Radeon HD 7990 versus GeForce GTX TITAN Review @ [H]

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,532
Radeon HD 7990 versus GeForce GTX TITAN Review - We follow-up with a look at how the $999 GeForce GTX TITAN compares to the new $999 AMD Radeon HD 7990 video card. What makes this is unique is that the GeForce GTX TITAN is a single-GPU running three displays in NV Surround compared to the same priced dual-GPU CrossFire on a card Radeon HD 7990 in Eyefinity.
 
Not much surprise here. Crazy that its $100 more than 2 7970's while being slower...
 
Yes, the price is a bit crazy....however, since the punt paid for mine, I'm happy and it's better than the crashes and shit, I was getting with the 690...25x16 is enuff res for me and the Titan does it easy..BTW, nice review v interesting data..
 
Sorry if I missed it but in the review was it (2) 7970's or one being evaluated? Also there wasn't a link to the discussion. If I missed if it was 2 or 1 I apologize.
 
Here you go:

http://www.hardocp.com/article/2013...ersus_geforce_gtx_titan_review/7#.UYc8GbUTGoM

1367219788M4TNFhfNAc_7_1.gif


Sorry if I missed it but in the review was it (2) 7970's or one being evaluated? Also there wasn't a link to the discussion. If I missed if it was 2 or 1 I apologize.
 
Thank you! I assumed it was CF, but I wanted to be sure before I link the article to my nephew. :)
 
It's a bit confusing because, at least under the Crysis 3 section, it says one 7970 beats out 680 SLI. I assume it's CFX but...yeah.
 
It's a bit confusing because, at least under the Crysis 3 section, it says one 7970 beats out 680 SLI. I assume it's CFX but...yeah.

Yeah, for a long time I thought they were burying the lead there, hiding some sort of super Hulk HD7970 since in all the graphs it didn't mention CF.

Phew.
 
The Dual GPU vs single GPU is a bit weird since the card that is called Tesla and has the price of almost triple Titan is made for a different market. Also the design has next to nothing to do with gaming in general it is just stacking of compute units until they ran out of space.

Titan is only here because of Nvidia slump in sales for Tesla (or maybe they made to many chips who knows). While with gaming Titan may do okay, the market for this card is so limited that there not making them anymore, can't say that about the AMD cards.
 
Sorry if I missed it but in the review was it (2) 7970's or one being evaluated? Also there wasn't a link to the discussion. If I missed if it was 2 or 1 I apologize.

It would have been best to clearly label the cards in the actual review vs pointing them to an older review to compare results just to determine what is going on. Also, the link to the discussion is still not functional.


the graph's didn't have it saying crossfire but i mean ffs they said 7970 ghz edition crossfire a million times in that review, not exactly sure how people missed it. if you're not actually reading whats being written in the review then maybe you should be.


Not much surprise here. Crazy that its $100 more than 2 7970's while being slower...

single card vs dual cards, 100 dollars is worth it for the slightly less performance you get, especially for the people that can't run crossfire.


The Dual GPU vs single GPU is a bit weird since the card that is called Tesla and has the price of almost triple Titan is made for a different market. Also the design has next to nothing to do with gaming in general it is just stacking of compute units until they ran out of space.

Titan is only here because of Nvidia slump in sales for Tesla (or maybe they made to many chips who knows). While with gaming Titan may do okay, the market for this card is so limited that there not making them anymore, can't say that about the AMD cards.

if i remember correctly before it's release there were rumors AMD was going to be releasing the 8k series in Q1. my guess is Titan was a "oh shit we don't have anything to release" type card, but yeah i agree with you. though i'd laugh if the release of the titan backfires and the 700 series doesn't meet the expectations of even the titan. i guess all we can do is sit and wait.
 
Last edited:
From the article:

In all cases, Radeon HD 7970 GHz Edition CrossFire from two separate video cards was still the fastest, and best gameplay experience, for a lesser price.

Does "best gameplay experience" include smoothness? In other words, does the article imply that 7970 crossfire is comparable in gameplay fluidity as the Titan and 680 SLi configurations? Please be explicit in your reply, not vague.
 
While it's impressive that a single GPU is keeping up with a pair of 680's and 7970's, I'm most impressed by how the 7970's performed. Looks like AMD have really nailed their drivers with this current batch. Considering they're under $400 a piece now, I think that's a bigger story than the Titan, at least IMO.
 
Great article. Can I nitpick one point on Page 7?

Just think, GeForce GTX TITAN is within 30% performance of GTX 680 SLI and Radeon HD 7990, but overall saving 165W compared to HD 7990 and 152W compared to GTX 680 SLI.

Here you switch from a percentage value to absolute values. This is very confusing. I've taken the liberty of making the power figures more presentable (to me, anyway, YMMV).

Given the base of 67W we get the figures below:

Code:
Card                         Idle / Delta      Load / Delta
No GPU                        67 / +0          n/a
Radeon 7990                  148 / +81         536 / +469
GTX 680 SLI                  112 / +45         523 / +456
Radeon 7970 GHz CF           145 / +78         667 / +600
Titan                         96 / +29         371 / +304

So we can see that, excluding the 67W system power, at load you have to use 54% more power with the 7990, 50% more power with the 680 SLI, and 97% more power with the 7970 GHz CF.
 
the Radeon HD 7970 Ghz CF is labelled as Radeon HD 7970 Ghz edition in the performance charts. please change that. AMD's frame pacing driver will make the HD 7970 Ghz CF even more attractive from a price performance point of view. july can't come soon enough.
 
I am still wondering about the smoothness issue with the ATI cards. I have read too many reviews (even here on HARDOCP) that ATI Crossfire setups don't have the smoothness of the Nvidia SLI setups.

For me - I am debating my next upgrade. I am still sitting on an older CPU(i7 930). Should I go ahead and get a new MB/CPU or get a 2nd 680 and go SLI?
 
With the entirely objective "playability" metric used here, smoothness should always be taken into account. I recall several reviews where the author claimed to have had to reduce quality settings because of smoothness issues, so the same should have been done here if necessary.
 
I am still wondering about the smoothness issue with the ATI cards. I have read too many reviews (even here on HARDOCP) that ATI Crossfire setups don't have the smoothness of the Nvidia SLI setups.

For me - I am debating my next upgrade. I am still sitting on an older CPU(i7 930). Should I go ahead and get a new MB/CPU or get a 2nd 680 and go SLI?

it is interesting it's not mentioned. Kyle and Brent have both been saying for a long time that CF simply was not as smooth. The article would seem to encourage someone to buy two 7970's if they were shopping today. Given the recent frame metering topic, I would think that would be a huge negative in recommending a CF product right now.
 
CF at the moment should not be an option under any circumstance as it is currently.

That being said, the beta cat drivers look promising and my statement above could and hopefully will be rendered completely inaccurate within the next few months.
 
CF at the moment should not be an option under any circumstance as it is currently.

Not surprising you should say that.
You invested $1,000 in a Titan, so you have a vested interest in believing that you would not have been better off with a Crossfire system. Any information to the contrary you'll probably reject/ignore, unless you're completely objective and have no brand loyalty (rare around [H]).
 
CF at the moment should not be an option under any circumstance as it is currently.

That being said, the beta cat drivers look promising and my statement above could and hopefully will be rendered completely inaccurate within the next few months.

Really? Im running 2 6970s in one machine in 3 7970s in another and the only game that gives me trouble is Farcry 3. 3 7970s in tri-fire for $1050 smokes one Titan SC at $1050 there is no arguing that at all.
 
CF at the moment should not be an option under any circumstance as it is currently.

That being said, the beta cat drivers look promising and my statement above could and hopefully will be rendered completely inaccurate within the next few months.

From the article.

"We are using Catalyst 13.5 Beta2 for AMD Radeon HD 7990 and 7970 GHz Edition CrossFire."

So I guess the new drivers have finally satisfied the [H] crew when it comes to smoothness. If it had been an issue they would have mentioned it... a lot. Good news.

Oh yea, and no Gold Award for the 7970GE CF setup? *tisk* It's what I would recommend to a customer for a dual GPU setup as of right now. I really need to get another 7970 and do some of my own testing.
 
I'd just like to say thank you for all the [H]ard work the [H] crew put into this article.
 
From the article:

In all cases, Radeon HD 7970 GHz Edition CrossFire from two separate video cards was still the fastest, and best gameplay experience, for a lesser price.

Does "best gameplay experience" include smoothness? In other words, does the article imply that 7970 crossfire is comparable in gameplay fluidity as the Titan and 680 SLi configurations? Please be explicit in your reply, not vague.

Could you please reply to post #17 in this thread?
Thanks!

Since he's probably catching up on sleep after exiting mad scientist mode, I can attempt to reply for him.

When we (and I mean all video card editors at [H] collectively) talk about the "best gameplay experience", it means the highest playable settings that we can set for each game while still maintaining a smooth and enjoyable playing experience. We do capture fps data, but that is more to have something to graph to show our readers than it is something that we base our judgment on - we don't even keep the frame counter on screen when we are deciding playability. If the experience isn't smooth, regardless of frame rate, we will lower settings until it is smooth. We aren't running benchmarks, we are actually playing the game. This has been how [H] has done reviews for many years and has addressed the issue of smoothness long before the whole frame time debacle ever started.

Therefore, when Brent makes the statement that you quoted, he is speaking from the perspective of smoothness. Yes, on the publically available (and even non-publically available drivers), Crossfire with Eyefinity is NOT as smooth at a given frame rate as is SLI, however, since the Radeon cards usually deliver a faster frame rate, those extra frames help to smooth out the experience to the end user/gamer.

If you've got some time for some reading, feel free to look back across our Crossfire/SLI articles from the past few years where you will see what I have described being documented and written about by us....
 
Since he's probably catching up on sleep after exiting mad scientist mode, I can attempt to reply for him.

When we (and I mean all video card editors at [H] collectively) talk about the "best gameplay experience", it means the highest playable settings that we can set for each game while still maintaining a smooth and enjoyable playing experience. We do capture fps data, but that is more to have something to graph to show our readers than it is something that we base our judgment on - we don't even keep the frame counter on screen when we are deciding playability. If the experience isn't smooth, regardless of frame rate, we will lower settings until it is smooth. We aren't running benchmarks, we are actually playing the game. This has been how [H] has done reviews for many years and has addressed the issue of smoothness long before the whole frame time debacle ever started.

Therefore, when Brent makes the statement that you quoted, he is speaking from the perspective of smoothness. Yes, on the publically available (and even non-publically available drivers), Crossfire with Eyefinity is NOT as smooth at a given frame rate as is SLI, however, since the Radeon cards usually deliver a faster frame rate, those extra frames help to smooth out the experience to the end user/gamer.

If you've got some time for some reading, feel free to look back across our Crossfire/SLI articles from the past few years where you will see what I have described being documented and written about by us....

David,

I really appreciate the quick and informative reply! The clarification on testing procedures helped. I guess I'm just a little confused; wouldn't a higher frame rate only make the stuttering more apparent, since stuttering is essentially very brief periods of zero FPS (or dropped frames)? I would think going from 70 fps to zero occasionally would be MORE noticeable than 50 fps to zero occasionally. But maybe I just don't understand it properly.
 
Good grief, I don't typically get into it but could you let your bias show any more here Brent?

If this was the same article with an AMD GPU you'd be going on about how it isn't worth it because of the lower settings, framerates, and price-performance. I know because I've seen it. Because it's Nvidia you're ranting about how "impressive" it is that it's ONLY 25-35% slower (frequently with lower AA or other in-game settings).

Give it a rest, we get that you prefer Nvidia but can you at least try to be objective with these things? The Titan got hit by a lightning bolt in this and you're more or less telling us that at least it didn't get fully immolated.
 
David,

I really appreciate the quick and informative reply! The clarification on testing procedures helped. I guess I'm just a little confused; wouldn't a higher frame rate only make the stuttering more apparent, since stuttering is essentially very brief periods of zero FPS (or dropped frames)? I would think going from 70 fps to zero occasionally would be MORE noticeable than 50 fps to zero occasionally. But maybe I just don't understand it properly.

The higher frame rate causes the gaps to be smaller between the frame delivery which can help it overcome the stuttering that you feel when playing. I.e. at 30fps, you might have a .2 second gap in a given second, but if you're at 60fps, that .2 second gap becomes a .1 second gap. Its not really a going from 70FPS to 0FPS matter, it is how bunched together those frames are during a given second. I could go find some prior links, but in short, we have observed that a higher average frame rate will negate the observable impact of the Crossfire microstutter.

Going back to how we do things - we frankly don't care what the frame delivery time is. We select the most challenging level within a given game (that uses all of the graphical features and hits the cards the hardest) and we play through the level. If it is choppy we drop settings, even if we're averaging 100FPS (ok, that's extreme, but to illustrate the point).
 
Going back to how we do things - we frankly don't care what the frame delivery time is.

That's a great way to look at it! It's all about the overall experience, and hearing that the 7970 CFX delivered the best experience out of all of those options is sure a nice change of pace from everyone focusing on frame delivery statistical metrics instead of the total gaming experience.
 
What would be the best solution for 7680x1600 gaming? Right now I have 3x GTX 670 (4GB) and I'm not necessarily happy with it.

I saw in one of the other reviews here that 3x Titans absolutely SMOKED 3x 7970s and 3x 680s at 5760x1200, often being 50-100% faster.

What is the reason for that though? I could definitely see the 680s being VRAM bottlenecked (and possibly the 7970s, although much less so), which wouldn't apply to my 4GB cards so I wouldn't expect such a dramatic performance increase.

The 384 bit bus in the Titans could also be a major factor as well.

Or it could be mostly the different chip.

Three Titans is a lot of money though and they're going to depreciate pretty HARD so it's an investment I'm cautious about making. I am trying to gauge how they would compare to three GTX 780. If the memory bus is the main factor making the Titans so good at multi-gpu multi-monitor, then it would probably be worth it to pay the extra to get 384 instead of 256/320 speculated on the 780s. If it's mainly chip design and VRAM contributing to the Titan's performance, they might not be worth the price premium.
 
Good grief, I don't typically get into it but could you let your bias show any more here Brent?

If this was the same article with an AMD GPU you'd be going on about how it isn't worth it because of the lower settings, framerates, and price-performance. I know because I've seen it. Because it's Nvidia you're ranting about how "impressive" it is that it's ONLY 25-35% slower (frequently with lower AA or other in-game settings).

Give it a rest, we get that you prefer Nvidia but can you at least try to be objective with these things? The Titan got hit by a lightning bolt in this and you're more or less telling us that at least it didn't get fully immolated.
Apparently you didn't even bother to read where he praises the AMD setup as better than Nvidia...:rolleyes:
In all cases, Radeon HD 7970 GHz Edition CrossFire from two separate video cards was still the fastest, and best gameplay experience, for a lesser price.
In terms of price and nothing else, the $999 AMD Radeon HD 7990 video card is a better purchase than one single GeForce GTX TITAN. We know two TITAN's in 2-way SLI will blow HD 7990 out of the water, but it is also $1000 more expensive. However, the real show stopper is still two separate Radeon HD 7970 GHz Edition video cards with CrossFire enabled. Those can be purchased at $100-200 less expensive than GeForce GTX TITAN or AMD Radeon HD 7990, and provide a lot more performance. 7970 GHz Edition CrossFire will win every time, and by a larger margin than Radeon HD 7990.
 
Apparently you didn't even bother to read where he praises the AMD setup as better than Nvidia...:rolleyes:

What Runningflame570 is getting at, is that when AMD "loses" in one of these reviews, he feels that Brent isn't giving them any saving graces. Whereas in this review where AMD clearly came out on top, Brent went out of his way to maintain a positive outlook on nVidia - such a courtesy is not afforded to AMD when it loses.
 
Back
Top