To build crossfire ready or not, an upgrade path, 2 years from now (build advice)?

[mephiston5]

I am planning a new PC build.

I am going to put a R9 390 card in my machine, with the assumption that I'll upgrade the card in 2 years.

Would I be better served to put a crossfire ready PSU in my build, with the intention of buying another R9 390 in 2 years to create a crossfire setup, or buying a lesser PSU with the intention of replacing my aging R9 390 with the best similar tier card I can find 2 years from now? From a overall performance perspective, I should say.

Thanks.

 

[ThatBuzzkiller]

Crossfire option seems to be the best at this rate since it's going to take ages for GPUs to evolve from now on when there's a clear slowdown in advance of transistor technology until EUV hits ...

 

[bystander36]

If there is a slow down in technology, there will also be a slow down in game settings, so if it comes to that, the GPU upgrade will just be pushed back.

 

[Doubtless]

Personally I'd just recommend that you stick to a single GPU setup unless you're wanting to run triple monitor, or looking to go 4k very soon. I used a similar strategy when I put together my first build several years ago, and I quickly found that I would have been better off just selling my single GPU and upgrading to a faster one (I ended up using 6950 crossfire and being fairly unhappy with it). The noise of two cards is a significant knock, as is the increased upfront cost of getting a multi GPU ready PSU.

I also think that GPU's will continue to move forward at a reasonable pace for the next couple of years. Look at how much improvement there is between early 28nm cards and the current flagship cards. A 980ti is faster than 680 SLI, so it stands to reason that a Pascal refresh flagship two years from now will be faster than 390x crossfire, not to mention that you might run into VRAM issues depending on how games move forward.

 

[3DVagabond]

I don't think Grenada/Hawaii will be significant in 2 more years. Unless you are concerned and are going to be replacing your current PSU anyway, I wouldn't bother until you actually have a need.

 

[mephiston5]

Personally I'd just recommend that you stick to a single GPU setup unless you're wanting to run triple monitor, or looking to go 4k very soon. I used a similar strategy when I put together my first build several years ago, and I quickly found that I would have been better off just selling my single GPU and upgrading to a faster one (I ended up using 6950 crossfire and being fairly unhappy with it). The noise of two cards is a significant knock, as is the increased upfront cost of getting a multi GPU ready PSU.

I also think that GPU's will continue to move forward at a reasonable pace for the next couple of years. Look at how much improvement there is between early 28nm cards and the current flagship cards. A 980ti is faster than 680 SLI, so it stands to reason that a Pascal refresh flagship two years from now will be faster than 390x crossfire, not to mention that you might run into VRAM issues depending on how games move forward.

I don't think I'll go 4k within two years. I will be running a dual monitor setup, however. Your logic regarding the 980ti vs. 680 SLI makes sense. I'll stick with running one card.

Thanks for the advice, everyone.

 

[Silverforce11]

Don't CF two mid-range cards for a new build its not a good idea. In 2 years time, these class of GPUs will be low-end.

Just get something to use that's enough now without breaking the bank and later upgrade to a new 14/16nm ff GPU.

 

[KingFatty]

What is the price difference between the two power supplies (one for the crossfire build, and one for the single-GPU build) that you are considering between?

 

[mephiston5]

What is the price difference between the two power supplies (one for the crossfire build, and one for the single-GPU build) that you are considering between?

The Antec 650W 80+ that I am looking at is going for about $100. I haven't really researched/priced anything with more wattage, but the SeaSonic SS-750KM3 750W seems nice, it runs about $130.

 

[KingFatty]

Hmm, $30 could be worth it just for insurance, and the peace of mind knowing that you always have the option if you want to take it. It might make you mentally regret not having the option, even if you never end up using it.

 

[Mondozei]

If there is a slow down in technology, there will also be a slow down in game settings, so if it comes to that, the GPU upgrade will just be pushed back.

Basically /thread.

I'd also repost something I wrote a bit earlier today, but which is still very relevant to this thread:

Lastly, compare a game like Ethan Carter, which can run surprisingly well on even mid-range GPUs, to most mainstream games in 2010. Of course there's been progress - and plenty of it. But a lot of devs just want to focus on brute force to solve everything. Alien: Isolation and MGS: GZ both proved that you can get very good effects on modest hardware if you spend time optimising the game, something which for some reason seems to be seen as a luxury these days.

If there is a slowdown in GPU progression then expect games to become increasingly better optimised. Still, here's why this recent mania about GPU tech slowdowns are overblown in the short term.

14 nm GPUs will be powerful. We know that. So you got yourself a year there. After that, you could see the introduction of a new arch.

Nvidia might just go for GP104 next year(while the GP100, the "full fat" version could wait until 2017 for consumers. Tesla cards would be out and possibly a new Titan earlier). This would ensure that if we don't see a node shrink in 2018, they could just introduce VOLTA that year instead on 14 nm.

Just like they introduced GM200 this year on 28 nm.

So 14 nm GP104 for 2016, 14 nm GP100 for 2017 and either node shrink and/or Volta for 2018. It probably depends how fast Samsung/GloFo/TSMC can get a decent 10 nm node working.

I don't think the next 2-3 years will see any signficiant GPU slowdowns. But after 2018 and beyond, that is possible. I'd still think you'd be better off getting a R9-390 this year and upgrade it 2 years from now to a midrange GPU. It's probably going to be at least 70-80% better than what you have now considering we're looking at two, not one, node jumps in one go AND a new arch.

 

[tential]

Imo you're insane not to ensure your own site psu can handle 2 cards. I can't see how you can't find a psu capable of doing that in your price range.

My biggest regret right now is worrying about using 2 cards. I could easily pick up multiples of the 290 instead of the 390 right now and get great perf far cheaper than even a single 390 but yet I have to worry about my powersupply so I've had to take every single component out currently (hard drives) before I even consider going crossfire lol

 

[Mushkins]

This question is asked almost daily, and the answer is always the same:

Unless you're going for eyefinity/4k max settings right now, you will almost always get better performance with less hiccups and issues by getting a single powerful card.

Buy a good card now, then sell your current good card and buy a new good card years down the road when it's time for you to upgrade. Technology moves too quickly to play games trying to future-proof, and you might not even be able to get the same card in the same hardware revision years from now when you decide to go SLI, which means even more potential problems.