Discussion in 'Reviews' started by W1zzard, Aug 20, 2013.
To read this review go to: http://www.techpowerup.com/reviews/MSI/GTX_760_HAWK/
Fast card no doubt about it, low power consumption and with some nice O/C headroom, the GTX x60 series is now mixing it with the 7970, how things have moved on.
"This could be a good indicator of what to expect when you buy the card, or we could have gotten a very lucky sample."
A quick look at other reviews show that overall, this just may be the best overclocking GTX760 to this day
Great review and very good card, little to much premium for me.
Hey W1zz what would happen if you plug just 1 x 8pin in? It only used 230w...
probably wont start.
the 760 hawk is built well. but the price needs to come down. the GTX 760 has no chance against HD 7950 cards selling for USD 220 - 280 at newegg. these cards come clocked at 925 - 1000 mhz out of the box and on average OCs hit 1100 - 1150 mhz. at those speeds HD 7950 is faster than GTX 760(1.3 ghz)
also TPU's performance summary average is disproportionately affected by Starcraft II which has the AMD cards running at just above 50% of Nvidia cards perf. If TPU can include a game which affects the average disproportinately there are many games which run much better on AMD cards which can find a place in tpu suite. eg:COH 2 with SSAA 4x, Dirt Showdown with Global Illumination
In other reviews on the web the HD 7970 Ghz is on par or slightly ahead of GTX 770 and the HD 7970 is faster than GTX 670. TPU shows a stock GTX 670 faster than stock HD 7970 which is not a correct reflection of how these cards perform in the latest and most popular games.
amd fanboy or review troll too?
Agreed. Mine clocked to the moon for my review!!!
The ram almost hit 2K in fact.
ok, so as i understand the review properly, the functionality of the ln2 bios switch wasnt tested, right?
All it does is disable OCP/OVP and allow a 185% power limit.
I could clock the same on the original bios as I was not hitting the power limit in the first place.
that's correct. see earthdog's response for what it does
Wizz, you know more about bios' and how they work when I do. Is there a way manufacturers can implement a second bios without having to reinstall drivers? Its like the system thinks its a different card... Is that possible?
first get your reviewing methodology right. stop benching cards which cost USD 170 and USD 700 at same settings
you have perf numbers which are meaningless. there is no min fps at all which is one of the most important factors in determining playability. no fraps graph. no frametimes using fraps or fcat. no video of the scene/level being tested as techreport and hardwarecanucks do or even explaining the level and exact gameplay scene being tested as hardocp do. how many benchmarks are actual gameplay benchmarks ? how many are using in game benchmarks which in most cases is not a good indicator of game performance. pick 8 - 10 of the latest and most important games. focus on quality and not quantity. try and bring more credibility into your benchmarking.
and yes if you believe a 760 is better than a HD 7950 run a HD 7950(1150 mhz) vs a GTX 760(1300 Mhz) benchmark faceoff. also do not believe you are the only site on the internet benchmarking GPUs. many sites do a better job than you and there are even end users who do a better job than you. good luck. if you want to get vengeful and ban me goodbye.
so with ovp gone, can i change the voltage freely right away with the default app or do i have to ask for a special licence?
Voltage is still limited. I will be testing out the MSI AB beta14 tonight to see if I can get 1.3x volts out of it though. I REALLY want 1400MHz core out of this thing on air... was close at 1359MHz for some benchmarks.
1. Minimum framerates are important to an extent. Is anyone going to base a buying choice over what one FPS over THOUSANDS through the benchmark are? I think if one were to do a minimum FPS, it should be presented as time below XX FPS which gives one a better idea of how long it is hanging out down there. The value of that metric goes down the higher the FPS are anyway.
2. Is there a point to test single card frametime? Last I heard this problem was with SLI/CFx which, for the most part, has been fixed outside of DX9 and multiple monitors...
3. I would like a more clear explanation of what sections of games are tested. I think he lists it when he adds it? But after that, I don't see a mention.
3a. As far as manual walkthroughs to benchmark. I thought the same thing you did a while ago until I actually sat down and tested it. If you try to do the same things you did on each run through and not deliberately go off course so to speak, it is remarkably consistent. I mean sure you may not reload in the same place or duck for cover or whatever, but the longer that it is tested, the less overall variance you will have between results for the most part. Also, not a lot of games have canned benchmarks either. So do we(reviewers) leave out a AAA title like BF3 because it doesnt have a canned benchmark? That is the question reviewers ask themselves on this issue (well our team anyway, LOL!).
I really think there is a much better way to get your point across than to post in the manner you did. Wizz is a well respected reviewer and his reviews are right up there in places I look first, and many others do as well. If I were to nitpick, my only concern with the reviews, all of them here, is the scoring. I do not think there is an empirical method behind it. For example, what is the difference between a 8.5 and 8.7? How do you take off tenths of points? It just seems 'different' each time. That said, I never look at the scores but the data, which is plenty to make an informed decision off of. On the flip side, the site I review for uses Approved, MEH, and Fail and have each clearly defined.
No review site or method is perfect... we all try to do the best we can to bring the best information to the readers. Give him some slack or be CONSTRUCTIVE in your criticism.
Wizz, keep up the good work you and thousands of others know you do. Don't let punks like this slow you down.
thats why a fraps fps graph is useful in seeing how much time it spends below 30 fps. even then it has been shown that frametime graphs are the best measure of gameplay smoothness.
yes there is. just because it has been fixed in past games does not mean frametime consistency issues will not occur in future titles. reviewers need to make sure Nvidia/AMD don't let up.
no. reviewers should avoid in game benchmarks and look at benchmarking gameplay runs. obviously they need to try and keep the run as much consistent as possible. so as you mention it involves not deviating from a planned course or set of actions.
yeah i would have if he had not called me a fanboy. If he believes he is non-biased why would he respond with personal attacks.
tpu reviews need to reflect the trends in benchmarking and the focus on gameplay quality (frametimes) rather than raw perf(avg fps). anyway thats for tpu to figure out.
Perhaps you do not understand. That issue is with MULTIPLE GPUs, not single... so why would that be an issue in THIS review?
Bias and personal attacks have nothing to do with each other... I dont get the point there.
That said, I'm out. Nice review as always Wizz!
Thanks for the review. This is my first post (new to PC gaming) and just got this card from online retailer. I am very impressed though I'm noticing driver instability in some benchmarks, notably 3DMark demo and Metro Last Light benchmark. Also had a few games that would fail to start. Had to revert Nvidia driver back to 320.49 from 320.63 I believe it was. Now Metro Last Light is working but 3DMark demo is still crashing about 2mins in. Looks like a new driver might be out today so gonna give it a shot this evening. Still very happy with this GPU as it runs all my games in HD very well and is super quiet. Paid $289 like the review says. Was worth the extra $20-$30 over the gaming version to me, but I have an MPower board and wanted the look along with the quality.
BTW, I love Techpowerup and its Android app. Helps me stay informed even with a busy schedule. Thanks again and please keep providing these great reviews!!
I used the 326.xx beta driver that is at the Nvidia site, and 3DMark (Firestrike) worked just fine. We do not test Metro:LL though, sorry.
the topic of frametime consistency problems on single GPU became a huge issue with techreport's article on AMD HD 7950 last dec.
after that a lot of sites have started bringing in fraps or fcat based frametime benchmarking.
Fraps frametime is not good at showing the big picture. So MEH on that techreport article. THEY needed to do it right, LOL! One needs the FCAT behind it...Here is a quote from our article/take on this whole thing...Please specifically go to the Anandtech link which should clear a lot of things up for you.
Regarding single card issues (from anand article):
Also note, for single cards, it was mostly fixed since march:
More on FRAPS being MEH for this testing:
Single card testing:
FRAPs is not the tool to use to determine frame times to begin with. It does not read the frames that get transferred through the cable onto the monitor.
And again, frametime issues are virtually non existent on single GPU solutions, it may show up from time to time, but compared to multi GPU solutions its a non issue.
there is no methodology for the numeric score. it's based on my personal feel, mostly relative to other cards of the same series. the idea is to condense the whole review into a single number, which is kinda impossible as we all know, but it helps some people decide whether they even want to read the conclusion (after skipping to that page immediately of course)
then why are you even here wasting your time commenting?
I just searched your recent posts and made an observation
minimum fps is a useless metric. in many cases it measures when a harddisk access or similar occurs that slows down an individual frame, ignoring the rest of the benchmark data. you play for an hour and determine overall playability by a sub-second instant in time?
imo the correct way to use minimum fps is to provide a histogram of frametimes, probably with some rejection of outliers, which no normal user would understand
games which have a benchmark use the benchmark, everything else (the majority) uses normal gaming.
publishing which scenes you use is the same as using a canned benchmark because drivers will be actively optimized for it by NV and AMD
It's a maneuver to plug the chasm that’s between their $260 and $400 offerings, but 13% more money for 9% performance, its value isn't stunning. OC brings that value position up, but hard to say how given BF3 would translate across various games. I'd like to see a head-to-head between this and some" Top-Shelf" 7950 Boost. The last set of numbers for such a card was Nov 2012, for a HIS 7950 IceQ X² Boost 3 Gb, hard to correlate (1920x1200) such results. While I concur the Summary isn't the whole story, you need to peruse the individual games.
Given the AMD "sell-off" here in the States... Newegg there's been various 7950 w/Boost for $180 working codes and rebates, plus 3 games from Radeon Gold Rewards, or a nice Sapphire (100351SR) 7970 OC with Boost down at $260, and 3 titles to pick from, it's hard to be ecstatic for this. Essentially Nvidia is now dropping a binned version off "first-round" GK104 (aka a GTX660Ti) they’ve accumulated for months and still believe the credence it’s a $300 offering.
In about 4 weeks this might not appear to be any value, I perceive a shake-up in the mainstream segment.
Separate names with a comma.