GTX 660 Vs. HD7850

[p_monks33]

Looking for some real world usage of these two chips, I will be using one of the two in my new M-ITX build and can't decide which one I want to go with. The card will be in a Silverstone SG05 so the length needs to be less than 10". I would also like something that runs pretty cool, and most likely can exhaust air out of the case all while doing so quietly. I can overclock if need be, but really need to keep temps down. I know either card will be slightly CPU bottlenecked by the I3 2125 I will be using in some instances, but that is ok and I figure will be a rare occasion outside of BF3 multiplayer. I don't care about compute so that is a non-factor. I will only be gaming and streaming movies with it. Thanks in advance.

 

[Termie]

The only pure blower exhaust designs are going to be loud, as they are reference HSFs. In fact, I'd say there is no real blower on the AMD side, with the closest thing being this model: http://www.newegg.com/Product/Produc...82E16814131464

On the nVidia side, you can get this: http://www.newegg.com/Product/Produc...82E16814130826. Definitely a blower design, but the cooler is known to be a bit loud, especially at idle (it vibrates). It's 9.5" long, by the way, just within your limit.

With an SG05, though, won't most of the air flow back out of the side vent anyway? If so, your best bet might be to go with the MSI Twin Frozr design of either card:

GTX660: http://www.newegg.com/Product/Produc...82E16814127702
HD7850 TF3: http://www.newegg.com/Product/Produc...82E16814127663
HD7850 PE: http://www.newegg.com/Product/Produc...82E16814127688

Those 7850s are the shortest models out there at around 8", while the 660 is still 9.25".

By the way, the 660 is definitely faster stock-to-stock, but the 7850 will equal the 660 with overclocking. Note that with an eye towards deals, you can get the 7850 2GB for $185 or less after rebate.

BTW, I don't have real-world usage, but I've researched both thoroughly as I'm interested in them for my HTPC.

 

[p_monks33]

Thanks Termie, you are right about the side vents on the SG05, I will probably use a non reference cooled card. I have seen some really good deals on the Sapphire 7850 here lately, and the MSI TF 660 looks very nice as well. I have heard some people complain about stuttering and audio issues with netflix and streaming video via the 7850 so I was leaning towards the 660. Anyone out there using a 660 yet? I am about to bite on one.

 

[LOL_Wut_Axel]

The only pure blower exhaust designs are going to be loud, as they are reference HSFs. In fact, I'd say there is no real blower on the AMD side, with the closest thing being this model: http://www.newegg.com/Product/Produc...82E16814131464

On the nVidia side, you can get this: http://www.newegg.com/Product/Produc...82E16814130826. Definitely a blower design, but the cooler is known to be a bit loud, especially at idle (it vibrates). It's 9.5" long, by the way, just within your limit.

With an SG05, though, won't most of the air flow back out of the side vent anyway? If so, your best bet might be to go with the MSI Twin Frozr design of either card:

GTX660: http://www.newegg.com/Product/Produc...82E16814127702
HD7850 TF3: http://www.newegg.com/Product/Produc...82E16814127663
HD7850 PE: http://www.newegg.com/Product/Produc...82E16814127688

Those 7850s are the shortest models out there at around 8", while the 660 is still 9.25".

By the way, the 660 is definitely faster stock-to-stock, but the 7850 will equal the 660 with overclocking. Note that with an eye towards deals, you can get the 7850 2GB for $185 or less after rebate.

BTW, I don't have real-world usage, but I've researched both thoroughly as I'm interested in them for my HTPC.

It will be 5-10% faster with overclocking. The HD 7850 overclocks 30-35%, and the 660 10-15%.

 

[p_monks33]

So you are saying when overclocked vs overclocked the 7850 wins out?

 

[Termie]

So you are saying when overclocked vs overclocked the 7850 wins out?

I haven't found a review directly comparing the two, but hardocp found the 660 to be seriously lacking in headroom.

http://hardocp.com/article/2012/09/19/asus_geforce_gtx_660_directcu_ii_overclocking_review/

That being said, the OC'd 660 actually matches an 1100MHz 7870 in the four games tested, so my hunch is that even a 1200MHz 7850 will do no better.

 

[TheSisko]

I'm similarly weighing these two cards for an upgrade of my HTPC.
I've found Anandtech's GTX 660 review to be curiously unique in it's conclusion of the superior efficiency and power usage of the GTX 660 vs. HD7850. I really want a card I can keep cool quietly and effectively in a small case.

Sent an email to Ryan but no reply yet...pasted the interesting parts of it below.

but your review implies that the GTX 660 is mostly competitive with the HD7850 in terms of power consumption while offering HD7870 level performance.

However, after reading multiple other reviews, I've been left wondering if the single Metro 2033 measurement in your review is really representative of the relative power consumption of the HD7850 vs. the GTX 660. In the reviews linked below, the power consumption of the GTX 660 is reported as significantly higher than the HD 7850.

http://www.tomshardware.com/reviews/geforce-gtx-660-geforce-gtx-650-benchmark,3297-23.html
http://www.legitreviews.com/article/2023/11/
http://www.techspot.com/review/572-nvidia-geforce-gtx-660/page11.html
http://www.bit-tech.net/hardware/2012/09/13/nvidia-geforce-gtx-660-2gb-review/9
http://hothardware.com/Reviews/NVIDIA-GeForce-GTX-660-RoundUp-MSI-ZOTAC-Gigabyte/?page=11

Perhaps a more thorough investigation is in order? Do you have any data from other games that reflects the Metro 2033 results? What settings did you use for the power consumption test? It would be great if future articles used more than one game and included some details on the settings used for the power benchmarks.

Techpowerup's review (with raised +20% PowerTune limit to eliminate throttling) shows GTX 660 drawing about 30% more power than a HD7850 in Crysis 2 at most resolutions.

Power consumption: http://www.techpowerup.com/reviews/Zotac/GeForce_GTX_660/25.html
Performance-per-watt: http://www.techpowerup.com/reviews/Zotac/GeForce_GTX_660/28.html

 

[Silverforce11]

Check my sig, SG05 with no compromise (not much anyway).

Edit: recently modded to improve airflow, with antec 620 exhaust out top taking all the hot air out. and a enermax fan 120mm intake at the front with filter. case is running whisper quiet and cool OC.

 

[LOL_Wut_Axel]

So you are saying when overclocked vs overclocked the 7850 wins out?

Yes.

The reason why the 660 overclocks so little is that NVIDIA set the power limit very strictly and there's no way around that for now.

 

[LOL_Wut_Axel]

I haven't found a review directly comparing the two, but hardocp found the 660 to be seriously lacking in headroom.

http://hardocp.com/article/2012/09/19/asus_geforce_gtx_660_directcu_ii_overclocking_review/

That being said, the OC'd 660 actually matches an 1100MHz 7870 in the four games tested, so my hunch is that even a 1200MHz 7850 will do no better.

LOL, HardOCP.

They're one of the most biased of the big review sites. Not only do they select a limited number of games, but the games they DO select are in their majority games where NVIDIA has an unusual advantage. I wouldn't trust a single thing they say.

Again, this is pretty simple math. GTX 660 overclocks 10-15% and HD 7850 30-35%. Stock the GTX 660 is ~11% faster than HD 7850.

 

[96Firebird]

Again, this is pretty simple math. GTX 660 overclocks 10-15% and HD 7850 30-35%.

Where are you getting this data from?

 

[LOL_Wut_Axel]

Where are you getting this data from?

http://www.techpowerup.com/reviews/Zotac/GeForce_GTX_660/30.html
http://www.techpowerup.com/reviews/AMD/HD_7850_HD_7870/29.html

 

[96Firebird]

What's with the rolling of the eyes? It is pretty standard to post sources to data instead of just blindly putting it out there as the truth.

And one review site, who's to say they aren't biased like you claimed HardOCP was?

 

[LOL_Wut_Axel]

What's with the rolling of the eyes? It is pretty standard to post sources to data instead of just blindly putting it out there as the truth.

And one review site, who's to say they aren't biased like you claimed HardOCP was?

Maybe because they test 16 games, unlike HardOCP who tested with 4 games in that "review". Unlike HardOCP, TPU tests everything from AMD-biased, NVIDIA-biased, and neutral games. Computerbase.de is another good, unbiased site.

 

[Termie]

LOL, HardOCP.

They're one of the most biased of the big review sites. Not only do they select a limited number of games, but the games they DO select are in their majority games where NVIDIA has an unusual advantage. I wouldn't trust a single thing they say.

Again, this is pretty simple math. GTX 660 overclocks 10-15% and HD 7850 30-35%. Stock the GTX 660 is ~11% faster than HD 7850.
...

Unless you can show me better OC to OC data, I'm sticking with what I've already said.

I know that the 7850 has 30-35% OC headroom. I also know it doesn't achieve one-to-one scaling, despite what a lot of enthusiasts have said on this forum. It's closer to 75% - here's Anandtech's findings at 1050MHz (22% OC): http://www.anandtech.com/show/5625/...-7850-review-rounding-out-southern-islands/19

That would mean a fully maxed out 7850, at 1200MHz, would score about 30% higher than stock, making it at most 10% faster than an OC'd 660. And that's assuming you get a star overclocker.

Point is, on average they are going to be very close when OC'd.

 

[Silverforce11]

Take only from reviews with lots of games to minimize selection bias? Well, TPU, computerbase.de, ABT, guru3d, xbitlabs etc.

 

[Silverforce11]

Unless you can show me better OC to OC data, I'm sticking with what I've already said.

I know that the 7850 has 30-35% OC headroom. I also know it doesn't achieve one-to-one scaling, despite what a lot of enthusiasts have said on this forum. It's closer to 75% - here's Anandtech's findings at 1050MHz (22% OC): http://www.anandtech.com/show/5625/...-7850-review-rounding-out-southern-islands/19

That would mean a fully maxed out 7850, at 1200MHz, would score about 30% higher than stock, making it at most 10% faster than an OC'd 660. And that's assuming you get a star overclocker.

Point is, on average they are going to be very close when OC'd.

Depends on the game the scaling can be 70-90%.

The AVERAGE OC is actually 40% from heaps of samples.

 

[zaydq]

Not sure how I feel about Hardocp's results nor their review. Seems a bit askew from the rest of the reviews out there.

Also, the 7850 is cheaper than the 660, if you could minimize the performance gap with some OCing then you save yourself a bit of money, and still have a decent performer thats efficient and small.

 

[LOL_Wut_Axel]

Unless you can show me better OC to OC data, I'm sticking with what I've already said.

I know that the 7850 has 30-35% OC headroom. I also know it doesn't achieve one-to-one scaling, despite what a lot of enthusiasts have said on this forum. It's closer to 75% - here's Anandtech's findings at 1050MHz (22% OC): http://www.anandtech.com/show/5625/...-7850-review-rounding-out-southern-islands/19

That would mean a fully maxed out 7850, at 1200MHz, would score about 30% higher than stock, making it at most 10% faster than an OC'd 660. And that's assuming you get a star overclocker.

Point is, on average they are going to be very close when OC'd.

Neither does the GTX 660, so what's your point?

Like I said before, it'll be 5-10% faster comparing OCs. I don't care if you think that means they're very close. You're the one that said earlier that they would be the same.

That being said, the OC'd 660 actually matches an 1100MHz 7870 in the four games tested, so my hunch is that even a 1200MHz 7850 will do no better.

HardOCP is notoriously unreliable if you want an unbiased source.

 

[Termie]

I think the OP will have to decide for himself whether the OC'ing advantages of the 7850 are worth it to him. I still have yet to see a review that shows an OC'd 7850 beating an OC'd 660, so thus no one has given him an answer to his question, which was whether the OC'd 7850 beats an OC'd 660.

I provided HardOCP data comparing the 660 to a 7870, and apparently none of you like it. But you haven't come up with anything better. I'm fully aware of all the stock-to-stock reviews out there.

Even XbitLabs, which is usually quite good about this, mysteriously benched the OC'd 660 against an OC'd 7870, not a 7850. It comes out 10% behind the 7870@1240 - would an OC'd 7850@1200 be able to keep up? I sincerely doubt it.

 

[Silverforce11]

Even XbitLabs, which is usually quite good about this, mysteriously benched the OC'd 660 against an OC'd 7870, not a 7850.

They also did benches of 7850 OC vs 7870 OC and other sites show clock for clock, 7850 is ~5% slower.

Very similar to the 5850 vs 5870, 6950 vs 6950... and the 7950 vs 7970.

 

[Termie]

They also did benches of 7850 OC vs 7870 OC and other sites show clock for clock, 7850 is ~5% slower.

Very similar to the 5850 vs 5870, 6950 vs 6950... and the 7950 vs 7970.

Thus a 7850@1240, which would be a 44% overclock, which in turn is well above average, would be approximately 5% faster than an OC'd 660 with a +100 offset over reference.

So the OP has his answer, in a sense.

 

[ShintaiDK]

http://www.anandtech.com/show/6276/...-review-gk106-rounds-out-the-kepler-family/10

Easy to pick. Plus you get PhysX for Borderlands 2 for example.

 

[Termie]

http://www.anandtech.com/show/6276/...-review-gk106-rounds-out-the-kepler-family/10

Easy to pick. Plus you get PhysX for Borderlands 2 for example.

If only it were that easy...

 

[aaksheytalwar]

Avg oc to avg oc
Avg fps usually vary by around 10%
And min fps by upto 15-20%

When comparing

Xx50 and xx70