Review: GeForce GTX 970 SLI review

Published by

Click here to post a comment for Review: GeForce GTX 970 SLI review on our message forum
#4931652
https://forums.guru3d.com/data/avatars/m/166/166569.jpg
I've never tried SLI, but it seems people always moan about drivers and scaling when using SLI.
#4931657
https://forums.guru3d.com/data/avatars/m/175/175739.jpg
Just wondering...... If you had the money to get either 1 Gigabyte GTX 980 G1 Gaming Or 2 Gigabyte GTX 970 G1 Gaming (for SLi) Which of the two will you choose? a Single 980 or 2 970 for SLi? (I know the choice can be easy to some folks, but is it worth it to SLi 970? or is it more worth it to get a single 980 (with no chance of Dual SLi cause that price is just high)?, 1 thing is certain with the price of the 980, its not easy to SLi atleast two of them since they are more expensive than the 970.)
2 x 970's without question. A 980 is expensive and only 10 frames or so different to a 970. You can get 2 970's and have hardly any problems with drivers and SLI but have so much power to destroy a 980 by a long shot and are paying the same amount as what 1 780ti used to be not to long ago.
#4932712
https://forums.guru3d.com/data/avatars/m/230/230236.jpg
People don't always moan about SLI - everyone else who has it working just don't have much to say because they're busy playing.
#4932728
data/avatar/default/avatar21.webp
If you have less than a 4K monitor, I would get 970 SLI. That is an incredible bang for the buck right now. If you have a 4K monitor, or plan to buy one, I would wait a little bit for AMD to come out with their next gen within 3-6 months AND for Nvidia to improve their drivers. We want to see where (multi) gpus can push 4K with a little bit of AA can hold SOLID 60+ or dare I dream 120 FPS with no AA in all games.
#4934119
https://forums.guru3d.com/data/avatars/m/115/115616.jpg
2 x 970's without question. A 980 is expensive and only 10 frames or so different to a 970. You can get 2 970's and have hardly any problems with drivers and SLI but have so much power to destroy a 980 by a long shot and are paying the same amount as what 1 780ti used to be not to long ago.
I had similar situation when buying cards in the past. 260 SLI was basically at the same price as single 285, and the performance was not even close. With 970/980 I'll hold back until I get high-refresh-rate monitor. With 1920x1200@60Hz, getting anything faster than the current 680 is pointless.
#4934331
https://forums.guru3d.com/data/avatars/m/230/230236.jpg
You're not playing the right games 😉
#4993647
https://forums.guru3d.com/data/avatars/m/79/79189.jpg
I truly love your reviews, one of my favorite sites on the net. Keep up the good work. I did feel that you did the review of the GTX 970 SLI a disservice though as the results both game and synthetic are all extremely low based on my testing. All be it you test at default settings and I may tweak my cards with a slight overclock even at default you are way behind my scores. For example I hit over 18400 in Firestorm (oc of +200 core and + 400 mem) http://www.3dmark.com/fs/3716216 and you score 13355. You have a 6 core intel cpu at 4.6 and I have a 4 core cpu at 4.6 so would have expected better results from your rig. At default my cards are scoring 15892... In BF4 using your settings I am scoring 139fps where you are scoring 108, not sure why your scores are so much lower but I will say that this two card solution sure was worth every penny. Thanks again for the great reviews. Greg
You should not compare results from overclocked cards to cards at stock clocks to invalidate performance results..
#4993729
https://forums.guru3d.com/data/avatars/m/79/79740.jpg
You should not compare results from overclocked cards to cards at stock clocks to invalidate performance results..
He also said... "At default my cards are scoring 15892...". So he does have a point. The 970 scaling in the review is only around 40%. I thought it could do better than that. Not sure if there was a scaling bug in earlier drivers that was fixed later on, but either way, a 13355 firestrike is a bit low. Single 970 in review is 9568.
#4993734
https://forums.guru3d.com/data/avatars/m/196/196284.jpg
He also said... "At default my cards are scoring 15892...". So he does have a point. The 970 scaling in the review is only around 40%. I thought it could do better than that. Not sure if there was a scaling bug in earlier drivers that was fixed later on, but either way, a 13355 firestrike is a bit low. Single 970 in review is 9568.
That poster is using an i7 4790K clocked to 4.6ghz whereas Hilbert is using an i7 3960K clocked at 4.6ghz. That poster's CPU is around 15-20% faster, clock for clock. That poster is using the 347.09 driver whereas Hilbert was using 344.07 drivers. There have been at least 9 driver releases since the 344.07 driver. Hilbert was using Windows7 x64 whereas that user is running Windows8.1. That poster's comparison is invalid. Period. Considering the hardware being compared and the difference in drivers between release and current, the noted differences are to be expected. BTW, here's my Firestrike score at my card's default clocks. http://www.3dmark.com/fs/3728665 Only 9961. What my card does on a modest OC... http://www.3dmark.com/fs/3772204
#4993741
https://forums.guru3d.com/data/avatars/m/79/79740.jpg
3960x = 6 cores / 12 threads 4790k = 4 / 8 That alone should put the 3960x (Sandybridge E) above (or at least equal to) 4790k in a firestrike bench (which is heavily multi-threaded). But 9 driver releases later (according to you), and if it puts Footman36's SLI score at 15892 @stock, then thats a +16% difference. Which from what I may have heard at some point or another since 970 release that there may indeed have been an SLI driver issue and is what I suspect to be the case. p.s. you also ignored the 40% poor scaling in the 970 SLI review score vs the 970 single card.
#4993745
https://forums.guru3d.com/data/avatars/m/79/79189.jpg
He also said... "At default my cards are scoring 15892...". So he does have a point. The 970 scaling in the review is only around 40%. I thought it could do better than that. Not sure if there was a scaling bug in earlier drivers that was fixed later on, but either way, a 13355 firestrike is a bit low. Single 970 in review is 9568.
I did bold the word disservice in the quote. Having a point does not make Hilberts review a disservice to the GTX 970. Since Hilbert did state the low score (13355) might be due to a launch driver issue.
#4993749
https://forums.guru3d.com/data/avatars/m/196/196284.jpg
3960x = 6 cores / 12 threads 4790k = 4 / 8 That alone should put the 3960x (Sandybridge E) above (or at least equal to) 4790k in a firestrike bench (which is heavily multi-threaded). But 9 driver releases later (according to you), and if it puts Footman36's SLI score at 15892 @stock, then thats a +16% difference. Which from what I may have heard at some point or another since 970 release that there may indeed have been an SLI driver issue and is what I suspect to be the case. p.s. you also ignored the 40% poor scaling in the 970 SLI review score vs the 970 single card.
I didn't ignore anything. Nor did I miscalculate anything. There's 9 drivers posted on NVidia's site for the GTX970. The driver Hilbert used is not among those drivers. So, unless NVidia pulled 1 or more drivers from their site, there have been at least 9 drivers released to the general public since the 970 launched. SLI scaling issues are common at launch of a new architecture.
#4993750
https://forums.guru3d.com/data/avatars/m/79/79740.jpg
I did bold the word disservice in the quote. Having a point does not make Hilberts review a disservice to the GTX 970. Since Hilbert did state the low score (13355) might be due to a launch driver issue.
Yeah, I have seen mention of SLI driver issue with 970 early drivers here and elsewhere. But to go back to sykoziz suggestion that it may be hardware (CPU differences), its not: 4790k vs 3960x firestrike score: http://www.guru3d.com/articles-pages/core-i7-4790k-processor-review,19.html
#4993753
https://forums.guru3d.com/data/avatars/m/196/196284.jpg
At default my cards are scoring 15892...
Hilbert appears to have used reference cards in his SLI review, whereas here they're being compared to a pair of Asus Strix 970's that have higher default clock speeds.
#4993754
https://forums.guru3d.com/data/avatars/m/79/79740.jpg
I didn't ignore anything. Nor did I miscalculate anything. There's 9 drivers posted on NVidia's site for the GTX970. The driver Hilbert used is not among those drivers. So, unless NVidia pulled 1 or more drivers from their site, there have been at least 9 drivers released to the general public since the 970 launched. SLI scaling issues are common at launch of a new architecture.
And which I noted the following "Which from what I may have heard at some point or another since 970 release that there may indeed have been an SLI driver issue and is what I suspect to be the case."
#4993757
https://forums.guru3d.com/data/avatars/m/79/79740.jpg
I did bold the word disservice in the quote. Having a point does not make Hilberts review a disservice to the GTX 970. Since Hilbert did state the low score (13355) might be due to a launch driver issue.
Sorry, missed that. I did not mean he has a point with the word 'disservice', but rather he has a point with the difference in score. Good that Hilbert mentioned the suspect driver issue, would not like to see others/outsiders finding fault with the review. Hope that Hilbert can do an update though.

Previous Page

Next Page