How long since AMD was on top or within 5% of the best card?

I sold my 7970's and purchased 2x GTX 680's because xfire was a POS at that time. The 680's in SLI were much smoother.

The good news was xfire improved with the 290's and I sold the 680's to get a pair of those.

The 7970s got much better in Crossfire with subsequent drivers -- AMD had just gotten hammered in the press for their frame-pacing issues with the HD6000-series and had committed to fixing them, but those fixes weren't available upon the 7000-series release.

I rarely pick up cards upon release, but while holding a pair of HD6950s and looking at the HD79x0, the GTX6x0 cards looked much better running in pairs during initial reviews. As noted, the HD7970 did get much better with time, you know, except for the heat ;).
 
The 7970s got much better in Crossfire with subsequent drivers -- AMD had just gotten hammered in the press for their frame-pacing issues with the HD6000-series and had committed to fixing them, but those fixes weren't available upon the 7000-series release.

I rarely pick up cards upon release, but while holding a pair of HD6950s and looking at the HD79x0, the GTX6x0 cards looked much better running in pairs during initial reviews. As noted, the HD7970 did get much better with time, you know, except for the heat ;).
To be fair they were on a 384bit bus, had an extra 1gb of RAM and were compute monsters at the time. If you looked outside of gaming, these cards had good value.
 
To be fair they were on a 384bit bus, had an extra 1gb of RAM and were compute monsters at the time. If you looked outside of gaming, these cards had good value.

Without a doubt- AMD has been focusing hard on building compute cards first, and probably has sacrificed gaming prowess as a result. Nvidia appeared to do the same up until the GTX600-series, and have since started releasing maxed-out gaming parts a la the 1080Ti. That's something I'd like to see from AMD, which I feel they're capable of doing and perhaps have the internal capital to pull of with Big Navi.
 
The 7970s got much better in Crossfire with subsequent drivers -- AMD had just gotten hammered in the press for their frame-pacing issues with the HD6000-series and had committed to fixing them, but those fixes weren't available upon the 7000-series release.

I rarely pick up cards upon release, but while holding a pair of HD6950s and looking at the HD79x0, the GTX6x0 cards looked much better running in pairs during initial reviews. As noted, the HD7970 did get much better with time, you know, except for the heat ;).

power consumption 680 vs 7970 was nearly the same

Unless you mean OEM cooler vs OEM cooler
 
But also 3 to 5dB(A) louder, which is at least twice as loud: https://www.anandtech.com/show/5699/nvidia-geforce-gtx-680-review/19

Also, seeing the GTX680 faster than the HD7950 Boost in 2018, however that compares to an HD7970: https://www.techspot.com/article/1588-geforce-gtx-680-revisit/


Yeah... my memory was correct, and I'd definitely still take the GTX680 ;)

I'm not getting baited on this discussion. It is what it is. If you love Nvidia, by all means, go for it because I know you have a passionate hate for anything AMD.


Fact is the 7970 was the better card overall, the R9 290x (For gaming) was the better overall card. (I owned a Kepler Titan) The newer architectures were flops. Big Navi is still not here and the drivers are still bad.
Please do not respond to my message because I am not going any further with you.
 
If you love Nvidia, by all means, go for it because I know you have a passionate hate for anything AMD.

Neither of these are true.

Fact is the 7970 was the better card overall

...not at all what Anandtech found at launch, or Techspot six years later, as linked above...

R9 290x (For gaming) was the better overall card

Again, not really. Tossup with Nvidia at release, but now 50w more power draw for 6dB(A) more noise under load: https://www.anandtech.com/show/7457/the-radeon-r9-290x-review/19


I'm not getting baited on this discussion.
Please do not respond to my message because I am not going any further with you.

It's not a discussion, and you're not being baited -- you posted a personal accusation and you're now dodging facts that disagree with your position.


I'm not looking at either company with rose-colored glasses. I prefer to see them for what they are, and what they can do for me.
 
At least he admits the 7970 is faster than the 680, then and even moreso now.
Some people will pay a premium for a quieter and less power hungry card even if it might be slightly slower.

buuuuut there is such a thing as undervolting if you're willing to tweak it (like most people here are willing to do)
 
The 680 launched later than the 7970 but it was faster...that's not the case post launch especially with the 2GB vs 3GB ram buffer. The 7970 was the better card in the long run imo, and 7970 overclocked better. They were both good products. Idiotincharge always claims neutrality while taking the side of Nvidia, what's new?

idiot, you can't use a comparison of a 680 vs 7950 c'mon dude.
 
It was the Bat Man era HD 5870 / HD 5850 that was the real movement as Nvidia was worried at the time and CrossFire was the shit on my x58 eVGA 3 Way SLi Board with i7 - 930 in triple channel 1600Mhz .. as I ran dual HD 5850's at the time .. but cut my teeth with Nvidia SLi in Twin eVGA 7900 GTO's
 
This is utterly false for the 7970.
Basically all 7970s could do 1125Mhz on the core (Stock was 925 im pretty sure).
The 680 came out like 6 months later and just barely beat a stock 7970 and had a 1/3rd less VRAM.
Whoever owned a 7970 didn't buy a 680 because it was faster at all.
The 7970 OC'd like a boss, and man, them legs........

Who here today would take a 680 over a 7970 for gaming???
................
That's what I thought.

Back on topic, I agree with the 290x, although the FuryX was pretty close to the 980ti (stock) as long as your never got close to the 4GB VRAM limit lol. OC is a WHOLE nutha story though.

I still have a 290x PCS+ that runs stock at 1040 or 1050MHz can't remember. To the guy that said he had 20 290x's and none were stable @1000MHz, I say: You're doing it wrong lol! You're problem is not the Video cards

I remember spanking the 680/770s with mah 7970s all over ocn. Oh those old days...

https://www.3dmark.com/fs/3470060
https://www.3dmark.com/fs/3004700
 
I still have my almost new Sapphire Tri X R9 - 280 rebrand with the box but it's not Batman like .
 
The last time AMD had a clear performance advantage over nvidia, was on the 9700/9800 era (more so the former). No OC needed no voltage tweaking, no memory tweaking. Ever since, they have traded blows a few times but for the most part nvidia has dominated.
No.
X1950XTX clearly defeated 7900GTX.
 
I still have my almost new Sapphire Tri X R9 - 280 rebrand with the box but it's not Batman like .
I've got the tri x fury, runs great with a small undervolt. Now I run a fury nano in my itx... Still waiting for something to come out that is worth replacing it with.
 
No.
X1950XTX clearly defeated 7900GTX.

That's not AMD, but rather the ATI we all loved and miss ;)

AMD entered the GPU scene with the purchase of ATI and the release of the 2900XT, the poorest release since ATI was trying to use Rage 128 graphics cards to compete with 3Dfx and Nvidia.

To give credit to AMD, they've yet to repeat that failure, which much resembled Nvidia's FX5800 release in that neither architecture was really targeting the software being developed.


Idiotincharge always claims neutrality while taking the side of Nvidia, what's new?

I'm not a fan of fanboism. I'm a fan of performance and fitting alternatives to solutions.

You'll note, if you were to review my post history rather than sharing your uninformed impressions, that I recommend those products that provide the best fit regardless of vendor, and that lately on the CPU side and on the mid-range GPU side, that's been almost exclusively AMD.

That's not because I have some love or dislike for the company, but because I'm more interested in the best currently available solution for a given requirement.
 
I'm also going to say it hasn't been since they 290X launched. Driver tweaks later on boosted performance measurably, and it became the best bang for the buck in it's post-price drop segment.

I'm hopeful for Navi2, but I won't be holding my breath.
 
That's not AMD, but rather the ATI we all loved and miss ;)

AMD entered the GPU scene with the purchase of ATI and the release of the 2900XT, the poorest release since ATI was trying to use Rage 128 graphics cards to compete with 3Dfx and Nvidia.

To give credit to AMD, they've yet to repeat that failure, which much resembled Nvidia's FX5800 release in that neither architecture was really targeting the software being developed.

I know this but my point to stoly was the the 9700/9800 was certainly not it. There's been several generations after that time that ATI/AMD had the faster solution.
 
I know this but my point to stoly was the the 9700/9800 was certainly not it. There's been several generations after that time that ATI/AMD had the faster solution.

I think it's a difference in perspective; there's been periods of a handful of months where AMD released earlier than Nvidia and provided a challenge, but when comparing generation to generation overall and between the release of both companies' products for that generation and the release of the next generation, AMD has only ever gotten close.

But yes, for a few of the occasions where they've released their top-end part for a new generation a few months ahead of Nvidia, they've managed to edge slightly ahead.
 
I think within 5% of the best is quite a low bar. Manufacturers have months to target a performance envelope and meet or just pip the existing market leader. The real question is when did AMD last have clear market leadership at the top (20% or more) in single card performance.
 
I think within 5% of the best is quite a low bar. Manufacturers have months to target a performance envelope and meet or just pip the existing market leader. The real question is when did AMD last have clear market leadership at the top (20% or more) in single card performance.

Well, that, and how long?

With AMD, you pretty much have to plan for the three to six months afterward to get the cooling, BIOS, and drivers all lined up. I'm always happy when there's an exception but those are pretty rare.

So, when's the last time they exceeded their competition by a significant amount and got everything else lined up for more than a month or two?

I'm inclined to believe they haven't; when I've bought AMD it hasn't been because they were the fastest, but because they were a better value for the given requirement.

What do y'all think?
 
The ref R- 290 I bought just for Battlefield at the timing was worth far more to bitcoin and I sold it and made enough to buy my 280 while I waited for non-ref 290x which I bought way cheaper $269 with a game .. so the 290 made enough profit to pay for a 280 and 290x which I still have both of them .
 
AMD cards age better than Nvidia cards, when the 290x came out it wasn't near the GTX 780, but it did get better with time due to drivers, and now it is. I'd say the last time AMD was on top is the HD 7970, specially the GHz edition.
 
AMD cards age better than Nvidia cards

This was more true when AMD was releasing pretty horrific drivers -- that's not really the case today, and further, by the time AMD got around to really fixing up their drivers, Nvidia had new hardware. So if you waited for AMD to get the drivers figured and their partners to ship decent coolers -- you could just get something from Nvidia that was better right out of the gate.

Today, the 'fine wine' is mostly in the past, though, or at least, I hope it is. You shouldn't have to wait the better part of a year or longer for the vendor to squeeze full performance out of a part, or worse, wonder if they're ever going to get their software stack optimized in the first place. And since AMDs drivers have gotten much better in the past few years, it's not really prudent to consider their products on the basis of performance scaling up over time.
 
9700 pro - never forget the day that hardware outpaced softwear for a moment
 
9700 pro - never forget the day that hardware outpaced softwear for a moment

That was my first ever ATI card and it was an absolute beast. I remember all my mates around me at a lan and all there jaws were on the ground watching me play the original Farcry.
 
I bought the 9700 pro to play Half Life 2 (and the leak ..sh..) that was a leap forward in both software and hardware. It would be awesome if AMD did it again for HL thr.. err Alyx.
 
I bought the 9700 pro to play Half Life 2 (and the leak ..sh..) that was a leap forward in both software and hardware. It would be awesome if AMD did it again for HL thr.. err Alyx.

Same!

Well, actually Counter-Strike: Source, that came with the purchase of Half Life 2, but I also played Half Life 2.
 
This is utterly false for the 7970.
Basically all 7970s could do 1125Mhz on the core (Stock was 925 im pretty sure).
The 680 came out like 6 months later and just barely beat a stock 7970 and had a 1/3rd less VRAM.
Whoever owned a 7970 didn't buy a 680 because it was faster at all.
The 7970 OC'd like a boss, and man, them legs........

Who here today would take a 680 over a 7970 for gaming???
................
That's what I thought.

Back on topic, I agree with the 290x, although the FuryX was pretty close to the 980ti (stock) as long as your never got close to the 4GB VRAM limit lol. OC is a WHOLE nutha story though.

I still have a 290x PCS+ that runs stock at 1040 or 1050MHz can't remember. To the guy that said he had 20 290x's and none were stable @1000MHz, I say: You're doing it wrong lol! You're problem is not the Video cards

I find this funny....because the 7970's went up against Kepler's midrange die...at launch the 7970's were so bad the were fighting with GK104-400-A2 (Kepler's mid range die).
If AMD had been close, they would have had to be within 5% of the GK110 dies, not the GK104 dies.

The problem when people look at PR names and not look under the hood and see what is really going on.

EDIT:
The QUADRO line names showed the "proper" Kepler line-up:
Quadro K4000 GK106
Quadro K5000 GK104
Quadro K6000 GK110
 
Last edited:
I find this funny....because the 7970's went up against Kepler's midrange die...at launch the 7970's were so bad the were fighting with GK104-400-A2 (Kepler's mid range die).
If AMD had been close, they would have had to be within 5% of the GK110 dies, not the GK104 dies.

The problem when people look at PR names and not look under the hood and see what is really going on.

EDIT:
The QUADRO line names showed the "proper" Kepler line-up:
Quadro K4000 GK106
Quadro K5000 GK104
Quadro K6000 GK110

Oh boy.......

Where were these GK110s when the 7970 launched in Dec 2011????????
The GTX 680 was Nvidia's fastest GPU and it came out in Mar 2012. They could name it whatever they wanted and that doesn't change the fact that the GTX680 (GK104) was the fastest they had at the time regardless of whether it was considered a "mid range die".

You could literally say the same thing about Tahiti, it was only the midrange die with a die size of 352mm2. Hawaii was 438mm2 and would be considered the "high end" die. They literally could have done the same type of naming as Nvidia (like GCN-104 for 7970 and GCN-110 for 290x) and your "point" would be non-existent.

What you mean to say is look under the hood at the die size as that will tell you if the chip is big or not.
If "looking under the hood" is only looking at code names, then you need to change that.

By the time the 780Ti 3GB came out (Nov 2013), it had to deal with the 290x, and we all know how much of a beast it was and still is.
 
Last edited:
That was my first ever ATI card and it was an absolute beast. I remember all my mates around me at a lan and all there jaws were on the ground watching me play the original Farcry.

It was the first time I could check 8xAA and play a game - it was glorious.
 
Oh boy.......

Where were these GK110s when the 7970 launched in Dec 2011????????
The GTX 680 was Nvidia's fastest GPU and it came out in Mar 2012. They could name it whatever they wanted and that doesn't change the fact that the GTX680 (GK104) was the fastest they had at the time regardless of whether it was considered a "mid range die".

You could literally say the same thing about Tahiti, it was only the midrange die with a die size of 352mm2. Hawaii was 438mm2 and would be considered the "high end" die. They literally could have done the same type of naming as Nvidia (like GCN-104 for 7970 and GCN-110 for 290x) and your "point" would be non-existent.

What you mean to say is look under the hood at the die size as that will tell you if the chip is big or not.
If "looking under the hood" is only looking at code names, then you need to change that.

By the time the 780Ti 3GB came out (Nov 2013), it had to deal with the 290x, and we all know how much of a beast it was and still is.

Do we need to do this every time?
NVIDIA's SKU's are following a pattern:

Low range: GTX 450 (GF106, 128 bit mem-bus, 238 mm2) -> GTX 650(GK106, 128 bit mem-bus, 118 mm2) -> GTX 650(GM106, 128 bit mem-bus, 227 mm2) -> GTX 1050(GP107, 128 bit mem-bus, 132 mm2) -> GTX 1650(TU116, 128 bit mem-bus, 284 mm2)

Mid range: GTX 460 (GF104, 256 bit mem-bus, 332 mm2) -> GTX 680(GK104, 256 bit mem-bus, 294 mm2) -> GTX 980 (GM104, 256 bit mem-bus, 398 mm2) -> GTX 1070(GP104, 256 bit mem-bus, 314 mm2) -> RTX 2070(TU104, 256 bit mem-bus, 545 mm2)

High end: GTX 480(GF100, 384 bit mem-bus, 529 mm2) -> GTX 780 Ti (GK110, 384 bit mem-bus, 561 mm2) -> GTX 980 Ti (GM200, 384 bit mem-bus, 601 mm2) -> GTX 1080 Ti(GP102, 384 bit mem-bus, 471 mm2) -> RTX Titan (TU102, 384 bit mem-bus, 754 mm2)

That was Fermi -> Kepler -> Maxwell -> Pascall -> Turing

That you have problems telling a high end die from a mid range die of a given architechture is a bad argument.

Now tell me when the last time NVIDIA had a high end SKU with:
Gx1x6 SKU name
256 bit Mem bus
<450 mm2

PR names are useless, you need to look at the die...in the BIG picture...otherwise fancy PR will confuse you...like now.
 
Do we need to do this every time?
NVIDIA's SKU's are following a pattern:

Low range: GTX 450 (GF106, 128 bit mem-bus, 238 mm2) -> GTX 650(GK106, 128 bit mem-bus, 118 mm2) -> GTX 650(GM106, 128 bit mem-bus, 227 mm2) -> GTX 1050(GP107, 128 bit mem-bus, 132 mm2) -> GTX 1650(TU116, 128 bit mem-bus, 284 mm2)

Mid range: GTX 460 (GF104, 256 bit mem-bus, 332 mm2) -> GTX 680(GK104, 256 bit mem-bus, 294 mm2) -> GTX 980 (GM104, 256 bit mem-bus, 398 mm2) -> GTX 1070(GP104, 256 bit mem-bus, 314 mm2) -> RTX 2070(TU104, 256 bit mem-bus, 545 mm2)

High end: GTX 480(GF100, 384 bit mem-bus, 529 mm2) -> GTX 780 Ti (GK110, 384 bit mem-bus, 561 mm2) -> GTX 980 Ti (GM200, 384 bit mem-bus, 601 mm2) -> GTX 1080 Ti(GP102, 384 bit mem-bus, 471 mm2) -> RTX Titan (TU102, 384 bit mem-bus, 754 mm2)

That was Fermi -> Kepler -> Maxwell -> Pascall -> Turing

That you have problems telling a high end die from a mid range die of a given architechture is a bad argument.

Now tell me when the last time NVIDIA had a high end SKU with:
Gx1x6 SKU name
256 bit Mem bus
<450 mm2

PR names are useless, you need to look at the die...in the BIG picture...otherwise fancy PR will confuse you...like now.

You completely missed the point about the GK104 being the fastest and highest end chip Nvidia had for gaming for almost 2 years.
Whether its considered a mid range part or high end part is totally irrelevant.
What is relevant is actual performance.
The only thing that those numbers tell you is that there is going to be a bigger faster chip on the same arch. In some cases (680 to 780Ti), it will be a while...

Also:
gtx 1080 = GP104
rtx 2080 = TU104
 
Last edited:
You completely missed the point about the GK104 being the fastest and highest end chip Nvidia had for gaming for almost 2 years.
Whether its considered a mid range part or high end part is totally irrelevant.
What is relevant is actual performance.
The only thing that those numbers tell you is that there is going to be a bigger faster chip on the same arch. In some cases (680 to 780Ti), it will be a while...

That does alter the fact that NVIDIA has a pattern...and their midrange SKU was enough to pick a fight with with AMD's high end SKU.
And GK110 was not absent for 2 years...because NVIDIA could fend for it self with the GK104 in consumer space...they could get the higher margins in enterprise.
Look at the release date for the QUADRO K6000 (GK110)....do I need to explain economics 101 for you too?

When your mid range SKU (GK104) is all you need to stay competitive...ot makes perfect business sense to have your high end SKU(GK110) haul in the enterprise margins.
 
That does alter the fact that NVIDIA has a pattern...and their midrange SKU was enough to pick a fight with with AMD's high end SKU.
And GK110 was not absent for 2 years...because NVIDIA could fend for it self with the GK104 in consumer space...they could get the higher margins in enterprise.
Look at the release date for the QUADRO K6000 (GK110)....do I need to explain economics 101 for you too?

When your mid range SKU (GK104) is all you need to stay competitive...ot makes perfect business sense to have your high end SKU(GK110) haul in the enterprise margins.

Now you are starting to make sense at least.
For the purposes of gaming, the enterprise market is pretty irrelevant unless you buy Quadros for gaming.
Knowing that Nvidia has a higher end GPU somewhere that is almost unreachable due to price (Quadro) does nothing for us consumers, other than let us know that we're being milked.
Stuff like that makes me mad because we "know" that something faster is coming, but they're not going to release it yet until competition catches up, and if not, they may or may not release it at a highly inflated price (2080Ti), but sadly that's standard business practice for Duopolies.

My point is, who cares who has what in the works, how big or how small, performance is the thing that makes a GPU high end or not. I'm pretty sure everyone here understands how Nvidia's codenames work
 
Now you are starting to make sense at least.
For the purposes of gaming, the enterprise market is pretty irrelevant unless you buy Quadros for gaming.
Knowing that Nvidia has a higher end GPU somewhere that is almost unreachable due to price (Quadro) does nothing for us consumers, other than let us know that we're being milked.
Stuff like that makes me mad because we "know" that something faster is coming, but they're not going to release it yet until competition catches up, and if not, they may or may not release it at a highly inflated price (2080Ti), but sadly that's standard business practice for Duopolies.

My point is, who cares who has what in the works, how big or how small, performance is the thing that makes a GPU high end or not. I'm pretty sure everyone here understands how Nvidia's codenames work

Guess why I never owned a GTX 680?
People fall for PR names all the times (and get angry when they spot rebrands)...but nothing seems to anger certain people more than the GK104 was all NVIDIA needed to compete with AMD's highend...I don't care what the PR name of a SKU is...I care for the SKU name....because that gives me the full story.
 
Guess why I never owned a GTX 680?
People fall for PR names all the times (and get angry when they spot rebrands)...but nothing seems to anger certain people more than the GK104 was all NVIDIA needed to compete with AMD's highend...I don't care what the PR name of a SKU is...I care for the SKU name....because that gives me the full story.

People buy the most performance that they can afford most of the time. The few people that want the best if they can afford it will wait for the xx0/x1x/Titan variants as they know they will be the fastest for the longest time usually.
People who wanted the fastest GPU Nvidia had in 2012 bought the GTX 680. Companies bought Quadro.
You waited because something bigger was coming, which is the case ALL of the time, now how long it takes to make it worth it is up to the person. Apparently you were fine waiting a while until something faster than the GTX680 came out.
Thanks for letting us know why you didn't purchase a GTX 680.
 
Back
Top