Ok, both the Arcadia III case and the Asus Strix GTX 980 Ti have arrived, and I put everything together for a quick test. Seems to be working fine so far, the card isn't too loud even under full load, as the new case is well ventilated. Quick comparison between the GTX 970 and the GTX 980 Ti via the built-in benchmark from Tomb Raider (2013) using Ultimate settings at 1080p, with TressFX turned on as well:

GTX 970

• Min FPS: 68.0
• Max FPS: 126.0
• Average FPS: 97.6

GTX 980 Ti

• Min FPS: 114.0
• Max FPS: 192.0
• Average FPS: 154.5

In both cases, the test was performed under Windows 10 LTSC 21H2 using Nvidia drivers 560.94. And yeah, the game is much more playable on the GTX 980 Ti.

Also, I tested Prince of Persia: Sands of Time with 8xSGSSAA at 1600x1200 under WinXP. It runs a bit better than on the GTX 970, but it's still not perfect. In particular, during the final battle with The Vizier, the performance still tanks when he casts that summoning/illusion spell from behind the (transparent) curtains. Otherwise, the game mostly runs fine at those settings.

Joseph_Joestar wrote on 2025-05-13, 19:53:

Ok, both the Arcadia III case and the Asus Strix GTX 980 Ti have arrived, and I put everything together for a quick test. Seems […]
Show full quote

Ok, both the Arcadia III case and the Asus Strix GTX 980 Ti have arrived, and I put everything together for a quick test. Seems to be working fine so far, the card isn't too loud even under full load, as the new case is well ventilated. Quick comparison between the GTX 970 and the GTX 980 Ti via the built-in benchmark from Tomb Raider (2013) using Ultimate settings at 1080p, with TressFX turned on as well:

GTX 970

• Min FPS: 68.0
• Max FPS: 126.0
• Average FPS: 97.6

GTX 980 Ti

• Min FPS: 114.0
• Max FPS: 192.0
• Average FPS: 154.5

In both cases, the test was performed under Windows 10 LTSC 21H2 using Nvidia drivers 560.94. And yeah, the game is much more playable on the GTX 980 Ti.

Also, I tested Prince of Persia: Sands of Time with 8xSGSSAA at 1600x1200 under WinXP. It runs a bit better than on the GTX 970, but it's still not perfect. In particular, during the final battle with The Vizier, the performance still tanks when he casts that summoning/illusion spell from behind the (transparent) curtains. Otherwise, the game mostly runs fine at those settings.

Wow, that's an immense performance difference! But again, it's almost double the processing units.

H3nrik V! wrote on 2025-05-14, 06:46:

Wow, that's an immense performance difference! But again, it's almost double the processing units.

Yeah, I was a bit surprised as well. But then again, the Asus Strix is factory overclocked.

From what I gather, the performance difference between a GTX 970 and a GTX 980 Ti is usually around 25% but that can vary from game to game. This Techpowerup review compares the Asus Strix model to many contemporary GPUs and goes into a lot more detail.

Joseph_Joestar wrote on 2025-05-14, 07:39:

H3nrik V! wrote on 2025-05-14, 06:46:

Wow, that's an immense performance difference! But again, it's almost double the processing units.

Yeah, I was a bit surprised as well. But then again, the Asus Strix is factory overclocked.

From what I gather, the performance difference between a GTX 970 and a GTX 980 Ti is usually around 25% but that can vary from game to game. This Techpowerup review compares the Asus Strix model to many contemporary GPUs and goes into a lot more detail.

That's cool. Bu yes, 2816 vs 1662 Shader units should give something. My 980Ti is "non-factory" overclocked, though 😉

A few notes on the Raijintek Arcadia III case. I got this one since it was cheap (around 33 EUR), and that's probably because it only comes with a single (exhaust) fan. Fortunately, I had a pair of BeQuiet! Silent Wings 3 in the old case, which I'm now reusing as my intake fans. The whole system is fairly quiet until it reaches about 50% load, at which point the GPU and CPU fans gradually ramp up. Even so, the noise level is acceptable to me, and there is no unpleasant "whirring" or anything like that, just a gentle hum.

One big issue that I have with the Arcadia III is how SATA SSDs are mounted to the back of the case. Essentially, they are screwed in, so if you ever need to replace them, you have to take out the motherboard first. This baffles me, as it would have made much more sense to include a detachable caddy or just rubber grommets that could be used for mounting the SSDs. I guess they really needed to save those 50 cents or whatever that would have cost. On the plus side, the drive cage next to the PSU is easily accessible and can be quickly detached if you need to take out the HDD or SSDs mounted there.

I get that my use case is a bit unusual for this day and age, as I have 4 SATA SSDs and one 3.5" HDD in this system, but it's still annoying that those back mounts were such an afterthought. Also, due to the orientation of the mounted SSD, it's impossible to use angled SATA cables, so I had to order a bunch of straight ones. Still waiting on those to arrive, and that's the main reason why the build isn't fully finished yet.

And it's done. Updated the pics in the first post, as well as the benchmarks.

Interestingly, there wasn't that much difference in performance between the GTX 980 Ti and the GTX 970 in the games that I've tested. It looks like I'm CPU bound, so the performance benefit will only be felt when forcing advanced forms of anti aliasing like SGSSAA and such. I had to ditch the internal SATA DVD drive, since the new case has no room for it, but I have an external USB DVD drive which will work just fine for playing physical copies of my games. I also removed the PCI network card as it wasn't being used much. Lastly, I replaced the stock exhaust fan that came with the case with a Noctua. All in all, I'm pretty happy with how this rig turned out. 😀

Joseph_Joestar wrote on 2025-05-15, 12:09:

And it's done. Updated the pics in the first post, as well as the benchmarks.

Wow. That GPU really is sagging. Looks like a great build once you sort the sag out.

As for performance, yeah, not surprised to be CPU limited. The GPU is several years newer than the platform its running in. I think the advantage over the 970 is noise. It least in my case the 980 Ti was much quiter. Expecially if CPU bound and the 980 Ti is barely breaking a sweat

RetroPCCupboard wrote on 2025-05-15, 13:58:

Wow. That GPU really is sagging. Looks like a great build once you sort the sag out.

Yeah, I think I'll have to build my own support stick from wood. All the ones that are available to me locally have a magnetic bottom, and I don't want to put that right above the hard drive cage.

One more thing to note, the cooler assembly of this card is thinner near the mounting bracket than it is at the tail end of the card. That might skew the perspective a bit and make it seem more crooked than it actually is.

RetroPCCupboard wrote on 2025-05-15, 13:58:

I think the advantage over the 970 is noise. It least in my case the 980 Ti was much quiter.

My particular 980 Ti (Asus) and my 970 (MSI) are about the same in terms of noise. Both cards are practically inaudible when under 50% load, provided that the case has good airflow. Above 50% load, the GPU fans gradually start ramping up, but never get to unpleasant noise levels. Of course, the 980 Ti doesn't hit 50% load as quickly as the 970, which effectively makes it the quieter card.

I tested a few more WinXP games on the 980 Ti using the 1600x1200 resolution, with 8xSGSSAA and 16xAF forced on top of that via Nvidia Profile Inspector:

Splinter Cell (2003)

Prince of Persia: The Sands of Time (2003)

Far Cry (2004)

Battlefield 2 (2005)

At those settings, only Sands of Time can (very rarely) increase GPU load over 90% when lots of transparency effects are on screen. Otherwise, it hovers around 30-50% during normal gameplay. Splinter Cell, Far Cry and Battlefield 2 usually stay below 30% load. This means that the entire system is practically noise-free, even after enabling the best possible AA/AF.

Repasted the GTX 980 Ti with some Thermal Grizzly Kryonaut.

I made the card run for 5 minutes under full load (with V-Sync disabled) and the (boost) core clock increased by 30 MHz compared to the previous result. With V-Sync enabled, temperatures dropped by about 7°C from before changing the paste. For the before/after comparison, I used the approach suggested in this video.

7 degrees is a very decent reduction in temperature. Nice. I may do that to mine. I haven't actually checked it's running temps yet

RetroPCCupboard wrote on 2025-06-17, 08:39:

7 degrees is a very decent reduction in temperature. Nice. I may do that to mine. I haven't actually checked it's running temps yet

The temperature is kinda difficult to gauge with these cards, since they start thermal throttling at a pre-defined point. Mine does it at 83C, though that can be changed via MSI Afterburner.

You can still get a good temperature reading if you load up a game, turn V-Sync on and keep the GPU load steady (e.g. fire up Doom 2016 or something and just look at a wall). But the better indicator is the boost clock value at full load and with a fixed fan speed, as shown in Jay's video linked above.

I repasted my GeForce 9800 GTX and GeForce GTX 260 with just MX-4. Standard 9800 GT had paste that was so hard it had to be scraped off die by debit card. Temperatures dropped by about 3-4'C immediately. All my other newer cards haven't been repasted yet and are more noisy than they should be. Repasting seems to be needed about every 5 years. I will probably buy Thermal Grizzly Kryonaut for the better cards. I wonder how often Kryonaut needs to be repasted.

MX-4 is 8.5 W/mK
Thermal Grizzly Kryonaut is 12,5 W/mK

I wonder how the famous Intel TIM does after 10 years.

Alex get Roman's new Duronaut it's he's new paste. He says it has a longer life and better performance than Kryonaut. I'm going to get a tube of it soon.

You're right. Kryonaut has serious problem with high temperatures and durability. I'm going to get Duronaut next time I order parts for my AM2+ system.

I know he also said that Duronaut is a lot easier to spread. When I order parts to finish a computer for my kids I'm going to order a tube of it as well.

Honestly i hate the idea of "repasting" as maintenance. People kill so much hardware doing this... Also there is so much more to it than just thermal conductivity and manufacturers often chose thermal interface based on many factors, like longevity and consistency of application. Thermal conductivity from specs is practically useless for judging how well it will work since it can be measured in different ways and there are different other factors which affect heat transfer.

I've seen 20 year old thermal interface in some servers and some videocards which was still in great condition and changing it did not improve things, except i am sure thermal compound i've used will not last 20 years...

Then there is the issue that there is a lot of randomness involved - it can be clearly observed on modern high power GPUs with bare die and ability to see hotspot temperature. You apply new fancy thermal compound, "GPU temperature" is a few C down, but suddenly hotspot is 20C higher. You remove it, reapply again the same way, and suddenly it is fine. You simply would not know this on older card with no ability to see hotspot temperature and then it'll fail some time later for unknown reason while being seemingly kept cool. A lot of effort goes into preventing this from the factory.

So i personally prefer to change thermal interface only if there is some indication it is needed. Like temperature increase over time, or being excessively old, like from 90s, with unknown history.

I also, for some unknown and seemingly illogical reason, dislike Thermal Grizzly. Something just feels... wrong. Too aggressive marketing (sponsoring every youtuber in existence), obscene prices, obvious signs of trying to get every bit of perceived performance at any cost. Most comparable to charging li-ion batteries to higher limit of what's possible without them exploding instead of normal voltage, which a lot of manufacturers used to do a lot and some still do. +5% capacity, 10x worse longevity, but if it allows better scores in reviews - anything goes. After all reviews never show anything related to longevity.

What i've wanted to try for a while is MX-6, higher viscosity is a good thing, should make it last longer in practice.

Archer MX-6 is a good paste but it does suffer from pump out pretty good

chrismeyer6 wrote on 2025-06-24, 23:41:

I know he also said that Duronaut is a lot easier to spread.

Yeah, Kryonaut has the consistency of a bubble gum that's been sitting on the pavement for hours during a hot summer day. 😁 I'll also be switching to Duronaut in the future. It wasn't yet available for purchase in my local shops when I was repasting this card.

Archer57 wrote on 2025-06-25, 00:39:

Honestly i hate the idea of "repasting" as maintenance. People kill so much hardware doing this...

Yup, it can be dangerous. That said, the disassembly of my Asus Strix GTX 980 Ti was pretty straightforward. But I'm certainly not looking forward to opening up my Xbox 360 S or my PlayStation 4 Pro when the time comes. The latter is almost 10 years old now, and the fan has started to get pretty loud, despite being cleaned regularly.

Joseph_Joestar wrote on 2025-06-25, 05:41:

But I'm certainly not looking forward to opening up my Xbox 360 S

I remember the regular Xbox 360 being pretty straight forward in the inside when I opened it to replace the incredibly loud fans years ago, but the case itself was a huge pain in the ass. I couldn't get those damn plastic clips open without scratching the hell out of it.

Hopefully they made the slim less annoying to open.