Notes:
This is a simple/quick translate for shiver76 (#4 as of today in 3dmark11 hall of fame), as he's too busy and I post it on behalf of him. After you guys see LN2 sessions here every day, time for a change! I believe it reflects the limit of air cooling, and please don't be harsh if it cannot catch up with your LN2 scores!
Prologue:
Though I've received several MSI GTX580 Lightning, I didn't have much spare time to play with them. Finally got a chance today. This test includes the basic overclocability of MSI GTX580 Lightning, temp/noise comparison against the reference 580s, the ultimate aircooling oc of 4-way SLI, and some criticisms on the GTX580 Lightnings. Miss it, your loss
Test Setup:
Mobo: ASUS P6T7 WS
CPU: i7 980X
RAM: Corsair Dominator GT 2G x 3 2000 7-7-7-20 1T
VGA: MSI GTX580 Lightning x 4, Reference GTX580 x 4 (flashed with ASUS BIOS)
Driver: 270.51 & 270.61
HDD: Crucial C300 256G
OS: Win7 x64 Ultimate
PSU: Silverstone ST1500W + TT Power Express 650W
Ambient temp: 20C
First of all, group photo of old members and new members in the family. Can't say the different brands of the reference cards please my eyes..
Now the test results of each single Lightning. Since it's the "Lightning", the most important point is about overclockability! However all the 5 Lightnings I've played were only giving slightly better results than the reference cards under aircooling conditions, and this is a bit disappointed compared with what the GTX480 Lightning did in the old times... In the order of time I opened the boxes:
Card A: default 1.013v; OC 1.213v 930MHz (WTF?!)
Card B: default 1.000v; OC 1.100v 970MHz
Card C: default 1.013v; OC 1.113v 1000MHz
Card D: default 1.025v; OC 1.175v 1000MHz
Card E: default 1.013v; OC 1.113v 970MHz
Though I could further bump the voltage in many tests, the gain was rather neglegible, probably due to the limitation of aircooling. I've tried to play with various switches on the back of the Lightning but none of them helped with my OC under room temperature. Maybe it'll make a big difference during LN2 sessions?
The following is a test of a relatively good sample among these Lightnings. Core +100mv for 1GHz:
Now some simple comparison between the reference and the Lightning:
(This section is chopped by me)
Time for benchmarking! Is this the world-first of 4-way GTX580 Lightning?
Core/Mem/Aux = 137mv/100mv/20mv, fan 100%, clock 970/1940/1200 (not every sample could achieve 1GHz and this is the current limit I can achieve for 4-way). Breaking X9000! When I submitted the results I was #7 (and I assume all the top 6 were LN2s and I'm the top aircooling?)
In order to comply with ORB and enter hall of fame:
275.27 beta:
Now time for Crysis @ 960/1920/1200, map = avalanche. Breaking 100fps average is so simple! Why not Crysis 2? Sorry neither I'm interested nor these cards...
Now the criticism. MSI please answer my questions: as a top-end product, why the hell is it so unfriendly towards multi-way SLI? Look,
1. The PCB is not raised the same time as the cooler, which makes it impossible to use the 3-way/4-way SLI hard bridge. It is nasty to use the soft bridges and such bending is not good for longevity!
2. As a product meant for 3-way/4-way SLI, and advertised as a two-slot card, why the f* would the fan blade take some extra space making it a 2.1-slot card? Since it's already g* distance between cards, the fan blade can easily stab the back of the next card! I had to insert some soft stuff between them. This is not only far away from elegancy but also stupidity!
Conclusion:
1. MSI GTX580 Lightning is overall a nice product, with overclockability slightly better than the reference cards (but also with a chance to get hit by bad ones), good-looking and making. The excellent cooling performance with only a 2-slot (2.1-slot) design makes those 3-slot cards ashamed.
2. A blemish in an otherwise perfect thing: not friendly towards multi-way SLI. Hope MSI improves it in the future.
Bookmarks