Once again Chocolate Box returns, this time for a massive upgrade! Even at this early stage it's obvious what will be going down and i'm very very excited about it. After the previous platform upgrade to Z170 there was certainly no shortage of CPU speed or efficiency and now it's time to apply the same ethos over to the GPU side of things with just a hint more brutish ignorance.
Momentarily I had planned to swap in a 980Ti K|NGP|N which would have been a significant undertaking of swapping PSU's, many cables and lots of little tweaks to make it all fit. But then while contemplating how to orchestrate that elaborate transplant a Titan X Pascal was bestowed upon me. So now I must go forth and do what I have to do!
The super vigilant among you, or just everyone, may have noticed that was slightly too many TXP's to fit on an ITX board. Well you were correct, and no I won't be using 4 this time, sorry. However I will be using the best one of this set, because 'a' TXP clearly wouldn't have been good enough for a build with one of the best 61280K's and DDR4 kits in the wild.
So to determine a clear victor and a worthy companion for the current hardware a huge amount of overclocking and benchmarking took place. Here are some of the highlights...
Now you may have expected that one card in the pack shone through, clocked higher, scored like no other and those 5 screenshots were all from one of the finest examples of a TXP to ever exist. Well I must admit, I did too, but that wasn't the case at all.
Card - Core - Mem - FS eff - FS - Heaven - Valley
TXP1 - 220 - 600 -
31877 - 32074 -
4145 - 6679
TXP2 - 220 -
700 -
31836 -
32214 -
4170 - 6707
TXP3 - 220 -
700 - 31791 - 31621 - 4117 -
6809
TXP4 -
250 -
700 - 31739 -
32279 - 4124 -
6838
On all four cards I tested the highest achievable core and memory offsets with the stock cooler, once I established they could all easily achieve +200mhz core and +500mhz memory I ran Firestrike Standard graphics with the same settings on each card to determine the relative efficiency. The same thing as an ASIC value attempts to represent, the deviation was quite significant and it gave me a great insight into what I could expect from them without investing hours and hours into watching Firestrike. The two most efficient cards were able to achieve the highest Unigine Heaven scores while the higher leakage cards fared better at Valley. Firestrike made things very interesting as it took advantage of the very high clocks TXP4 was able to achieve while at the same time the more balanced blend of efficiency and clocks TXP2 offered also did well, unlike earlier in Valley.
With all that information picking the optimal card wasn't a simple decision as they all performed reasonably well under different conditions. In the end it came down in favor of TXP2 as it had great efficiency and was still able to achieve relatively high clocks, visible in its strong performance across all 3 benchmarks. The shear speed and leakage of TXP4 were very alluring and i'd be excited to see what it could do at lower temperatures, however with only a single 120mm radiator efficiency seemed like a more suitable asset. Plus I did a little SLI testing and TXP2 was by far the most cooperative as a primary card.
So enough of the boring reading, many sexy pictures!
All the credentials are now ready for action, I just need a little watercooling gear to suit the new card, few cables, few tubes and I will be ready to hit LAN's with some next level Farm Simulator settings. No seriously, only a few weeks until i60 and there is lots of farming to be done!!!
JR