Best Graphics Cards For The Money: October 2014 (Archive)

[anthony8989]

GPUs are usually not memory bandwidth starved,

They usually are.

The only reason the R9-285 manages to catch up with the R9-280 is because AMD invested more processing resources in texture compression and other bandwidth optimization tricks, same goes for Maxwell. But those tricks can only go so far

No, they aren't. If they were, Nvidia would have gained a lot from using a 384-bit memory interface for the GTX 980 and 970. They didn't, because it wasn't worth it.

Or - in my opinion - Nvidia didn't because it was an unnecessary expense. They already have all the Radeon cards beat by a significant margin on 256b interfaces. Stretching to 384b - no matter which way you look at it - only further increases the already large gap in performance at literally no benefit to the Geforce brand.

It is easy to imagine that 256b is cheaper than 384b. It's also easy to imagine that larger memory buffers equate to higher power consumption. So why jump to a more expensive card that consumes more power just to beat an already beaten competition. That's sadism!

 

[anthony8989]

Your own benchmark results are not on par with your hierarchy chart. Isn't the hierarchy chart based on Tom's Hardware benchmarks?

This article is more on the editorial side, based on what *I* benchmark. Indeed, I've seen cards like the 760 fall behind in newer benchmarks (check our Radeon 285 article).

Our Charts benchmark suite is kind of old, but it's in the process of an update, so we should see things reflect more accurately in the future.

Cleeve! The GTX 760 is 10-15% stronger than the GTX 580 and you have it one tier lower. Explain yourself!

 

[InvalidError]

No, they aren't. If they were, Nvidia would have gained a lot from using a 384-bit memory interface for the GTX 980 and 970. They didn't, because it wasn't worth it.

You are forgetting the efficiency gains from Maxwell again. Those are one-shot improvements and without them, Nvidia would not have been able to afford the memory interface reduction.

Wider memory interfaces are expensive to implement and memory is not showing any signs of getting any faster any time soon either - at least not in convenient and cost-effective form factors - so both AMD and Nvidia are stuck having to put most of their effort in reducing bandwidth dependence if they want to have anything to push for when die shrink season comes around.

 

[Brogan]

When GTX 980Ti scheduled for release? I hope before holidays.

Will there be a GTX Titan Black II based on Maxwell?

Is a Ti version even planned for 9 series?

 

[anthony8989]

When GTX 980Ti scheduled for release? I hope before holidays.

Will there be a GTX Titan Black II based on Maxwell?

Is a Ti version even planned for 9 series?

Scheduled for release 8-12 hours after the R9 390x trolol

 

[jwl3]

I've said this before and I'll say it again - why in the world are AMD cards so much cheaper than Nvidia? Is it generally accepted that Nvidia is a higher quality with better drivers? Look at this October issue - 90% of the recommended cards are AMD because of the price.

Just a note: I've never been a fanboy of either brand but I buy the price/performance sweet spot card all the time. I've owned 80% AMD, 20% Nvidia because of pricing. No single hardware issue on any. But only AMD has given me such headaches with software drivers.

 

[jwl3]

Sakurra:
Yeah well you can get a more powerful R9 280 for 170 bucks, so the GTX 760 is just not a good buy at the moment.



Yep, that's exactly the reason why I ended up buying another crud AMD card. The 280 was just that much cheaper than the equivalent card and I hate having to pay a premium for the same performance.

 

[cobra5000]

This is awesome and I love it! Now, I have so many options!

 

[Ninjawithagun]

@ Don Woligroski --> This article is very sloppy! You had better do your homework better next time. Dedicated Tom's Hardware members expect much better than this!! Your charts are outdated and incorrect beyond all belief. The impression is that you didn't take any time to research the market thoroughly and just slapped in single solutions at each price point. There are tons of options in each price category that will satisfy both Nvidia and AMD customers.

 

[Skorpius]

I believe the 970 GTX is the best bang for your dollar card if you are building a new PC for gaming.1-For my new unit what is nice is there are many factory over clocked models available, (well sold out atm, I can see why) and I am new to the PC building space so I really do not know (yet anyways) how to properly O.C. a GPU. 2-The companies making GTX970’s are going to release further bios/software updates down the line once they realize there is a TON more head room for more overclocking. 3-You can use the stock coolers as they are all looking quite impressive. When compared to the competing cards out there (R 290) they run way cooler which equals to less noise. Last but not least 4-HUGE savings compared to the R9 290 what I was originally planning to get.

You ask, HOW to you save even more money? I will explain; it’s all in the cooling. The stock coolers on all the R9 290's only work at stock speed. When mildly O.C. these cards run SUPER HOT which leads to TONS of noise and heat. The only way to really cool these cards right and get the maximum efficiency out of them is to water cool them. This leads to at least Two Hundred Dollars or more spent on water blocks, Radiators, fittings etc. Any ways I believe for people new to building there new units/rigs, whatever you want to call it; they need to keep this in mind when looking and comparing the full price of GPU they are planning to go with.

 

[larsoncc]

Like the other posters above me, I have to say - the hierarchy chart no longer represents performance metrics. I don't know who puts the 580 at the same level as a 760, or the 760 and 670 on different levels. It's messed up throughout the chart. It requires fixing and should be based on the actual performance that Tom's themselves publishes. As is, it may lead people to make unfortunate purchases!

 

[Sakkura]

I believe the 970 GTX is the best bang for your dollar card if you are building a new PC for gaming.1-For my new unit what is nice is there are many factory over clocked models available, (well sold out atm, I can see why) and I am new to the PC building space so I really do not know (yet anyways) how to properly O.C. a GPU. 2-The companies making GTX970’s are going to release further bios/software updates down the line once they realize there is a TON more head room for more overclocking. 3-You can use the stock coolers as they are all looking quite impressive. When compared to the competing cards out there (R 290) they run way cooler which equals to less noise. Last but not least 4-HUGE savings compared to the R9 290 what I was originally planning to get.

You ask, HOW to you save even more money? I will explain; it’s all in the cooling. The stock coolers on all the R9 290's only work at stock speed. When mildly O.C. these cards run SUPER HOT which leads to TONS of noise and heat. The only way to really cool these cards right and get the maximum efficiency out of them is to water cool them. This leads to at least Two Hundred Dollars or more spent on water blocks, Radiators, fittings etc. Any ways I believe for people new to building there new units/rigs, whatever you want to call it; they need to keep this in mind when looking and comparing the full price of GPU they are planning to go with.

What a load of rubbish. The R9 290 does not require water cooling at all. Just a good non-reference air cooler is more than plenty.

 

[RedJaron]

I believe the 970 GTX is the best bang for your dollar card if you are building a new PC for gaming.1-For my new unit what is nice is there are many factory over clocked models available, (well sold out atm, I can see why) and I am new to the PC building space so I really do not know (yet anyways) how to properly O.C. a GPU. 2-The companies making GTX970’s are going to release further bios/software updates down the line once they realize there is a TON more head room for more overclocking. 3-You can use the stock coolers as they are all looking quite impressive. When compared to the competing cards out there (R 290) they run way cooler which equals to less noise. Last but not least 4-HUGE savings compared to the R9 290 what I was originally planning to get.

You ask, HOW to you save even more money? I will explain; it’s all in the cooling. The stock coolers on all the R9 290's only work at stock speed. When mildly O.C. these cards run SUPER HOT which leads to TONS of noise and heat. The only way to really cool these cards right and get the maximum efficiency out of them is to water cool them. This leads to at least Two Hundred Dollars or more spent on water blocks, Radiators, fittings etc. Any ways I believe for people new to building there new units/rigs, whatever you want to call it; they need to keep this in mind when looking and comparing the full price of GPU they are planning to go with.

Methinks you're a little confused here. When using the reference clocks and cooler from AMD, yes, noise and temps can be a concern. But when getting a factory OC'd board from partner vendors using their own cooling solutions, the 290s run quite well with no worries of thermal throttling.

 

[akula2]

I've said this before and I'll say it again - why in the world are AMD cards so much cheaper than Nvidia? Is it generally accepted that Nvidia is a higher quality with better drivers? Look at this October issue - 90% of the recommended cards are AMD because of the price.

I'm quite interested in a particular AMD model - Firepro W9100

Many reviews say that it's beating Quadro K6000 cards (own a few of them for business tasks). I'm tempted to buy a Firepro and do the tests myself, but awaiting for the new Maxwell based Quadro cards. I'll wait for a few weeks to learn about a new Nvidia launch...

 

[Sakkura]

I've said this before and I'll say it again - why in the world are AMD cards so much cheaper than Nvidia? Is it generally accepted that Nvidia is a higher quality with better drivers? Look at this October issue - 90% of the recommended cards are AMD because of the price.

I'm quite interested in a particular AMD model - Firepro W9100

Many reviews say that it's beating Quadro K6000 cards (own a few of them for business tasks). I'm tempted to buy a Firepro and do the tests myself, but awaiting for the new Maxwell based Quadro cards. I'll wait for a few weeks to learn about a new Nvidia launch...

When it comes to FirePro and Quadro, look at benchmarks for the software you need to run. Performance varies wildly from one program to the next in a way you practically never see when it comes to games. Imagine an R9 290 getting 120 FPS in Crysis 3 and 10 FPS in Far Cry 3, while a GTX 970 got 15 FPS in Crysis 3 and 130 FPS in Far Cry 3.

 

[akula2]

When it comes to FirePro and Quadro, look at benchmarks for the software you need to run. Performance varies wildly from one program to the next in a way you practically never see when it comes to games. Imagine an R9 290 getting 120 FPS in Crysis 3 and 10 FPS in Far Cry 3, while a GTX 970 got 15 FPS in Crysis 3 and 130 FPS in Far Cry 3.

No, I don't spend several thousands of dollars on each high-end workstation solely by looking at review benchmark. I not only run my businesses, also into design/modeling/simulation tasks (review), write code using CUDA (review/tests/etc) and more. E.g., simulate 50,000 atoms to take a good measure of such workstation cards, and most of the hardware in the workstation in general. More system load brings out any hidden issues such as temperatures, cooling performance, noise annoyances etc. I custom design PCBs to solve a problem or some value addition or product (IP). I don't think any reviewer does the kind of things myself and my employees do in the real world applications (R&D and Manufacturing in Life Sciences and Medical Electronics).

My requirements revolves around Graphics (non-gaming) and Computational capabilities. But a straight forward single GPU doesn't work in all machines. E.g., I use Titan Blacks, Quadro K6000 or in SLI or Quadro K6000 + Tesla K40. It all depends on what I run (say double precision), which I take care of long before building a high-end workstation. Hence, I'm taking my good time to build ten x99 machines (five 5960Xs and five Xeon E5 v3s) in the upcoming holiday season. But I must see Maxwell based new Quadros and Teslas so that I can compare them with FirePro W9100 card. I don't have 'enough' info (from my channels) on their launch.....

 

[Sakkura]

No, I don't spend several thousands of dollars on each high-end workstation solely by looking at review benchmark.

So you decide on a whim instead? That's worse. The only way to gain any insight ahead of time is to check reviews.

 

[aberkae]

Just ordered a gtx 980 g1 planning to oc it to 1600/8000 available from newegg.

 

[siman0]

Well im just going to jump on ebay real quick and grab a few r9 290x cards for 250 usd. They are still fen crazy, and a decent card. Cant beet picking them up cheap in the used market.

 

[Eggz]

What's going on with the hierarchy chart?

You have the GTX970 below GTX780ti and the GTX980 in the same level with GTX780ti, while the real performance at page 8 clearly shows GTX970 on par with GTX780ti and GTX980 one level (at least) ahead.

Your own benchmark results are not on par with your hierarchy chart. Isn't the hierarchy chart based on Tom's Hardware benchmarks?

From the link you give, it seems like you're comparing the performance per dollar chart. That has no bearing on pure performance. If you get an old VooDooFX for free, it will have infinite performance per dollar, but it'll still be slow as mud.

Take a look at the relevant part of the 970 analysis, and you'll see it isn't as fast as the 980 and 780 ti:

http://www.tomshardware.com/reviews/nvidia-geforce-gtx-980-970-maxwell,3941-16.html

 

[Brogan]

^^ That's why I bought a 980 G1. (which are out of stock again on NE)

My Unigine benchmarks freakin tripled from my old card. Woof

Inquisition is gonna look siiiiiiii.......

 

[Memnarchon]

What's going on with the hierarchy chart?

You have the GTX970 below GTX780ti and the GTX980 in the same level with GTX780ti, while the real performance at page 8 clearly shows GTX970 on par with GTX780ti and GTX980 one level (at least) ahead.

Your own benchmark results are not on par with your hierarchy chart. Isn't the hierarchy chart based on Tom's Hardware benchmarks?

From the link you give, it seems like you're comparing the performance per dollar chart. That has no bearing on pure performance. If you get an old VooDooFX for free, it will have infinite performance per dollar, but it'll still be slow as mud.

Take a look at the relevant part of the 970 analysis, and you'll see it isn't as fast as the 980 and 780 ti:

http://www.tomshardware.com/reviews/nvidia-geforce-gtx-980-970-maxwell,3941-16.html

Well it seems that you don't understand how the chart at p8 works.

As you can see the red bar is the performance that the card has, compared to 295X2 at 1920x1080.
The black bar does the same but for 3840x2160.
The blue bar is an average of black and red.
These have nothing to do with price.
Price is the yellow dots in the chart. The yellow line connects the dots in order to see how much the price goes up for each price segment.
So you can make the calculations and see what is a good buy and what's not.
And here was the confusion. At this chart, GTX970 and GTX780ti have both the same performance.
But Cleave already answered this in a post before you.
These are outdated benchmarks and we shall see an updated version Soon™.

 

[Eggz]

^^ I understand the performance per dollar chart. I also understand that it summarizes of a whole lot of information. Maybe there was a typo or something. But considering the underlying information as a whole tends to show that the 970 isn't as fast as the 980 or 780 ti. That's pretty clear. . . .

My point is that the hierarchy chart seems fine. The comment I quoted above just seemed overstated, since it appeared to suggest that the hierarchy chart was inaccurate with respect to placement of the 970, 980, and 780 ti. It's not. Except for that red bar, Tom's articles are consistent with that too.

 

[Memnarchon]

^^ I understand the performance per dollar chart. I also understand that it summarizes of a whole lot of information. Maybe there was a typo or something. But considering the underlying information as a whole tends to show that the 970 isn't as fast as the 980 or 780 ti. That's pretty clear. . . .

My point is that the hierarchy chart seems fine. The comment I quoted above just seemed overstated, since it appeared to suggest that the hierarchy chart was inaccurate with respect to placement of the 970, 980, and 780 ti. It's not. Except for that red bar, Tom's articles are consistent with that too.

Nah it wasn't a typo.
Cleeve already answered, as I said, that charts at p8 show outdated benchmarks.
They will soon update the p8 with the new benchmarks.
We might get a review for the new benchmarks as well.

 

[Matrony]

I was wondering why the new apu's from amd are not in this chart? you have the intel cpu's graphics on the chart, but not the amd apu's?