xentr_theme_editor

  • Please do not post any links until you have 3 posts as they will automatically be rejected to prevent SPAM. Many words are also blocked due to being used in SPAM Messages. Thanks!

MSI GTX 580 Lightning Xtreme Edition (XE) 3GB Review

Status
Not open for further replies.
xentr_thread_starter
Just Cause 2 (DX10)

Just Cause 2 (DX10)


Just Cause 2 has quickly become known as one of the best-looking games on the market and while it doesn’t include DX11 support, it uses the full stable of DX10 features to deliver a truly awe-inspiring visual experience. For this benchmark we used the car chase scene directly following the Casino Assault level. This scene includes perfectly scripted events, some of the most GPU-strenuous effects and lasts a little less than four minutes. We chose to not use the in-game benchmarking tool due to its inaccuracy when it comes to depicting actual gameplay performance.


1920 x 1200

GTX-580-LIGHTNING-XE-47.jpg


GTX-580-LIGHTNING-XE-48.jpg


2560 x 1600

GTX-580-LIGHTNING-XE-49.jpg


GTX-580-LIGHTNING-XE-50.jpg
 
xentr_thread_starter
Lost Planet 2 (DX11)

Lost Planet 2 (DX11)


Lost Planet is a game that was originally released on consoles but in its port over to the PC, it gained some highly impressive DX11 features. For this benchmark, we forgo the two built-in tools and instead use a 2 minute gameplay sequence from the second level in the first chapter. The reason we use this level is because it makes use of three elements that are seen throughout the game world: jungles, water and open terrain.


1920 x 1200

GTX-580-LIGHTNING-XE-52.jpg


GTX-580-LIGHTNING-XE-53.jpg


2560 x 1600

GTX-580-LIGHTNING-XE-54.jpg


GTX-580-LIGHTNING-XE-55.jpg
 
xentr_thread_starter
Metro 2033 (DX11)

Metro 2033 (DX11)


There has been a lot of buzz about Metro 2033 which has mostly centered on its amazing graphics coupled with absolutely brutal framerates on even the best GPUs on the market. For this test we use a walkthrough and combat scene from The Bridge level which starts at the beginning of the level and lasts for about 3 minutes of walking, running and combat. Famerates are measured with FRAPS and Advanced PhysX is turned off.


1920 x 1200

GTX-580-LIGHTNING-XE-56.jpg


2560 x 1600

GTX-580-LIGHTNING-XE-57.jpg
 
xentr_thread_starter
Unigine: Heaven v2.1 (DX11)

Unigine: Heaven v2.1 (DX11)


Unigine’s Heaven benchmark is currently the de-facto standard when it comes to simple, straightforward DX11 performance estimates. While it is considered a synthetic benchmark by many, it is important to remember that no less than four games based on this engine will be released within the next year or so. In this test we will be using a standard benchmark run with and without tessellation enabled at three resolutions,


1920 x 1200

GTX-580-LIGHTNING-XE-58.jpg


GTX-580-LIGHTNING-XE-59.jpg


2560 x 1600

GTX-580-LIGHTNING-XE-60.jpg


GTX-580-LIGHTNING-XE-61.jpg
 
xentr_thread_starter
8x MSAA Testing

8x MSAA Testing


In this section we take a number of games we have tested previously in this review and bring things to the next level by pushing the in-game MSAA up to 8x. All other methodologies remain the same.

BattleField: Bad Company 2 (DX11)
Note that 8x MSAA is enabled via the game’s config file for the NVIDIA cards since it is not a selectable option within the game menu

GTX-580-LIGHTNING-XE-40.jpg


Dirt 2 (DX11)

GTX-580-LIGHTNING-XE-46.jpg


F1 2010 (DX11)

GTX-580-LIGHTNING-XE-69.jpg


Just Cause 2 (DX10)

GTX-580-LIGHTNING-XE-51.jpg
 
xentr_thread_starter
Giving 3GB a Workout / Performance Impressions

Giving 3GB a Workout


In this section we will do some simple tests: use two notorious GPU memory hogs -Shogun 2: Total War and Metro 2033- to see how far we have to push things in order to reach the memory limit of a standard MSI Lightning (at its factory overclocked levels). We’ll then compare and contrast those results to those we encountered while using the 3GB Lightning XE.

For Metro 2033 we used our standard gameplay playback.

We decided to bypass Shogun 2’s built in benchmarking utility and instead use a playback from a custom battle. This effectively removes any CPU bottlenecks from the equation since AI is disabled in playback mode. Additionally, the config file was modified to allow the 1.5GB card to read as having 3GB of texture memory available which is necessary as any other option would limit the available video card settings.

GTX-580-LIGHTNING-XE-70.jpg

First up we have Shogun 2, an impressive looking game when DX11 is enabled but it is also extremely tough on hardware. As a purely anecdotal point, we read a whopping 2.1GB of texture memory being used at 2560 x 1600 with 8x MSAA enabled.

It seems like even with such a demanding game, the core actually becomes the bottleneck here rather than the available memory bandwidth. The 3GB card can easily win when the settings are cranked but it doesn’t make a difference since playable framerates were impossible to achieve with MSAA enabled at 2560 x 1600 regardless of the card being used. At 1920 x 1200 up to 4x MSAA things are a bit easier but since available video memory wasn’t an issue, both cards drew even.

GTX-580-LIGHTNING-XE-71.jpg

Metro 2033 shows much of the same thing as Shogun: the 3GB framebuffer does make a difference but once again the GPU itself becomes the limiting factor when asked to push around massive rendering workloads. On the positive side of things, the 3GB card was actually able to play Metro 2033 with everything maxed out….albeit at a paltry 17 frames per second. Minimum framerates on the other hand were absolutely dominated by the Xtreme Edition.


Performance Impressions


As you may have already observed Lightning Xtreme Edition’s performance in all of our tests (including those above) was interesting to say the least. The majority of our tests obviously just didn’t push the 1.5GB Lightning’s memory bandwidth past its breaking point. However, in other cases we saw that at high resolutions with extreme instances of MSAA, additional bandwidth can at the very least increase minimum framerates.

First up we have something that shouldn’t be too much of a shocker. On single monitor systems the current crop of DX11 games just isn’t bandwidth limited on a 1.5GB 384-bit card like the GTX 580. You’re much more likely to run smack into a rendering limitation rather than a situation which demands excess memory. All the video memory in the world won’t save your card’s bacon if its core struggles to process draw calls. Not once did the expanded GDDR5 layout provide playable framerates in situations where the 1.5GB version struggled.

This isn’t to say that 3GB GTX 580 cards are pointless in today’s market. It is quite obvious that the second generation DX11 games like Metro 2033 and Shogun 2 have a voracious appetite for memory resources and upcoming titles will likely do little to change this trend. So it is conceivable that within Lightning XE’s expected lifecycle there will be a few titles which will benefit from its expanded memory allotment…but only when playing at ultra high resolutions with anti aliasing pushed to mind boggling levels.

We saw throughout the standard battery of tests that memory controllers have additional overhead due to the increased memory allotment per controller. Memory latency has also increased through the use of 64x32 modules rather than the reference version’s 32x32 module layout. The result is a negligible framerate decrease when compared to 1.5GB card. It is for this reason that we highly recommend choosing wisely when looking at GXT 580 cards; if you’re using a 24” or smaller monitor, better performance is attainable by less expensive 1.5GB cards.
 
xentr_thread_starter
Temperature & Acoustics / Power Consumption

Temperature Analysis


For all temperature testing, the cards were placed on an open test bench with a single 120mm 1200RPM fan placed ~8” away from the heatsink. The ambient temperature was kept at a constant 22°C (+/- 0.5°C). If the ambient temperatures rose above 23°C at any time throughout the test, all benchmarking was stopped. For this test we use the 3DMark Batch Size test at its highest triangle count with 4xAA and 16xAF enabled and looped it for one hour to determine the peak load temperature as measured by GPU-Z.

For Idle tests, we let the system idle at the Windows 7 desktop for 15 minutes and recorded the peak temperature.


GTX-580-LIGHTNING-XE-63.jpg

The extra 1.5GB of GDDR5 does seem to push heat production up by a good amount but that alone doesn’t account for the higher temperatures we saw with the Lightning XE. We’re guessing our previous Lightning sample featured a lower leakage core which is why we are seeing this slight shift. Nonetheless, the temperatures are still very, very good and speak volumes about the Twin Frozr III heatsink design.


Acoustical Testing


Yes, we have finally added decibel testing to our repertoire and this section will expand in future reviews. What you see below are the baseline idle dB(A) results attained for a relatively quiet open-case system (specs are in the Methodology section) sans GPU along with the attained results for each individual card in idle and load scenarios. The meter we use has been calibrated and is placed at seated ear-level exactly 12” away from the GPU’s fan. For the load scenarios, a loop of Unigine Heave 2.5 is used in order to generate a constant load on the GPU(s) over the course of 20 minutes.

GTX-580-LIGHTNING-XE-41.jpg

The increased temperatures also cause the heatsink’s fans to spin up to higher RPMs than the Lightning XE’s predecessor. It is still quieter than the vast majority of high end cards on the market and can barely be head over the noise most case fans make.


System Power Consumption


For this test we hooked up our power supply to a UPM power meter that will log the power consumption of the whole system twice every second. In order to stress the GPU as much as possible we once again use the Batch Render test in 3DMark06 and let it run for 30 minutes to determine the peak power consumption while letting the card sit at a stable Windows desktop for 30 minutes to determine the peak idle power consumption. We have also included several other tests as well.

Please note that after extensive testing, we have found that simply plugging in a power meter to a wall outlet or UPS will NOT give you accurate power consumption numbers due to slight changes in the input voltage. Thus we use a Tripp-Lite 1800W line conditioner between the 120V outlet and the power meter.

GTX-580-LIGHTNING-XE-62.jpg

With higher clock speeds and a full 3GB of overclocked memory, it’s no wonder that the Lightning XE consumes more power than past cards. With that being said, when working with an ultra high end graphics card power consumption miracles just can’t be expected.
 
xentr_thread_starter
Overclocking Results

Overclocking Results


Overclocking on this card was a joy due to MSI’s AfterBurner software and we were actually able to increase core clocks past the 1Ghz mark. This was actually stable for normal gaming which is surprising since GTX 580 cards rarely hit this point on air cooling while retaining stability. Memory overclocking was a bit tougher which is likely due to the 256MB ICs.

All results were achieved with voltages of 1.23V on the GPU and an additional 40mV on the GDDR5 memory. Naturally, these increases led to a spike in heat production so the fans were increased to 75% in order to compensate.

Core Clock: 1009Mhz
Engine Clock: 2018Mhz
Memory Clock: 4380Mhz (QDR)


GTX-580-LIGHTNING-XE-72.jpg


GTX-580-LIGHTNING-XE-73.jpg
 
xentr_thread_starter
Conclusion

Conclusion


With the GTX 580 maturing like a fine wine, it has been interesting to see what NVIDIA’s board partners had in store for the market. We’ve already looked at Gigbayte’s Super Overclock and MSI’s Lightning and there even more custom cards on the horizon from companies like ASUS and EVGA. With this long list of ultra expensive cards making a splash, one would think that yet another overclocked GTX 580 would fail to make people sit up and take notice. However, the Lightning Xtreme Edition has caused a considerable amount of buzz within enthusiast circles ad with good reason too.

The XE is an extremely ambitious card since it combines the high speeds seen in the previous MSI Lightning with a massive amount of memory. Granted, this combination does come with an ultra high price of some six hundred bucks (or possibly more!) but that’s a price many will likely be willing to pay for what will likely be the highest performing GTX 580 available in North America.

From framerates to thermal numbers to its extremely quiet operation, the Lightning XE is a must have card for enthusiasts with some extra coin burning a hole in their wallets. However, the additional 1.5GB of GDDR5 doesn’t come without some sacrifices. MSI’s flagship card does tend to suffer a bit from the increase latency that comes with higher memory densities. The performance drop-off this caused against the standard Lightning wasn’t noticeable within games but it was notable nonetheless since it showed up in nearly every game. On flip side of this coin, the extra memory did allow the Lightning XE to post a few impressive gains in high resolution, high AA scenarios, particularly in the minimum framerate department.

Another issue we saw was a repeat of our past experiences with cards sporting non-reference memory allotments. Even in the case of the GTX 580, it seems like the underlying GF110 architecture runs out of gas far before the memory bandwidth limits are reached. In both Metro 2033 and Shogun 2 a 3GB framebuffer did allow for the game to run at ultra high detail settings but the experience was still a slideshow since the overclocked core was pushed well past its rendering capabilities. Granted, multi monitor resolutions would have likely seen the 3GB Lightning XE pulling off some impressive wins but we weren’t given a pair to play around with. Plus, spending $1200 or more on two of these cards is more than all but a few upper-crust enthusiasts are willing to spend.

In our opinion, MSI has a winner on its hands that likely won’t be eclipsed for some time. 3GB of memory may be dubious in its usefulness in some scenarios but with DX11 games demanding more and more memory resources, we’d highly recommend this type of layout for upcoming games. If ultra high resolution gaming isn’t your thing, the “vanilla” Lightning is out there but Xtreme Edition still is one hell of a card that’s well worth its asking price…but only to those who can absolutely afford it.


 
Status
Not open for further replies.

Latest posts

Back
Top