12" HP 2570P GTX760@4Gbps+c-EC2 (PE4L 2.1b) + Win7 [bjorm]

[bjorm]

Because of that GTX680 does not give any tangible advantages over the GTX670 I returned to the idea of ​​a mini eGPU . Generally carrying my trunk is ok but why not try to reduce it ?

Unfortunately GTX670 mini is almost unavailable in Poland , I started looking for GTX760 mini. According to test of these cards on the same portal, it is clear that the GTX760 performs better in games than the GTX670. Probably because of newer driver.

At the beginning I found the power supply, small SFX from Tacens Radix Eco 350W appeared as the ideal, but immediately after turning it on, I decided to find something else, to me is just too loud in idle. I came to SFX from Fortron, led satisfaction with currently used Fortron 350W big brother, I bought a SFX Fortron in the hope that this will more quiet than Tacens. And it is. Much more quiet. Tacens has a very strange whistling at idle and do not want to mix with the bearings.

When got SFX and card, I started to connecting it. Then zonk. GTX760 mini has 8 input pin adapter but I thought this adapter 8pin - 2x molex. No no no - we have 8pin from 2x6pin. Well, shit happens. Fortron as SFX has a lot molex as much as 3 but it is unfortunately one lacking Oh well , I'll try to connect like an idiot, 2xmolex - 6pin - 8pin as one branch and a second branch created from one (supposed to have two) molex.

After connecting everything, it turned out that the card fired! No howling , does not go out , etc. have been so tests in games and 3dm11 . For stock went nicely 3dm11 giving result 7104 GS . After trying to OC for the same level as on test card on TechPowerUp, failed to pass the entire 3dm11, the highest setting at which the test was over 100 core 200 memory. This gave a result of 7488. Results weak compared to the i7-3630QM + GTX680 or i7-3630QM + GTX670 but as you can see, once again, the points do not play particularly in the realities of the eGPU. In the end, it was able to draw a card with a score of GS most comparable to PS

After a few runs 3dm11, mode high efficiency and throttlestop set to x29 (the maximum possible stable multiplier for this CPU), I had a max temp 79C on the card. Alarmingly much. 80C is a throttling limit for a series of GTX7 ** . Few minutes after the test card hanging himself in idle at 33C (at 1400 rpm [ btw 1400 rpm max is observed for GTX670/760/680 series TwinFrozr]). The stress max is 2340 rpm. Not bad.

Below are some pictures (unfortunately poor quality) cards with Fortron SFX. It is a comparison of the card + SFX to the ATX and SFX card comparison to trunk which is fitted almost ideal normal ATX power + GPU. Ultimately, I want to make the case for a mini eGPU which is to have the external dimensions:

• height 12.5cm
  
• length 18cm
  
• max width of 10cm.
  

So we can say that I have an eGPU in size of ATX AC!

GTX760 mini + Fortron SFX

GTX760 mini + Fortron SFX vs Fortron ATX

[bjorm]

If You don't have already GTX660Ti so better buy GTX670.

here You have comparison for BF4, multi, 64x64, paracel, medium + FHD. today gameplay for GTX760 only for ten minutes, tomorrow I'll try to run it for 12-14 minutes, then we will se if differences will be bigger or smaller. For now, we have mean differences around ... 8% between OC versions GTX760 and GTX680. In my opinion it is not worth of paying extra cash for GTX680.

red: GTX760 OC mini

green: GTX670 OC MSI TF

blue: GTX680 OC MSI TF

Another picture present differences in temperature and TDP between i7-3630 and i7-3632. In my opinion is really worth to get CPU 35W TDP instead of 45W TDP (or downcloch it to the level 35W TDP, as You like ). We can also detect that there is the same CPU and GPU utilization for both CPU's.

[Tech Inferno Fan]

Unfortunately GTX670 mini is almost unavailable in Poland , I started looking for GTX760 mini. According to test of these cards on the same portal, it is clear that the GTX760 performs better in games than the GTX670. Probably because of newer driver.

ASUS GTX 760 DirectCU Mini 2 GB Review | techPowerUp shows a stock-clocked GTX670 being 7% faster than a stock-clocked GTX760 based on their cumulative benchmarks and game testing. Something I'd expect since the GTX670 has 16.7% more (1344 vs 1152) CUDA cores though runs lower clocks. Even your graphs show the GTX670 being faster than the GTX760. Run a GTX670 at the GTX760's higher stock or OC clocks and the performance gap widens further.

I was lucky to find a ASUS GTX670 mini. It's about 60% the size of my HD7970Ghz edition with a max 225W input power (vs 300W for the HD7970!!).

Another picture present differences in temperature and TDP between i7-3630 and i7-3632. In my opinion is really worth to get CPU 35W TDP instead of 45W TDP (or downcloch it to the level 35W TDP, as You like ). We can also detect that there is the same CPU and GPU utilization for both CPU's.

If you put your 35W CPU under extreme load (eg: linx) then you'll see Intel's turbo algorithm TDP throttle the 4-core turbo . It might go down to x27 or even less. For that reason I'd recommend getting a 45W TDP CPU to give more TDP headroom to play with. You have the flexibility of being able to use Throttlestop to limit the highest multiplier on a 45W CPU to bring temps down to 35W TDP CPUs if needed.

[bjorm]

If you put your 35W CPU under extreme load (eg: linx) then you'll see Intel's turbo algorithm TDP throttle the 4-core turbo . It might go down to x27 or even less. For that reason I'd recommend getting a 45W TDP CPU to give more TDP headroom to play with. You have the flexibility of being able to use Throttlestop to limit the highest multiplier on a 45W CPU to bring temps down to 35W TDP CPUs if needed.

But take a look that even BF4 as a game focused on CPU, uses CPU almost barely (so I don't need a 45W TDP efficiency). I'll check how it works after cores unlock/parking.