AliExpress!

I think all ASUS DirectCU II (DC2) GPU cards used the same fans, they might be interchangeable. Though I never owned any (let alone many) of them, so I could be wrong. Any part numbers or markings anywhere on the fan?

You can try your local craigslist. People always have dead/dying GPU cards they sell cheap (or just give away) for parts. Your local computer repair shop might have dead/dying cards with these fans, too.

If you live near a hackerspace or a plastics shop, you might be able to buy a 3D printed part at material cost.

Google shows that your problem is somewhat common. Some people glue/tape/splint their fan blades back together with mixed results, usually more wobbly and noisy than before plus it's just a matter of time before the repaired blades break off again. But you might be able to repair and balance it well enough, and it might hold indefinitely if used in your secondary GPU card.
I would repair the break with a strong superglue (any cyanoacrylate plus an accelerator/kicker). Then carefully cover and cut a "strong" or "thick" tape - metallic foil tape, aramid tape, carbon fiber tape, old vinyl decals, whatever - fully across both sides (this might need to be applied to opposite/other blades to keep the fan balanced and reduce bearing wear). The cross section of the break is very thin, putting strain on the glue/tape, so I'd reinforce it with a "splint" (made from metal wire or from any slender flexible toothpick-like bit of plastic) along each of the blade edges (and repeat the process on the opposite/other blades as needed to keep the fan balanced).

A perfect opportunity for a liquid-cooling upgrade ...

i already ordered the fan from ebay so all i have to do is order some thermal paste for the gpu when i take the card apart to plug the wires in.
whats a good thermal paste for a gpu?

Any "premium" TIM will outperform any "standard" TIM (including that infamous cheap pink metal oxide goop ASUS always uses for everything on every part).

I generally prefer trusty old Arctic Silver 5 (which I can buy any time from any local computer store) and Prolimatech PK-3 (because I still have a huge jar of it).
We use a lot of an unlabelled proprietary Shin-Etsu TIM (which I think is X23-7762) and sintered HOPG pads where I work - we engineer and manufacture custom mission-critical medical/hospital equipment, and we must comply with the very strictest regulatory/legal guidelines.
Most avid overclockers ("pro" or "amateur") seem to prefer Gelid GC-Extreme or IC Diamond. Gelid is firmly established as the "best" TIM for exotic/subzero cooling applications.

I think minor variations in specific thermal conductivity between the top-performing TIMs is essentially insignificant, an efficiency difference of 99.996% vs 99.994% basically translates into a difference of less than 0.1C or 0.2C in real measured temps, and it falls well within measurement error caused by numerous noncontrolled variables. And that's ignoring the most wildly random variable of all: the specific quirks and characteristics of any given part playing the overclocked "silicon lottery" game.

What matters more is how well the TIM fills the void between mating surfaces. A thin and runny TIM (like Gelid GC-Extreme or Prolimatech PK-3) is able to flow into and fill tiny gaps, it is a better choice for tight part tolerances and awkward part geometries. A thick and goopy TIM (like Arctic Silver 5) is typically more efficient at bulk thermal transfer, and it's a better choice for coverage of larger part surfaces. No single TIM can possibly be "perfect" for every possible application, regardless what champions of the brand will claim.

What matters most is how much TIM is used. Too many people (including ASUS, lol) seem to use far too much. But less is more, the idea is simply to fill the void and ensure full thermal contact between surfaces, it's not to stuff the gap until it leaks or pump it up with TIM (which has much better heat transfer than air but much worse than metal, lol). I generally prefer scraping the TIM on with a razor blade to attain consistent coverage across the entire part surface, but other application methods (so-called "X", "Dot", "Pea", "Rice Grain", and others) are all popular and effective on CPU/GPU parts. I also like to immerse the entire syringe/package into warm/hot water for a few minutes before application when working with a thick TIM (like Arctic Silver 5) - it'll flow better during application but adhere stronger as it cools down, plus it stays "wet" long enough to clamp on the heat blocks without introducing air bubbles.

Don't forget to allow some time for "dry-out" or "curing" before the TIM can perform at full spec.

Korth wrote:
IC Diamond is known to be abrasive and actually scratch chip packages - this can make subsequent TIM applications less efficient (unless you mirror-hone the surface each time) and can make reselling parts with "damaged" and worn-looking factory markings very difficult.

Gelid GC-Extreme has low viscosity, it's thin and runny. It seems second to none for extreme sub-zero cooling, but it does gradually "burn off" over normal use and requires comparatively frequent (maybe even seasonal) reapplication. Prolimatech PK-3 is a remarkably similar product (in terms of actual usage). Other top thermal pastes are Phobya HeGrease and Tuniq TX-4, I have no experience with either.

There are, of course, even better products like Indigo Xtreme and the Coollaboratory Ultras. These are expensive and highly sensitive to application (highly variable performance). They are actual chunks of metal, not metal-bearing pastes, which require a temperature reflow procedure to install. And they often require a lot of mechanically subtractive patience to fully remove.

There are dozens of websites dedicated to authoritative TIM measuring, testing, and comparisons. And very few agree on their top picks, although the same handful of products (mentioned above) tend to always appear in the top-tier slots. OC-enthusiast sites are filled with combative diehards who will defend the honour of their chosen TIM with endless fervour.

I personally use venerable Arctic Silver 5 for these sorts of applications. It may not be quite as efficient as other products but it consistently ranks high up in the lists and is a steadfast, proven performer (it's also cheap and readily available). And - unlike most extreme TIM products - it's a throwback from the days when TIMs were engineered to last "forever" instead of being reapplied with great frequency. Reviewers and benchmarkers and other online "laboratories" excitedly swap out their parts every week as new parts roll in for testing - they are hardly aware of TIM longevity or degradation over time - other people (like me) tend to build platforms which last and upgrade/swap our parts on a more annual basis.