Building a Budget PC: Should You Buy a Used Graphics Card?

[RealNeil]

All I'm saying is the days of a top tier card for around 500 are long gone and future 1080 tier cards will be at least double that from now on. I'd put money on it that not only will the prices not drop, but that they are not done peaking yet. The gamers are the ones who will resist the hardest, but all other sectors will just pay and go on about their business. Miners, artists, designers, developers, scientific community and the like, these are the ones who will continue to buy with little resistance to inflating prices and they will generate enough revenue for the manufacturers that they will not see any reason to reduce the prices.

You may be right, but I hope not.
The last GPU I bought cost me 650 bucks (MSI GTX 1080Ti GAMING-X TRIO) and I was choking on that. It was used for a short time until the owner upgraded.
According to current prices, I got it for about half of what it sells for retail.
So yes, buying used is a great idea as long as it works when you get it.

 

[maikal]

I may be a super exception to the rule, but I have built 4 PC's with nothing but used EBay parts. Bugged the sellers for days about the parts, verified warranties, verified build-on dates and so forth. Got R7 290X, RX 480 8gb, R9 390 and most recently used and later refurbished GTX 1070 card. All used, but bought within the year I bought them from the sellers. All of them for less than the retail price and all of them for less than the over-price on eBay. I refuse to buy from anyone but private re-sellers who have a few thing on eBay! I don't buy from these sellers who are like thrift store resellers or people who have like "10" sold or "5" left type of sellers. There are clues and signs all over eBay that will tell you who is a gamer that is just trying to pay for his upgrade and who is a retail reseller in disguise. You can appeal and negotiate to the gamer about the price, you cant with these resellers who buy a dozen GPU's or CPU's and gouge people on Ebay! I just refused to buy from them. But for one to have success you need to do your research, you need patience and you need to understand that the process will take time.

 

[maikal]

Folks, can you at least fix the titles of your artcles?

"Building a Budget PC: Should You Buy a Used Graphics Card?"

How about we say:

"Building a GAMING Budget PC: Should You Buy a Used Graphics Card?", because all you do is run games and gaming benchmarks. So what? Even with demanding real world software like PhotoShop and Solidworks, a fairly basic card with as little as 1GB of on-board memory will suffice. Nope, Solidworks will not render blindingly fast, but it would be good enough.

As a general rule, when TechSpot does articles full of gaming benchmarks, please call them what they are. Put the word GAMING in the title. Then I can afford to skip over them, having little interest in the topic.

You can tell the way this article is going from the beginning. If you had no interest in gaming, "Most gamers might scoff at a GTX 750 Ti in 2018, but it still ...". Its the third sentence! If you have no interest in gaming at that point you should have left! But instead you chose to read on so you could later ***** and complain about yet another gaming GPU comparison! Do us a favor, leave early and don't take up valuable posting space for your useless to the discussion post!

 

[ddferrari]

Folks, can you at least fix the titles of your artcles?

"Building a Budget PC: Should You Buy a Used Graphics Card?"

How about we say:

"Building a GAMING Budget PC: Should You Buy a Used Graphics Card?", because all you do is run games and gaming benchmarks. So what? Even with demanding real world software like PhotoShop and Solidworks, a fairly basic card with as little as 1GB of on-board memory will suffice. Nope, Solidworks will not render blindingly fast, but it would be good enough.

As a general rule, when TechSpot does articles full of gaming benchmarks, please call them what they are. Put the word GAMING in the title. Then I can afford to skip over them, having little interest in the topic.

Wow, Ben, you sound like a super busy guy- I can see why you're asking websites to cater to your needs! Are you a millennial, by chance?

 

[meric]

I wouldn't buy a used GPU even if it's very cheap. Especially if it's such an old generation card. Because I can never know how heavily it was used, was it used in a 7/24 mining rig? was it in a heavily overclocked gaming rig? how much was it stressed? how much more breath does it have? you know, 2nd hand ones usually dont come with a warranty, so that's a complete gamble. Instead, I'd go for a mediocre card from current gen. which is brand new and may cost a bit more.

 

[RealNeil]

I wouldn't buy a used GPU even if it's very cheap

Since my last post describing the GTX-1080Ti card (working perfectly by the way) that I bought used, I bought a pair of Gigabyte GTX-1070Ti Gaming G1 8GB cards. They'll be here tomorrow or Thursday. I got them slightly used for $400 bucks each.
For me, used is the way to go.

 

[Ben Myers]

Wow, Ben, you sound like a super busy guy- I can see why you're asking websites to cater to your needs! Are you a millennial, by chance?

First, I am not even close to being a millenial. Second, I work in this biz for a living. Third, the article spoke almost exclusively in terms of gaming systems, so why wouldn't the title actually indicate what the article is about? Finally, I buy used graphics cards in ones, twos and in lots. Some go in the recycling bin, others into refurb desktops. Snark!

 

[msroadkill612]

Just sayin, bu an option as a ~affordable disposeable place holder, a $100 2200g is an option. More muscular apuS than the 2400g are inevitable soon.

Further, to get fundamentals straight, its not gpuS that are scarce, its gpu ram, and only the apu sidesteps that problem, warts and all.

 

[captaincranky]

I bought a used 970 awhile back its still rockin like 2-3 years later? The trade off is no warranty... It's a gamble. With the rise in GPU prices........ Things are looking sad for new gamers.

Typically if a card is going to fail it's going to do so within the first month or it will be good for at least 5 years.

You can take allot of the gamble out by stress testing the card when you get it.

I'm not quite sure I follow your logic. If you buy a used card, stress test it and it breaks, you are at the mercy of the (ostensibly) private seller, to make good on a claim of, "money back guaranteed", and return shipping will almost surely be at your expense.

 

[Evernessince]

I'm not quite sure I follow your logic. If you buy a used card, stress test it and it breaks, you are at the mercy of the (ostensibly) private seller, to make good on a claim of, "money back guaranteed", and return shipping will almost surely be at your expense.

And yet and I can't tell you how many times I got a refund for doing just that. FYI return shipping is on the seller if the product is defective. Running a video card under full load and it crashing or having issues is considered defective by both AMD and Nvidia.

I've never had a single issue with returning a broken graphics card and I've never had to pay return shipping. The seller can refuse to take a return but eBay will step in and force them to take it anyways. Then again, I'm not trying to scam anyone. I've had multiple cases where the seller would insist the card was perfect when they tested it until I showed them video evidence and they take the return.

Stress test doesn't infer that you are doing something with the card that is designed to break it. It simply pushes the card to the maximum specifications it was designed to operate at. It's completely within spec and any card not working at those factory specifications is defective.

 

[RealNeil]

I'm not quite sure I follow your logic. If you buy a used card, stress test it and it breaks, you are at the mercy of the (ostensibly) private seller, to make good on a claim of, "money back guaranteed", and return shipping will almost surely be at your expense.

I usually run a few benchmarks with the used GPUs that I buy, but I don't run stress tests on any of them.
All of them are pretty solid. All are running in SLI except for the 1080Ti.

No issues at all.

 

[msroadkill612]

And yet and I can't tell you how many times I got a refund for doing just that. FYI return shipping is on the seller if the product is defective. Running a video card under full load and it crashing or having issues is considered defective by both AMD and Nvidia.

I've never had a single issue with returning a broken graphics card and I've never had to pay return shipping. The seller can refuse to take a return but eBay will step in and force them to take it anyways. Then again, I'm not trying to scam anyone. I've had multiple cases where the seller would insist the card was perfect when they tested it until I showed them video evidence and they take the return.

Stress test doesn't infer that you are doing something with the card that is designed to break it. It simply pushes the card to the maximum specifications it was designed to operate at. It's completely within spec and any card not working at those factory specifications is defective.

Do you forewarn them? If I were a private seller I would prefer to stress test it myself before shipping.

 

[msroadkill612]

Just so folks know, the authors mindset is "Based on the data here, you'll ideally want the GTX 760 or HD 7950 for a smooth experience.", referring to av/min scores of 61/52 fps & 59/50 fps.

Personally I thought the apuS fared well considering. Its no "gamer", but you can have fun gaming on it. An OC and a few extra tlc dollars on ram & cooling, & most games will be playable at reasonable settings.

with the apu u also become part of the cutting edge zen/vega ecosystem. all the development funded by professionals will filter down to your ~identical zen/vega world.

If/when u upgrade to a vega dgpu, your precious time invested in learning the APUs's Zen/Vega tools, will have been well invested.

Similarly, desktop apuS will benefit from "below", from what promises to be a massive incursion of the apu into the huge laptop market.

https://segmentnext.com/2018/06/28/intel-10nm-delay-resulting/

The apu alternative posed here, is a union of two orphaned dead zones.

The very fresh apu will improve radically from software/firmware within a year, just as ryzen1 & vega has (more so imo as better harmonising gpu & cpu on IF holds so much promise).

My take away was: You have to cop a lot of downsides to get not much better

 

[Evernessince]

Do you forewarn them? If I were a private seller I would prefer to stress test it myself before shipping.

The seller should always stress test before trying to sell their product. That said, it doesn't matter what the seller's description says they did, you always test new parts. If I were to trust every description I've read I'd be out at least $20,000 by now.

Both a stress test and playing a game push a graphics card to 100% (assuming you don't have a CPU bottleneck). Apparently, despite this being a tech forum, people are worried that this somehow could damage the card. In reality, a stress test can be any game or benchmark that achieves the prior stated goal. Anyone playing games on their graphics card are stressing it. The only difference is "stress test" infers that you specifically making sure that CPU isn't a bottleneck and that you are monitoring temperatures. That if anything makes it safer then blindly running a card to full load without measuring voltage and temps.

Heck I got a defective 4790K in the other day where the seller had no idea it was defective as issues would only appear under full load. Got on the phone with Intel and their CPU diagnostic utility itself stress tests the CPU and verified it was defective.

 

[msroadkill612]

Interesting. Could you give we newbs some links / guidelines, or should we just google it?

What are these tests?

 

[Evernessince]

Interesting. Could you give we newbs some links / guidelines, or should we just google it?

What are these tests?

The idea of a stress test is simply to ensure the video card doesn't have any issues like crashing, artifacting, lower than expected performance, ect.

You can choose to use any game you wish so long as you get GPU usage to 100%. Avoid games that are notoriously un-optimized like assassin's creed. Keep your temperature monitoring software up during the test and make sure voltages and temperature keep within a safe range.

For most games, you want to set the resolution to 1440p or higher and graphics setting to max. This will ensure a CPU bottleneck won't come into play for a majority of games. Remember the main purpose of this is to keep the GPU at 100% the entire time. If you aren't doing that then it isn't going to be effective.

Since this isn't a benchmark you can simply play the game normally, just make sure to do so for at least an hour. Tom's hardware uses a combination of games and benchmarks to stress test it's GPUs. You don't need to go that far, typically 1 hour of 100% GPU load will do the trick. Most tech websites also use pre-defined routes and locations in a game to ensure consistency but that also isn't required for a consumer.

If after the test you experienced no issues and the temperatures stayed within the expected range with no voltage irregularities, you are good to go. I say "within the expected range" as cards like the R9 290X run at 92c with the stock cooler. For those cards I always set a custom fan target of 82c.

 

[msroadkill612]

The idea of a stress test is simply to ensure the video card doesn't have any issues like crashing, artifacting, lower than expected performance, ect.

You can choose to use any game you wish so long as you get GPU usage to 100%. Avoid games that are notoriously un-optimized like assassin's creed. Keep your temperature monitoring software up during the test and make sure voltages and temperature keep within a safe range.

For most games, you want to set the resolution to 1440p or higher and graphics setting to max. This will ensure a CPU bottleneck won't come into play for a majority of games. Remember the main purpose of this is to keep the GPU at 100% the entire time. If you aren't going that they it isn't going to be effective.

Since this isn't a benchmark you can simply play the game normally, just make sure to do so for at least an hour. Tom's hardware uses a combination of games and benchmarks to stress test it's GPUs. You don't need to go that far, typically 1 hour of 100% GPU load will do the trick. Most tech websites also use pre-defined routes and locations in a game to ensure consistency but that also isn't required for a consumer.

If after the test you experienced no issues and the temperatures stayed within the expected range with no voltage irregularities, you are good to go. I say "within the expected range" as cards like the R9 290X run at 92c with the stock cooler. For those cards I always set a custom fan target of 82c.

Thanks. Much appreciate your efforts.