized heatsinks and much higher clock speeds. Let’s take a look at one such example. Temperatures remained at a fairly middle of the road 37°C when idle, and going up to 76°C under load.

Zotac GTX 670 AMP! Edition

If we take a look at the new Zotac GTX 670 AMP! Edition, we see some major improvements in the cooling solution. First off this is the exact same heatsink we have seen being used in the recently reviewed Zotac GTX 680 AMP! Edition. Measuring 10.5-inches, the GTX 670 AMP! Edition is also cooled by the massive heatsink, with the large copper pipes extending outwards and two big 80mm fans as the GTX 680 AMP! Edition.

Basically what this has allowed Zotac to do is to overclock the card to 1098MHz core clock (boosting to 1176MHz) and 3304 MHZ (6608MHz effective) memory speeds. That’s an overclock of 20% on the core clock and 10% memory speeds. That’s quite an incredible bump in speed over there. The best part, however, was that the temperature remained at 29°C when idling and touching just 60°C under load!

Zotac GTX 680 AMP! Edition Review

Taimoor Hafeez — Tue, 08 May 2012 13:00:55 +0000

Historically most of Zotac’s ‘AMP! Edition’ graphics cards have just had a mild tweak with slightly overclocked speeds. Sure, some of them have had an improved heatsink, but it wasn’t until the GTX 580 AMP²! Edition released late last year that we saw some significant improvements in the design. Today I’ll be looking at the GTX 680 AMP! Edition, but instead of just having a mild overclock, we get a significant heatsink as well.

Let’s take it from the top, and have a look at the goodies packed inside with the GTX 680 AMP! Edition. First off, we have three Assassin’s Creed games: Assassin’s Creed Director’s Cut, Assassin’s Creed II Deluxe Edition and Assassin’s Creed Revelations. These games will not in the least push the GTX 680, but they do look gorgeous.

Next up there are two molex to 6-pin power cables and one DVI to VGA adapter. Sadly there’s no SLI bridge included.

Now let’s have a look at the card itself. The most obvious thing is the huge heatsink on top of the PCB, which looks strangely disproportioned. A closer look will reveal some rather interesting characteristics.

The actual aluminum fins block on top of the PCB is slightly thinner than the overall width of the board. Popping out of it are four rather long heatpipes that stretch out an inch over the width of the PCB. I guess Zotac engineers found this to be the easiest way to dissipate heat, unorthodox as it is.

Stretching out over the aluminum fins is a metal cover, on top of which are the two 80mm fans. This metal cover itself stretches beyond, making the GTX 680 AMP! Editon into a 3-slot card. The hexagonal cuts for the fan further add to the bizarre design.

In total the Zotac GTX 680 AMP! Edition measures in at 10.5×2.2×5 inches. The width, as I mentioned is due to the copper pipes rising well out of the heatsink. Default factory speeds are 1176MHz core clock and 3304MHz (6608MHz effective) memory clock speeds. Compared to reference Nvidia speeds of 1006MHz core and 3004MHz memory clock speeds, that’s an effective overclock of 14.5 % and 10% respectively.

So all in all this is a pretty big card with some serious cooling solution and factory speed; let’s see if it can perform just as impressively.

ASUS HD 7870 DCUII TOP vs. XFX R7870 Black Edition

Taimoor Hafeez — Thu, 03 May 2012 08:23:59 +0000

With the impending release of the $1k GTX 690 and HD 7990, let’s not forget those cards that are still fighting the mid-range fight and vying for our $350. When the HD 7870 launched two months ago it turned out to be pretty impressive. Today I’ll be looking at two custom builds of this card from both ASUS and XFX. Both cards have special heatsinks and also come factory overclocked to some extent.

ASUS HD 7870 DirectCUII TOP

The ASUS HD 7870 DCUII TOP continues the trend of their legendary DirectCUII heatsink that encapsulates some of their best engineering efforts. However, the HD 7870 isn’t the monster that comes in the form of HD 7900 series, so don’t expect an especially heavy duty heatsink. However, what we have here is more than enough to keep the Pitcairn chip running cool.

The PCB itself is 9.5-inches, while the V2.0 updated DCUII heatsink extends out to measure a total of 10.5-inches. The GPU itself is connected to the base of the heatsink, with three heatpipes extending well into the rest of the aluminum fins above. The whole structure covers the entirety of the board, keeping everything cool from the two 75mm fans pushing air downwards into the fins.

The two 6-pin power connectors don’t demand more than the regular HD 7870, requiring a PSU of 450W. In addition to the card itself, the HD 7870 DCUII TOP comes with the Drivers CD and installation guide, plus a CrossFire bridge, a DVI to VGA converter and a 6-pin power cable extender.

Overclocking the ASUS HD 7870 DCUII TOP was a fairly easy deal, with the voltage already maxed out at 1.3v, and factory speeds at 1100MHz Core Clock and 1250MHz (5000MHZ effective) Memory speeds. Compared to the reference AMD speeds of 1000MHz Core and 1200MHz Memory speeds, we already have an overclock of 10% and 4.2% respectively. I managed to bump up these speeds to 1257MHz Core clock and 1304MHz (5216MHz effective) Memory speeds. That’s 14.3% core clock and 4.3% memory speed increase compared to factory settings; a whopping 26% core clock and 9% memory speeds compared to reference AMD speeds.

Nvidia announces GeForce GTX 690

Taimoor Hafeez — Sun, 29 Apr 2012 07:55:03 +0000

After many rumors about Nvidia’s dual GPU solution, now that the flagship GTX 680 has been launched, nothing concrete has come from Nvidia. That is until last night when the GTX 690 was officially announced.

As you would expect, the GTX 690 carries two GF110 GPUs, so you get double the benefits of a GTX 680. 3072 CUDA cores, 64 ROPs, 256-bit bus on each GPU and 4GB of RAM in total. Gone is the NF200 bridge used in the GTX 590, instead we have PLX’s PEX 874x switch that is PCIe 3.0 enabled, with 48 lanes.

To accommodate for two GPUs on one PCB some speed drops have been made to reduce heat, but thankfully not by much. Core clock speed is 915MHz (boosting to 1019MHz) compared to the GTX 680′s 1006MHz (Boosting up to 1058MHz).

Power comes in the form of two 8-pin connectors, with the GTX 690 rated at 300W TDP, compared to 375W TDP on the GTX 590. But not only does the GTX 690 dissipate less heat, but also runs cooler and quieter, 47dB compared to 51dB on the GTX 590.

The new heatsink on the GTX 690 is equally impressive, in that the aluminum fins on top of the GPUs are attached to chrome-plated aluminum cover, as opposed to the plastic cover on pretty much all of their other cards. Even the fan housing is magnesium alloy, which reduces noise and dampens vibrations. Nvidia will also have an LED which can be controlled by partners to indicate different things, changing colors or level of glow depending on the usage for instance.

All said, the number of monitors connected to the GTX 690 is limited to four, three via DVI and one mini-DisplayPort. The asking price for this impressively engineered product is $1000, which works out about the same as two GTX 680s. Availability will begin on 3rd of May, with more shipments coming in on 7th May. However, given the limited availability of the GTX 680s, don’t expect the GTX 690s to linger in stock for more than a few hours.

We’ll have full review coverage soon.

Nvidia regbadges old Fermi cards to series 600

Taimoor Hafeez — Thu, 12 Apr 2012 06:19:20 +0000

Rebadging previous generation cards to newer ones is not a new practice, as both Nvidia and AMD have done this in the past. A popular example by Nvidia is with their mid-range 9800GTX+ rebranded to GTS 250. Most recently AMD has done the same with their high-end mobile HD 6770M to HD 7690M XT, although rebranding in mobile GPUs is commonplace.

According to Fudzilla, Nvidia will be rebadging some of their old Fermi based entry-level cards to series 600; made popular with the new Kepler architecture of the GTX 680. Retail GT 610 will be a rebranded GT 520, the GT 620 will be a GT 430, and the GT 630 will be a GT440. The idea is that these cards will come with a new BIOS, faster clock speeds and, in the case of the GT 630, come with GDDR5 memory.

The GT 440; soon to be GT 630.

It seems that TSMC fabrication plant cannot keep up with 28nm production, and with the impending release of Intel’s Ivy Bridge and AMD’s A10 Trinity processors, the entry-level market for graphics will be extremely competitive. Under such circumstances it comes as no surprise that Nvidia is rebranding previous generation hardware to keep pace in the market.

ASUS HD 7850 Direct CUII Review

Taimoor Hafeez — Wed, 11 Apr 2012 08:33:01 +0000

The ASUS HD 7800 series launched early last month, and the HD 7870 left us quiet impressed. Now it’s time to see if the little brother in the series has enough chops for big league. And so today I’ll be looking at the ASUS HD 7850 Direct CUII graphics card which comes with a bigger heatsink and higher clock speeds than reference AMD design.

The Direct CUII heatsink is a well known name amongst PC enthusiasts as it represents ASUS’s best and biggest heatsink for their graphics card. PCB designs are also custom; usually smaller than reference designs, so that the heatsink has the maximum possible effect on the entire card.

Now usually Direct CUII heatsinks are found on flagship cards, such as the HD 7900 series, where the heatsinks are so big that we usually end up with a 3-slot card instead of the usual 2-slots. In the case of the HD 7850, however, we have a simple 2-slot solution. The 28nm Pitcairn chipset runs fairly cool, so it stands to reason that we don’t need an extraordinarily large heatsink.

Packaging

The ASUS HD 7850 Direct CUII comes with installation manuals, a drivers CD, a CrossFire bridge and a DVI to VGA adapter. Given that this is a mid-range card, the lack of extras bundled with it comes as no surprise. With a minimum PSU requirement of just 450W and a single 6-pin power connector, the HD 7850 is good enough to fit into HTPCs as well.

Heatsink

Now the card itself is very interesting to behold. The basic design of the Direct CUII heatsink on the HD 7850 looks to be identical to higher-end cards such as the recently reviewed HD 7970 Direct CUII TOP. Closer inspection will reveal major differences, though. The GPU and memory modules are covered with an aluminum block extending out to multiple thick fins. The GPU itself is connected to three copper heatpipes that then extend into a secondary aluminum block with thinner, and more, fins. The whole unit is then cooled off by two 7.2cm fans. The entire heatsink is covered in a plastic casing right on the top, unlike the metal one on HD 7970 DCUII TOP.

Overclocking

So what does this Direct CU II heatsink give us? Well, firstly the card comes factory overclocked to 975MHz core clock and 1250MHz (5000MHz effective) memory clock speeds. This itself is 115MHz and 50MHz faster than reference AMD core and memory clock speeds. When overclocking, the Direct CUII heatsink also afforded a respective core clock of 1031MHz and memory speeds of 1300MHz (5200MHz effective). That’s clocked 5.7% faster than factory and 19.8% faster than reference AMD speeds. Memory speeds are 4% faster than factory and 8.3% faster than reference speeds.

Temperatures remained at 30°C when idling, going up to 69°C under full load. While overclocked, idle temperatures remained the same, but maximum temperature touched 70°C. For the most part eh fans remained fairly quiet, but running Metro 2033, Battlefield 3, Unigine and 3DMark made the fans loud. Not loud enough to be annoying, but on our open testbed then were audible in the whole room. Inside a rig, of course, even that much sound will be minimal.

Test setup

The below testbed was used for testing the ASUS HD 7850 Direct CUII.

The below settings were used in each of the benchmark.

The tests were carried out at a resolution of 1920×1080 and drivers used for the ASUS HD 7850 Direct CUII were the latest Catalyst 12.3 release.

Nvidia GTX 680 SLI vs AMD HD 7970 CFX

Taimoor Hafeez — Thu, 29 Mar 2012 11:44:11 +0000

Now that the flagship single-GPU cards from both Nvidia and AMD are out, it’s time to see how well they perform in a dual-GPU setup. We know that the Nvidia GTX 680 beats the AMD HD 7970 without breaking a sweat; but history has taught us that SLI and CrossFireX don’t always have the same scalability one would hope.

So today I’ll be looking at the capability of a pair of GTX 680 and HD 7970, and how optimized both Nvidia and AMD drivers really are to take advantage of such heavy hardware.

Below is a spec comparison sheet between the GTX 680 and HD 7970 cards.

As you can see, on paper the HD 7970 seems superior to the GTX 680, in that it has a higher transistor count, more memory and shaders. Despite that, a standard GTX 680 will beat any HD 7970 in pretty much all our benchmarks. Increase the number of cards from one to two, though, and the picture changes colors.

Test Setup

For testing the graphics cards today, the following testbed was used:

The drivers used for the AMD HD 7970 cards were Catalyst 12.2 and for the Nvidia GTX 680 cards were GeForce 301.10.

The below settings were used in all of our benchmarks, run at resolutions of both 1920×1200 and 2560×1440.

Gigabyte HD 7770 OC Review

Taimoor Hafeez — Tue, 27 Mar 2012 06:28:59 +0000

In all the hoopla over the last few weeks with the Kepler launch, I overlooked a rather interesting card from AMD that also launched recently. While not quite in the same category as the other mid-range to high-end cards I have recently reviewed, the HD 7700 series nonetheless deserves a look over. True, I wasn’t much impressed with the AMD HD 7750, but the HD 7770 presumably improves the situation somewhat. So with much hopes, today I’ll be looking at the Gigabyte HD 7770 graphics card.

Layout & Cooling

The Gigabyte HD 7770 OC, as the name implies, comes with a factory overclock. What we have here is 1100MHz on the Core clock, up from 1000MHz from reference AMD speeds.

The overclock is obviously a result of the custom heatsink Gigabyte has put on top of the Cape Verde GPU. The large metal (presumably aluminum) heatsink has multiple fins spreading out; kept cool with a large fan that’s mostly quiet.

Measuring a total of 21cm, this card requires one 6-pin power connector with a minimum 450W PSU for smooth operations. Inside the packaging you’ll just get the (outdated) driver CD and installation instruction manual.

Nvidia CEO hints at Kepler coming to ‘superphones’

Taimoor Hafeez — Sun, 25 Mar 2012 08:21:06 +0000

With the recently successful launch of Nvidia’s flagship GTX 680 graphics card, the company’s CEO sent out an email congratulating his employees on their hard work developing the Kepler architecture. Furthermore, Nvidia’s CEO Jen-Hsun also talked a bit about where the Kepler architecture will be headed in the future.

The whole email can be seen below, as reported by AnandTech.

Notice the bolded part in the email where Jen-Hsun talk’s about the Kepler architecture coming to superphones. While not exactly a confirmation, nor an idea of their roadmap, the future of our phones certainly seems bright.

—–Original Message—–

From: Jensen H Huang

Sent: Thursday, March 22, 2012 9:48 AM

To: Employees

Subject: Kepler Rising

Today, the first Kepler – GTX 680 – is on shelves around the world!

Three years in the making. The endeavor of a thousand of the world’s best engineers. One vision – build a revolutionary GPU and make a giant leap in efficient-performance.

Achieving efficient-performance, great performance while consuming the least possible energy, required us to change our entire design approach. Close collaboration between architecture-design-VLSI-software-devtech-systems, intense scrutiny on where energy is spent, and inventions at every level were necessary. The results are fantastic as you will see in the reviews.

Kepler also cultivated a passion for craftsmanship – nothing wasted, everything put together with care – with a goal of creating an exquisite product that works wonderfully. Let’s continue to raise the bar and establish extraordinary craftsmanship as a hallmark of our company.

Today is just the beginning of Kepler. Because of its super energy-efficient architecture, we will extend GPUs into datacenters, to super thin notebooks, to superphones. Not to mention bring joy and delight to millions of gamers around the world.

I want to thank all that gave your heart and soul to create Kepler. You’ve created something wonderful.

Congratulations everyone!

Jensen