Which GPU do you think have aged the worst in the last 3 years?

[crisium]

660 and 670 still on average are slower today compared to their contemporary competition the 7850 and 7950. They aged even worse when you consider that when they launched the 660 was on the heels of the 7870 (it was the 650 Ti Boost that used to go 1:1 with the 7850) and the 670 traded blows with the original 7970, only when the Never Settle drivers came out did GCN really take the advantage over Kepler. And that advantage is certainly greater still today.

They still play games fine, of course. But there are only two companies with two lines of cards and thus Kepler has aged poorly in the face of the fact that the only other cards from then, GCN 1.0, have aged better. It's just relativity.

 

[Seba]

660 and 670 still on average are slower today compared to their contemporary competition the 7850 and 7950. They aged even worse when you consider that when they launched the 660 was on the heels of the 7870 (it was the 650 Ti Boost that used to go 1:1 with the 7850) and the 670 traded blows with the original 7970, only when the Never Settle drivers came out did GCN really take the advantage over Kepler. And that advantage is certainly greater still today.

Where can I find those new benchmarks for GTX 660 vs. HD 7850, with the more recent drivers?

 

[Goatsecks]

I have owned a pair of asus DirectCU 670 (SLI) since Janurary 2013. I have been extremely tempted to upgrade (I play 1080p, some newer titles, wanted a single card with more Vram.), But decided the gains/cost were not worth it with the main gain being the extra VRAM headroom.

I know many kepler owners. None are tempted to upgrade at the moment. Even my kepler owning office mate, who has a preference for AMD cards (and that is fine!), is not tempted to upgrade at the moment.

I understand that this is all heuristic and I am sorry to piss on certain parades, but: Kepler (3 years old now, no?) is more than capable of giving 4+ years of use.

 

[Headfoot]

R9 290/290X with reference cooler. Maybe not the worst case, but those deserves to be mentioned, just to balance this thread.

What? No. Not even close.

$400 290 @ release date. Earned me over its $400 in litecoin mining, started as benchmarking worse than the 780. Then it beat the 780. Now it beats the 780 Ti. Absolutely no way the $400 msrp 290 has lost value. "Balancing" by mentioning a card that has not lost value hardly at all doesnt make any sense. The 290 is probably the BEST value GPU over its lifetime on 28nm given starting at $400 when 780 was $650 and now it sells for $245... The 290 is probably in the top 5 best value gpus of ALL TIME at this rate. It's the 8800 GT of 28nm.

If the sound wears on you, you can use the plethora of AIO cooler mounting brackets that have come out. It's really a lot easier to mount AIO to GPU than it ever has been.

290x has a better case, since it was quite a bit more expensive for not very much more performance, but it still was cheaper than the 780 by a lot at the time and has held up significantly better.

 

[Seba]

No informed buyer will choose a second hand R9 290/R9 290X with reference cooler. And the problem with an aftermarket cooler (air or liquid) is not the difficulty of installing it, but the price of that cooler. You would be better of buying the right card in the first place.

 

[Flapdrol1337]

Reference cooler 290's are a safer secondhand choice than aftermarket ones.

Due to coinmining many cards ran with the fan at high speeds for very long times, the reference cooler has a ball bearing fan, non reference fans almost all use sleeve bearing fans, which wear out under such circumstances.

 

[Seba]

The probability of having been used for intensive coin mining makes the second hand R9 290/290X even more undesirable (reference cooler or not).

 

[DeathReborn]

I think 680 or 780 probably the worst.
AMD do better, the 290s are like duracell batteries, just keep going strong lol...

Now if you asked the worst card of all time,clue below.

Surely the Intel 740 trumps even the FX...

 

[crisium]

Where can I find those new benchmarks for GTX 660 vs. HD 7850, with the more recent drivers?

It's a trend that I've noticed in benchmarks over the past year.

Battlefront, Rainbow Six, Batman, Call of Duty, GTA5 the two cards are virtually equal:
http://gamegpu.ru/images/remote/htt...efront_Beta-test-starwarsbattlefront_1920.jpg
http://gamegpu.ru/images/remote/htt...inbow_Six_Siege_Beta-test-RainbowSix_1920.jpg
http://gamegpu.ru/images/remote/htt...tman_Arkham_Knight__GPU_v_2.0-test-1920_l.jpg
http://gamegpu.ru/images/remote/htt...ty_Black_Ops_III_Beta-test-BlackOps3_1920.jpg
http://gamegpu.ru/images/remote/htt...Action-Grand_Theft_Auto_V__v.3-gta_v_1920.jpg

In Witcher 3, the 7850 was actually originally faster until the patch, now they are equal:
http://gamegpu.ru/images/remote/htt...-The_Witcher_3_Wild_Hunt_v._1.06-w_1920_h.png

Not every game is unfriendly to Kepler of course, but the amount has vastly increased this year. But back in 2012-2013 you did not see this parity except in very AMD favored games. The 660 used to be decisively faster than the 7850 (at stock anyway, Pitcairn could OC hardcore).

In early 2013, Nvidia launched the GTX 650 Ti Boost to go head to head with the 7850, while the 660 was on the tail of the 7870:
https://tpucdn.com/reviews/NVIDIA/GeForce_GTX_650_Ti_Boost/images/perfrel_1920.gif

This is partly due to TPU's game selection and sometimes unusual benchmarks, true enough, but when the 980 launched they had 660 = 7870:
https://tpucdn.com/reviews/NVIDIA/GeForce_GTX_980/images/perfrel_1920.gif

Then look at the very last TPU charts to feature a 660, it seems to tell the same story as the GameGPU benchmarks. The 660 is equal to the 7850 (265 rebadge) while the 7850 is pulling ahead of the 650 Ti boost:
https://tpucdn.com/reviews/Gigabyte/GTX_960_OC/images/perfrel_1920.gif

It just seems that the 660 went from being in-between the 7850 and 7870, sometimes very close to the latter, to basically parity. See when the 650 Ti Boost launched it was priced to compete with the 7850 while the 660 price was competing with the 7870 (TPU 650 Ti Boost Review). Thus, poor aging relative to the only other competing architecture.

 

[Headfoot]

No informed buyer will choose a second hand R9 290/R9 290X with reference cooler. And the problem with an aftermarket cooler (air or liquid) is not the difficulty of installing it, but the price of that cooler. You would be better of buying the right card in the first place.

You can buy second hand reference 290s for $160. AIO+Bracket for $60. End solution = cooler, quieter and cheaper. You just lose the warranty. Coinmining is relevant to wear on the cooler, so if you replace it with an AIO+bracket it becomes irrelevant. I'd honestly be more worried about a reference 290 used for gaming since the default fan profile let those things get extremely hot where most coin miners are significantly more technically literate (undervolting, better fan profiles, open bench cooling...). 290 is not, and never was, a bad value. It has always been extremely good value and it has preserved that value extraordinarily well for a GPU.

 

[Seba]

It's a trend that I've noticed in benchmarks over the past year.

Battlefront, Rainbow Six, Batman, Call of Duty, GTA5 the two cards are virtually equal:
http://gamegpu.ru/images/remote/http...front_1920.jpg
http://gamegpu.ru/images/remote/http...owSix_1920.jpg
http://gamegpu.ru/images/remote/http...est-1920_l.jpg
http://gamegpu.ru/images/remote/http...kOps3_1920.jpg
http://gamegpu.ru/images/remote/http...gta_v_1920.jpg

In Witcher 3, the 7850 was actually originally faster until the patch, now they are equal:
http://gamegpu.ru/images/remote/http...6-w_1920_h.png

In all but one of those 6 benchmarks, GTX 660 is still above HD 7850, as it was from the begining. The difference seems small because they use improper settings (too demanding) for cards at that performance level.

 

[Deders]

Has anyone actually done any tests with mature drivers for the 780 compared to equivalent AMD? Most of the post 970/980 games were tested with drivers that were developed prioritising performance on those cards to help sales. It was often not until a couple of months later that Kepler's performance was optimal.

 

[Seba]

You can buy second hand reference 290s for $160. AIO+Bracket for $60.

And that proves the point that R9 290 with reference cooler has lost a great amount of it's initial value (purchase cost).

As the original owner I can only get $160 for my expensive (~$400, maybe more) two year old card (at most).
As a second hand card buyer, after paying $160 for the card, I have to pay another $60 (that is some cheap cooler, b.t.w.) for an aftermarket cooler, in order to reduce the noise and throttling.

 

[crisium]

In all but one of those 6 benchmarks, GTX 660 is still above HD 7850, as it was from the begining. The difference seems small because they use improper settings (too demanding) for cards at that performance level.

1 frame difference is equality. If you can't see the degree of difference with 660 of old and new compared to 7870 and 7850 then I'm sorry there's nothing I can do for you, the evidence is in front of your eyes as well in the TPU charts.

 

[Seba]

When you drop to 20 fps area, 1 fps is not equality. The test is not relevant, because the game is not playable at those settings.

From the beginning GTX 660 and HD 7850 were close, with GTX 660 slightly better in most games. The benchmarks you posted (while using improper game settings for GTX 660 and HD 7850 graphics card class) are showing the same thing.

 

[iiiankiii]

The GTX 980 Ti has done well in the DX12 benchmarks from what I've seen compared to the Fury, plus you get better DX11 performance so I'm guessing the 980 Ti will turn out to be a better buy than the GTX 780 was.




Yes, I agree. Attacking the whole Keplier lineup doesn't make sense. The low-end 660/670 did indeed age well. And I also tussled with RS a little over the 680. I'd say that it wasn't doing as bad as he suggested since it was performing about the same as the midrange GPU of the next series, the 770. Which is what has normally/historically happened.

Overall, I think the 600-series did an okay job.

Bro. I think the whole point of this thread is to show that Kepler isn't aging as well as GCN. It doesn't matter what Kepler card it is. Compared to GCN, it's falling behind going forward.

Having said that, most mid-range cards seem to age better because it is reasonable priced (~$200). Historically, mid-range cards usually gives you the best performance/$. I believe that's the reason the performance lost isn't too bad in respect to $.

However, Kepler, in respect to GCN, IS falling behind. Kepler used to BEAT, or go toe to toe with cards within the same tier (GTX 680 vs GTX 7970, GTX 770 vx R9 280x, GTX 660 vs 7850, etc.). Now, Kepler is losing ground to GCN no matter what tier it is. Keper just isn't aging well with respect to GCN.

 

[crisium]

When you drop to 20 fps area, 1 fps is not equality. The test is not relevant, because the game is not playable at those settings.

From the beginning GTX 660 and HD 7850 were close, with GTX 660 slightly better in most games. The benchmarks you posted (while using improper game settings for GTX 660 and HD 7850 graphics card class) are showing the same thing.

TPU showed that in 2013-2014 (see the two charts from that time frame), the 660 was closer to the 7870 than the 7850 in performance. Users who bought a 660 in that time frame in fact paid closer to the 7870 price than the 7850. In 2015, 7850 and 660 are virtually equal. They have aged worse. Please don't be a Baghdad Bob for the green defense force.

 

[Seba]

Anandtech GPU Bench 2013:
GTX 660 vs. HD 7850
GTX 660 vs. HD 7870
HD 7850 vs. HD 7870

GTX 660 is closer to HD 7850 than to HD 7870 in these games benchmarks. This was always the case, as still is.

 

[crisium]

Of course there is variance in benchmarks.

660 launch Sep 2012, before the Never Settle drivers. It is faster than the 7870:
http://cdn.sweclockers.com/artikel/diagram/3220?key=7e3aef8ea6ec3c8b912583434109a4b2

Mid 2013, things have stabilized to where 660 is in-between the two Pitcairns yet closer to the 7870:
http://cdn.sweclockers.com/artikel/diagram/4280?key=5e411b6fd0af87f9dbfb63257113c0f1

January 2015, 660 is radically slower than the 7870/270X and even basically equal to the 7850/265.
http://cdn.sweclockers.com/artikel/diagram/9103?key=93f1ae62c03c326b1cfff61415c7d19e

And it's gotten worse in 2015 games released since January (see iiiankiii for more recent games), but that's the last 660 I could find with Sweclockers.

I'm demonstrating trends over time, which you haven't done yet. Of course there is variance dependent on game selection. But in 2015, the variance leans heavily on parity where there used to be superiority. It's not a universal truth, yet despite this there is little room for doubt that mid-range Kepler line (660 through 680/770) has suffered relative to GCN 1.0 (7850 through 7970/280X). You're welcome to be the voice of dissent, but you stand with few.

 

[iiiankiii]

When you drop to 20 fps area, 1 fps is not equality. The test is not relevant, because the game is not playable at those settings.

From the beginning GTX 660 and HD 7850 were close, with GTX 660 slightly better in most games. The benchmarks you posted (while using improper game settings for GTX 660 and HD 7850 graphics card class) are showing the same thing.

That's right. The GTX 660 used to be faster than the HD 7850. BUT, in recent games, the HD 7850 is usually faster. However, it may not be a lot faster, but, instead of the HD 7850 losing to the GTX 660, the HD 7850 is now winning in most games. Also, I tried to pick the settings that made these lower end cards the most playable. Honestly, though, the difference is hardly noticeable. This is purely academic.

For some of the benches, no HD 7850 were used. A R9 265 were used. Since the R9 265 is a rebrand of the HD7850, it's close enough. But, a HD 7850 is ~5% slower than R9 265.

Oh, another interesting point. Most of these games are GAMEWORKS, too.

 

[tential]

I love how this whole conversation is based on which card ended up being faster...

So you guys care more about your GPU performance at the end of its life cycle, rather than the beginning of its life cycle (when you're turning down settings anyway and looking to upgrade).

That's wild.

 

[fourdegrees11]

What is up with Project Cars and a 960 2gb being noticeably faster then a 290X? I'm definitely impressed with the 290x at 1080 for every other benchmark. Makes me feel great about snagging a never opened XFX DD 290x for $235 on ebay!

 

[Kenmitch]

What is up with Project Cars and a 960 2gb being noticeably faster then a 290X? I'm definitely impressed with the 290x at 1080 for every other benchmark. Makes me feel great about snagging a never opened XFX DD 290x for $235 on ebay!

Heavily optimized for nvidia hardware. Translated to English = crappy AMD drivers some would lead you to believe.

 

[dogen1]

Heavily optimized for nvidia hardware. Translated to English = crappy AMD drivers some would lead you to believe.

We don't really know.

Pperhaps the game is heavily dependent on a certain dx11 feature amd hardware doesn't support on pc...

 

[tential]

What is up with Project Cars and a 960 2gb being noticeably faster then a 290X? I'm definitely impressed with the 290x at 1080 for every other benchmark. Makes me feel great about snagging a never opened XFX DD 290x for $235 on ebay!

For how much project cars is benchmarked, you would think a lot of people have purchased the game on PC. The complete opposite is true however.