tweaktown review GTX660Ti

[MTDEW]

The only ones nv are trying to please are themselves. There's no way this thing is the same value for a card as the 7950. Yet many pro nv websites paint it as equal or better value by adhering to nv's review guide and comparing it to only reference cards from the competition. That's doing a diservice to there readers. [H] review was a reach around sack fondling for nv. Some of the most realistic reviews came from xbit and tomshardware and even AT. No doubt they'll bear the wrath of nv's PR.

I usually stay out of these arguments, because they just end up in a flame war but....
Yeah, I still don't see why if you review a factory overclocked card, then why not just compare it to the competitors factory overclocked cards.
If you review a reference clocked card, then compare that to competitors reference cards.
It really is that simple and straight forward to post fair comparisons, yet it does seem to be avoided as much as possible on purpose.

They didn't sign an NDA.

But the supplier of the cards certainly did, that is why TweakTown avoided doing any tests which might reveal who they are.
TweakTown is not quite the "innocent victim" they claim to be.

But it really was was an honest straightforward review by TweakTown anyway despite their relationship with nvidia.

[H]'s review was just ludicrous. The stock GTX 660 Ti was paired against the GTX 460, and the 3GB model from galaxy went against 800MHz 7950s.

I never read [H]'s reviews because I dont care for how they are presented.
But since it was mentioned, i went and read it.

That is BY FAR one of worst/biased reviews I've seen in long time.

 

[zebrax2]

I was actually surprised that this didn't happen earlier for tweaktown.

Anyway their test suit isn't really that good but at least we have something to talk about a week or so prior to the release of the cards with their early preview/review.

 

[RussianSensation]

That is BY FAR one of worst/biased reviews I've seen in long time.

HardOCP has their moments of brilliance such as MSI GTX680 Lightning review and then they flop miserably with that 660Ti vs. 800mhz 7950 review, totally ignoring overclocking potential of MSI TF3 7950 or 950mhz Sapphire Dual-X 7950 that's $10 more than the Galaxy 660Ti card they used....:sneaky:

 

[Homeles]

TweakTown is not quite the "innocent victim" they claim to be.

I never stated they were.

 

[Cerb]

It's really surprising how much the performance varies by game. Sometimes it's right up there with a 7950, sometimes it can barely hang with a 7850, and then AA can just kill it. The asymmetrical memory and lack of ROPs does not make for easy performance level prediction, it appears.

 

[toyota]

the asymmetrical memory is NOT an issue like some of you think. I have a card that uses that and have tested the crap out of it. it scales fine with clocks and matches the faster cards just fine when overclocking. its a 1gb card and has no issues or drop off when exceeding 768mb of vram either so you guys need to stop thinking the 660ti is a 1.5gb card with another 512mb of useless ram.

 

[ShintaiDK]

Even AT's review doesn't appear to test a ref clocked 660 Ti, despite comparing to other ref clock cards on the graphs.

It's really surprising how much the performance varies by game. Sometimes it's right up there with a 7950, sometimes it can barely hang with a 7850, and then AA can just kill it. The asymmetrical memory and lack of ROPs does not make for easy performance level prediction, it appears.

Maybe you should look again. Whats that light green bar?

 

[BenSkywalker]

After looking across a bunch of reviews it is clear that the 660Ti event has been backed by nothing but nVidia shills, what I don't understand is why everyone hasn't called out the, by far, biggest nVidia shill in this whole debacle?

One group of people are claiming that the 7950 boards can only clock to 950MHZ if you take the voltage from 1.093V to 1.25V in a very obvious bald face lie that the people in this thread know to be wrong. Every 7950 hits at least 1.1-1.3GHZ at stock volts and every single board hits that without any issue whatsoever. But no, we have people just reporting as fact that just because that nVidia shill "AMD" decides to state something like that web sites are supposed to actually take that into consideration?

Seriously?

Clearly this "AMD" is nothing but a PR firm for nVidia and are doing nothing but spreading FUD to hurt the 7950. The people in this thread *know for a fact* what every single one of these 7950s will clock to and clearly this "AMD" only released this information to encourage reviewers to ignore the overclocking potential because of kick backs or pressure from nVidia.

Why would a responsible reviewer listen to this "AMD"? Who are they? Why would they know anything about the 7950 that the forum posters don't? It's a sad day that the reviewers aren't smart enough to ignore this "AMD" PR firm and instead report the fact that 7950s all overclock just like the samples they have, sadly they fall for nVidia's latest PR company ploy.

This "AMD" needs to be watched.

 

[Cerb]

Maybe you should look again. Whats that light green bar?

Something I completely missed, in my perusal of reviews D:.

 

[ViRGE]

Tweaktown has repeatedly circumvented NDA's and perhaps themselves have broken an NDA (not entirely sure on the latter) and now Nvidia is doing what they can to shut out Tweaktown from getting Nvidia products, regardless of whether the product is super 3LITE K!LLER or average.

Indeed. To break the NDA is a very serious thing. Even if they don't have a signed agreement it means that a partner is supplying cards to an unauthorized party, which of course would be a major violation of the supplier's agreement.

Worse, by giving TT cards it encourages TT and other parties to publish these godawful reviews. The single greatest thing about NDAs is that it allows real review sites to do proper benchmarking and analysis. Otherwise we'd be in a race to the bottom where the first guy to publish 3 benchmarks (having never updated his 7950 from the 12.1 launch drivers) would "win".

why does Wolfenstein say its DX in Precision? also I dont see how those benchmarks are so low for those Nvidia cards. my much slower card has never once dropped below the 60 fps cap in the game so far.

Long story short, the SP game is Direct3D9. The MP game is OpenGL. I have absolutely no idea why they did it that way.

 

[Crap Daddy]

One group of people are claiming that the 7950 boards can only clock to 950MHZ if you take the voltage from 1.093V to 1.25V in a very obvious bald face lie that the people in this thread know to be wrong

It is a group of people who happen to work in a company called AMD. They decided to release a BIOS which gives 1,25V to assure that EVERY 7950 will run STABLE at the advertised 950MHz.

As for: "Every 7950 hits at least 1.1-1.3GHZ at stock volts and every single board hits that without any issue whatsoever", I prefer to link you this:

http://forums.anandtech.com/showthread.php?t=2228094

The average on stock is 1050Mhz which is 1.05GHz

 

[zaydq]

It is a group of people who happen to work in a company called AMD. They decided to release a BIOS which gives 1,25V to assure that EVERY 7950 will run STABLE at the advertised 950MHz.

As for: "Every 7950 hits at least 1.1-1.3GHZ at stock volts and every single board hits that without any issue whatsoever", I prefer to link you this:

http://forums.anandtech.com/showthread.php?t=2228094

The average on stock is 1050Mhz which is 1.05GHz

I hit 1100 on stock volts... but it seems the new MSI TF3 7950s with 7970pcbs have been getting that. Must be specific to MSI.

 

[RussianSensation]

Maybe you should look again. Whats that light green bar?

Yes we know BF3 runs well on NV cards and an 800mhz 7950 is slow in that game. But it only takes a 925mhz 7950 to catch a 660Ti:

So what happens when we look at apples-to-apples 7950 vs. 660Ti overclocking in BF3?

HD7950 @ 1.025ghz

GTX660Ti @ 1.15ghz on GPU core (with GPU Boost hitting well over 1200mhz)

Of course we also know as BenSkywalker has alluded that cards such as the MSI TF3 7950 can hit 1.1-1.2ghz, which means an overclocked 660Ti can't win in BF3 and will get beaten overall by 20-30% in a wide variety of games since BF3 is one of 660Ti's better showings.

This is mysteriously ignored by the same crowd that has praised GTX460/470 overclocking last generation. :thumbsup:

It is a group of people who happen to work in a company called AMD. They decided to release a BIOS which gives 1,25V to assure that EVERY 7950 will run STABLE at the advertised 950MHz.

As for: "Every 7950 hits at least 1.1-1.3GHZ at stock volts and every single board hits that without any issue whatsoever", I prefer to link you this:

http://forums.anandtech.com/showthread.php?t=2228094

The average on stock is 1050Mhz which is 1.05GHz

The average is probably 1050mhz because it includes the crappy reference versions with poorly binned ASICs. Check a thread where people buy a good aftermarket 7950 and you'll see 1200mhz to be fairly common. No reason for anyone to buy a reference 7950 when Gigabyte Windforce 3x, Sapphire Dual-X and MSI TF3 7950 are often within $20 range of a 660Ti.

The stock voltage for Tahiti XT is 1.175V. I think Ben is talking about that stock voltage not necessarily the stock voltage of all 7950 cards since some ship with a low stock voltage of 0.993-1.0V. The point he is making it's laughable to use power consumption of HD7950 GPU Boost edition @ 1.25V OR to conclude that 7950 somehow needs 1.25V to overclock to 950mhz or even to 1200mhz. The GPU Boost if anything hurt AMD's marketing since we've known for months now that a good after-market 7950 can hit 1100-1200mhz on or below 1.175V all day.

 

[zaydq]

Yes we know BF3 runs well on NV cards and an 800mhz 7950 is slow in that game. But it only takes a 925mhz 7950 to catch a 660Ti:

So what happens when we look at apples-to-apples 7950 vs. 660Ti overclocking in BF3?

HD7950 @ 1.025ghz

GTX660Ti @ 1.15ghz on GPU core (with GPU Boost hitting well over 1200mhz)

Of course we also know as BenSkywalker has alluded that cards such as the MSI TF3 7950 can hit 1.1-1.2ghz, which means an overclocked 660Ti can't win in BF3 and will get beaten overall by 20-30% in a wide variety of games since BF3 is one of 660Ti's better showings.

This is mysteriously ignored by the same crowd that has praised GTX460/470 overclocking last generation. :thumbsup:




The average is probably 1050mhz because it includes the crappy reference versions with poorly binned ASICs. Check a thread where people buy a good aftermarket 7950 and you'll see 1200mhz to be fairly common. No reason for anyone to buy a reference 7950 when Gigabyte Windforce 3x, Sapphire Dual-X and MSI TF3 7950 are often within $20 range of a 660Ti.

The stock voltage for Tahiti XT is 1.175V. I think Ben is talking about that stock voltage not necessarily the stock voltage of all 7950 cards since some ship with a low stock voltage of 0.993-1.0V. The point he is making it's laughable to use power consumption of HD7950 GPU Boost edition @ 1.25V OR to conclude that 7950 somehow needs 1.25V to overclock to 950mhz or even to 1200mhz. The GPU Boost if anything hurt AMD's marketing since we've known for months now that a good after-market 7950 can hit 1100-1200mhz on or below 1.175V all day.

^ what Russian said

I hit 1200mhz on 1.168v, which is still below 7970 stock voltage of 1.175v, on my 7950

 

[piesquared]

Indeed. To break the NDA is a very serious thing. Even if they don't have a signed agreement it means that a partner is supplying cards to an unauthorized party, which of course would be a major violation of the supplier's agreement.

Worse, by giving TT cards it encourages TT and other parties to publish these godawful reviews. The single greatest thing about NDAs is that it allows real review sites to do proper benchmarking and analysis. Otherwise we'd be in a race to the bottom where the first guy to publish 3 benchmarks (having never updated his 7950 from the 12.1 launch drivers) would "win".[\b]
Long story short, the SP game is Direct3D9. The MP game is OpenGL. I have absolutely no idea why they did it that way.

On the other hand it also allows results to be unrealistically skewed in their favor because, well, review guides are %90 marketing. And thatis also no benefit to consumers. Case in point: nv is trying to manipulate consumers into believing this card is a better purchase than the 7950 when clearly it is not. It all comes down to perception and that is what the review guides are all about. but as we can see by the [h] review it isn't based in reality.

This whole thing with Tweaktown has nothing to do with an NDA. Sights have leaked early results many many times over the years. Nv is being a [mean] because these particular results show the real performance of the card, not the results the bean counters wanted to pitch. And that benefits consumer.

No profanity please
-ViRGE

 

[skipsneeky2]

Russian,it just simply seems people ignore overclocked numbers and just rely on stock clock numbers these days to crown victors so i honestly think you are not gonna get anywhere with the majority of the crowd.

I as someone who has owned both a 7970,and now 2 gtx670 cards,when i had that 7970 overclocked,it felt faster then a gtx670 for sure and heck when i got that 7970,i took its overclocking capabilities into mind.

OCN still holds it grounds when it is overclocked,that is a sure factor most appear to not getting at.

Got rid of my 7970 cause it was the lousy loud reference model.

 

[MrK6]

Russian,it just simply seems people ignore overclocked numbers and just rely on stock clock numbers these days to crown victors so i honestly think you are not gonna get anywhere with the majority of the crowd.

The thing is, all these cards are so close now you can make anything in the upper tier be "the best card" simply by changing the requirements for whatever is being asked/argued. I think RussianSensation, and some other posters, do a good job backing up their thoughts and arguments, but it's falling on deaf ears for the most part. The problem that people want to be "right" more so than come to a conclusion or some form of enlightenment (on the internet of all things, who knew ). It might be worth it to save some time and effort by just clarifying our discussions before they get heated or trolled.

 

[ViRGE]

On the other hand it also allows results to be unrealistically skewed in their favor because, well, review guides are %90 marketing. And thatis also no benefit to consumers. Case in point: nv is trying to manipulate consumers into believing this card is a better purchase than the 7950 when clearly it is not. It all comes down to perception and that is what the review guides are all about. but as we can see by the [h] review it isn't based in reality.

This whole thing with Tweaktown has nothing to do with an NDA. Sights have leaked early results many many times over the years. Nv is being a [mean] because these particular results show the real performance of the card, not the results the bean counters wanted to pitch. And that benefits consumer.

Pie, you've been here long enough that I know you're not a boob, so I can't imagine you seriously believe the conspiracy theory you're spouting. Violating NDAs has always been a surefire way to get a site blacklisted. TT is not the first site to play this game and to lose (badly).

Anyhow, who said anything about the reviewer's guide? Go read the AnandTech review and tell me that isn't a well written article with sound benchmarking. That's the kind of thing we'd lose in a free-for-all environment.

 

[Attic]

Pie, you've been here long enough that I know you're not a boob, so I can't imagine you seriously believe the conspiracy theory you're spouting. Violating NDAs has always been a surefire way to get a site blacklisted. TT is not the first site to play this game and to lose (badly).

Anyhow, who said anything about the reviewer's guide? Go read the AnandTech review and tell me that isn't a well written article with sound benchmarking. That's the kind of thing we'd lose in a free-for-all environment.

Ryan does a great job with his articles, but ommitting the avaible and popular non reference 7950 cards was IMO a big oversight.

It's clear that benching the non refernce 660ti against the reference 7950 paints the 660Ti in a very favorable but incomplete light given what prospective buyers have available to them. The current information and card performance should have been shown in Anandtech's 660ti review.

Ommiting the non reference 7950's,.. even the mention of them doesn't come across well and reeks of a review guide guideline. The review guides are certainly real things and I'd wager a large sum that nVidia, through the propaganda in their review guide, pushed for just the type of review that anandtech delivered.

 

[ViRGE]

It's clear that benching the non refernce 660ti against the reference 7950 paints the 660Ti in a very favorable but incomplete light given what prospective buyers have available to them. The current information and card performance should have been shown in Anandtech's 660ti review.

Que?

They have a reference clocked 660 Ti there. It's in light green.

 

[RussianSensation]

ViRGE, I think Attic's point is that since HD7870 goes for $280 or so, AT should have included 880-900mhz 7950s such as MSI TF3 or Gigabyte Windforce 3x since they go for $320-330 on Newegg and other places.

They tested a reference-based 925mhz GPU Boosted 7950 with 1.25V that frankly is an atrocious card. Not only does it cost $350, but the GPU Boost "works 50% of the time". For $350 they could have instead included a 950mhz Sapphire Dual-X that uses binned 7950s, or included those TF3 / Gigabyte Windforce 3x 7950s.

It's somewhat odd to include a $350 reference 7950 that has been overvolted from the factory when at least 3 other 7950s are better and cost less or the same.

Now what happens at the end of the review, power, temperature and noise of after-market 660Tis are compared against a reference overvolted 7950.....yet there are 3 awesome 7950s that weren't included?

Then once we get to the Overclocking section, 660Ti, including after-market versions, was overclocked to the max, but a reference 7950 with GPU Boost was left at stock speeds. I see the Gigabyte Windforce 2x card hit 1278mhz Overclock and then used here against a 925mhz 7950:
http://www.anandtech.com/show/6159/the-geforce-gtx-660-ti-review/20

^ That has got to be the most strange OCing head-to-head I've ever seen. A 925mhz reference 7950 was tested against a 1278mhz after-market max OCed 660Ti.

That's ummm.......strange to say the least. Why wouldn't the same review do Max OC 7950 vs. Max OC 660Ti reference + Max OC after-market 7950 (MSI TF3/Sapphire DX) vs. Max OC after-market 660Ti (MSI Power Edition)?

Instead, the review totally penalizes a reference based cooler of the 7950, high power consumption of the poorly binned 7950 chip that needed to be overvolted to 1.2v+ to get 925mhz, and in the end wasn't even overclocked in the OC section against overclocked 660Tis....how is that a fair review? It basically put 660Ti in the best possible light and 7950 in the worst possible light.

What consumer will choose a reference based 7950 for $350 over a $300 after-market 660Ti MSI Power Edition? No one. It's basically stating the obvious by picking the worst possible HD7950 SKU and not even overclocking it. The review wasn't written fairly imo to allow the gamer / consumer to pick the best card at $300-349 level because it never considered good after-market 7950s.

 

[BenSkywalker]

This is mysteriously ignored by the same crowd that has praised GTX460/470 overclocking last generation.

This is a very good point, I just can't seem to find the numerous links where nVidia came out and overvolted the 460/470 heavily and told all the reviewers that is the voltage that was needed to hit a relatively mild overclock on those boards, can you please link that up? Clearly we are being the impartial ones here, not that nVidia PR firm called "AMD".

The point he is making it's laughable to use power consumption of HD7950 GPU Boost edition @ 1.25V OR to conclude that 7950 somehow needs 1.25V to overclock to 950mhz or even to 1200mhz.

Actually the point I'm making is that AMD came out and told everyone that the 7950 isn't that good of an overclocker and needed large voltage bumps to hit a relatively minor overclock. If you are a reviewer and the company that makes the card tells you that is the case, you are supposed to assume they are liars and paint their card in a much better light then what they claim?

Ryan used the updated BIOS that AMD wanted them to, despite it not being available to consumers yet, and provided numbers on that part along with what people are actually willing to buy. They followed AMD's marketing propaganda and it still wasn't enough, they didn't go out of their way enough to set up the playing field to overclock the more expensive part to beat out the cheaper part.

So if you ignore AMD's statements, if your spend more money, if you manage to hit a ~20% overclock then the 7950 looks better then a stock clocked 660Ti and people are making it out like it isn't close.

Ryan does a great job with his articles, but ommitting the avaible and popular non reference 7950 cards was IMO a big oversight.

Ryan included AMD's PR BIOS for the 7950. Given that AMD is changing the voltage guidelines for parts and according to AMD those are supposed to be showing up at retailers by now what good would it really do? Is he supposed to come out and say AMD wasn't being honest and I'm going to bench a board that isn't going to be available anymore as configured because I think AMD is better then AMD claims they are?

 

[BenSkywalker]

Instead, the review totally penalizes a reference based cooler of the 7950, high power consumption of the poorly binned 7950 chip that needed to be overvolted to 1.2v+ to get 925mhz, and in the end wasn't even overclocked in the OC section against overclocked 660Tis....how is that a fair review?

AT has always done overclocking that way in all of their reviews.

http://www.anandtech.com/show/6025/radeon-hd-7970-ghz-edition-review-catching-up-to-gtx-680/18

The board getting reviewed gets overclocked, nothing else. I know, they should change that to make AMD look better at every oppurtunity, if AT isn't going out of their way and changing their review policy to always make AMD look as good as possible, even better then what AMD asks them to do, clearly they are nV shills......

 

[RussianSensation]

This is a very good point, I just can't seem to find the numerous links where nVidia came out and overvolted the 460/470 heavily and told all the reviewers that is the voltage that was needed to hit a relatively mild overclock on those boards, can you please link that up? Clearly we are being the impartial ones here, not that nVidia PR firm called "AMD".

It's not about overvolted 7950 or 460. It's about what other cards can you buy at the same price that are direct competitors. Like EVGA GTX460 FTW 850mhz was a direct competitor to the HD6870, so it was included in the HD6870 launch review. We can go now and buy at least 3 after-market 7950s that are better than the reference HD7950 card for $350 or less. Thus at least one of those cards should have been included since the review included after-market 660Tis.

Actually the point I'm making is that AMD came out and told everyone that the 7950 isn't that good of an overclocker and needed large voltage bumps to hit a relatively minor overclock.

AMD never stated anything of the sort. You or other people may have read the BIOS as such. AMD did this:

(1) If you don't know how to overclock, we'll release a factory reference 7950 with 925mhz GPU Boost for $350
(2) We think it's a competitor to the after-market 660Tis

AMD though never said that other 7950s are not competitors. Clearly the MSI TF3 and Gigabyte Windforce 3x are for sale on Newegg for $320 + shipping cost as 660Tis are for $300 + shipping. A tech site's job is to inform the potential buyers of what choices they have. We can use Newegg.ca/Newegg.com and Amazon prices to cover US and Canada. Prices for after-market 660Ti and HD7950 TF3 are also just £10-11 pounds apart in the UK.

So we have 3 countries, US/Canada/UK off the top of my head where the cards are priced $20 apart or £10-11. While the review finds it acceptable to include after-market 660Tis to use in OCing section and for noise/power consumption sections, it failed to include direct competitors to those cards barely costing more. Why?

Ryan included AMD's PR BIOS for the 7950. Given that AMD is changing the voltage guidelines for parts and according to AMD those are supposed to be showing up at retailers by now what good would it really do? Is he supposed to come out and say AMD wasn't being honest and I'm going to bench a board that isn't going to be available anymore as configured because I think AMD is better then AMD claims they are?

That doesn't explain why the 7950 wasn't overclocked at all vs. overclocked 660Tis in the last section. Also, since other 7950s are still available for sale, at least 1 of those should have been included. If and when AMD actually has after-market 7950 Boost Edition, a follow-up review can be made. However, the way the review was done completely ignored the existence of all current after-market 7950s. That's 100% unrepresentative of current or near term market conditions.

Instead our forum members had to go out of their way to do what should have been done by professional reviewers.

In the conclusion of the review I am reading:

"Its only downsides are that the $329 price tag puts it solidly in 7950 territory, and that the cooler is very average, especially when held up against what Gigabyte has done."

Why was it acknowledged that $329 is a pricing bracket of the 7950 but not a single $329 HD7950 was included in the review?

Even drop rebates for a moment. A $350 reference 7950 was included and a $350 after-market 7950 was not. A $350 reference 7950 was not overclocked, while each and every after-market 660Ti was overclocked.

I love AT's reviews but they dropped the ball big time on this one. At the very least, it would be a lot more objective to our readers if Ryan did a follow-up of after-market 7950s OCed vs. OCed 660Tis. It's only fair.

 

[RussianSensation]

The board getting reviewed gets overclocked, nothing else. I know, they should change that to make AMD look better at every oppurtunity, if AT isn't going out of their way and changing their review policy to always make AMD look as good as possible, even better then what AMD asks them to do, clearly they are nV shills......

You think there is some AMD conspiracy theory here? I think even in AMD's reviewers they should overclock AMD and NV cards. In this case not only did they ignore overclocking potential of the 7950 but they used after-market 660Ti vs. reference 7950. Even TechReport included the MSI TF3 7950 to at least have 1 after-market 7950. That's being objective 101.