Sign in with
Sign up | Sign in
Your question
Solved

Hd7870 or gtx660 ti?

Last response: in Graphics & Displays
Share
August 19, 2012 2:23:50 PM

im torned which card to buy..

More about : hd7870 gtx660

a c 1363 U Graphics card
August 19, 2012 2:33:34 PM

Most likely either will make you happy but I would not pay the same money for HD7870 vs GTX660Ti.
m
0
l
a b U Graphics card
August 19, 2012 2:46:51 PM

They're both good cards. I bought a HD7870 a few months ago an have been really happy with it.

A friend of mine in a different thread asked me to compare my OC'd 7870 with benchmarks for a 660ti. This is what I posted for him:

Other cards stats are from: http://www.overclockersclub.com/reviews/nvidia_gtx_660t...

This chart you will find on the link above:


My results for stock clock were very similar to the HD7870 on their chart. I have a Sapphire HD7870, they tested with the FLEX model. They score 36, I scored 35.7.

Now when I move my clocks to 1300/1525 this is my result:


It's a little higher than their result. But theres no telling if they reached the limit on their card or were just doing moderate overclocks. Overclockings are very subjective to the hardware and not all cards of the same model will overclock the same. I think it's a toss up and really depends on what games you prefer and what is on sale at the time of purchase.
m
0
l
Related resources
August 19, 2012 3:49:40 PM

thanks i was thinking of getting the Sapphire 7870 OCed version.. I play bf3, crysis2, alan wake, mostly fps games and third person games..
m
0
l
a b U Graphics card
August 19, 2012 3:53:56 PM

660ti is much better in BF3. HD7870 is a little better in Crysis2. Just try to spend less than $300 and I think you'll be satisfied with your purchase.
m
0
l
a c 87 U Graphics card
August 19, 2012 3:55:29 PM

I'd get a 7870. It overclocks better and has better tessellation performance as well as better AA performance when you overclock the memory. It also has substantially better performance in DirectC/OpenCL features of games. The GTX 670 is the only currently retailing Nvidia card that I'd consider if you want to go Nvidia. The 660 TI'sd 192 bit memory bus is too much of a hindrance in many situations. It has high average FPS, but it has low minimums and high maximums, not very consistent.
m
0
l
a b U Graphics card
August 19, 2012 4:15:32 PM

m
0
l
a c 87 U Graphics card
August 19, 2012 4:27:43 PM

Keep in mind that that is without tessellation and DirectC/OpenCL performance considered much at all. AA is put to use, but its affect is muddied with the many non-AA tests performed. THe difference in any given situation is much higher when you load up the graphics settings and don't have a CPU powerful enough to feed the maximum frame rates of the Kepler cards to mask the low minimum frame rates.

If anyone wants proof of the 660 TI's memory bandwidth bottle-neck, then consider this: The only difference between the 660 TI and the 670 is the memory interface and a minor GPU frequency reduction (not enough to affect performance noticeably with Kepler GPUs).
m
0
l
a c 87 U Graphics card
August 19, 2012 4:36:08 PM

That is because of the other sites using lower (or no, in many cases) AA and a faster CPU. Does OP have an i7-3930K or i7-3960X at 4.5GHz to let the maximum frame rates really mask the low minimums frame rates? I don't think so.
m
0
l
a c 1363 U Graphics card
August 19, 2012 5:03:10 PM

blazorthon said:
That is because of the other sites using lower (or no, in many cases) AA and a faster CPU. Does OP have an i7-3930K or i7-3960X at 4.5GHz to let the maximum frame rates really mask the low minimums frame rates? I don't think so.

It does not make majority of the sites wrong, Tom's review is the one that sticks out compared to the masses http://www.xbitlabs.com/articles/graphics/display/gefor...
Both Xbit and Techpowerup use AA and Tom is not using a slow CPU "Intel Core i7-3960X (Sandy Bridge-E), 3.3 GHz, Six Cores, LGA 2011, 15 MB Shared L3 Cache, Hyper-Threading enabled, Overclocked to 4.2 GHz"
m
0
l
August 19, 2012 5:06:26 PM

I have a 7870. How much should I overclock it? Like specific numbers. The default is 1000 MHz core and 1200MHz memory. What could I increase those to safely?
m
0
l
a c 87 U Graphics card
August 19, 2012 5:17:43 PM

rolli59 said:
It does not make majority of the sites wrong, Tom's review is the one that sticks out compared to the masses http://www.xbitlabs.com/articles/graphics/display/gefor...
Both Xbit and Techpowerup use AA and Tom is not using a slow CPU "Intel Core i7-3960X (Sandy Bridge-E), 3.3 GHz, Six Cores, LGA 2011, 15 MB Shared L3 Cache, Hyper-Threading enabled, Overclocked to 4.2 GHz"


Tom's CPU is not slow, but it is considerably slower.

Using AA is not the same as using only a little of it and in only a few tests and without tessellation and even worse, only in lower resolution tests (for the most part) where it is less important. OP's performance would be quite different from the other sites and even from Tom's because of OP's significantly slower CPU.
m
0
l
a c 87 U Graphics card
August 19, 2012 5:18:57 PM

voiidwulf said:
I have a 7870. How much should I overclock it? Like specific numbers. The default is 1000 MHz core and 1200MHz memory. What could I increase those to safely?


Around 1150-1250MHz should be where you can go safely on the GPU clock and the memory should easily hit well over 1500MHz. 1600-1800MHz is common for the memory, but it also depends on if your cooler covers the VRAM properly. Some don't and can't push the memory far at all. However, I suggest that you look into reviews that cover the 7870 overclocking (preferably with your model of 7870) for better details and tips.
m
0
l
a c 87 U Graphics card
August 19, 2012 5:29:04 PM

rolli59 said:
Would not call that low resolution http://www.xbitlabs.com/picture/?src=/images/graphics/g...


I wouldn't either, but it isn't the majority. There were many people saying that the 660 TI was better because of sub-1080p resolution tests that muddied up the averages and even then, they weren't that forgiving when AA was pumped up to bring performance down to levels that the monitors can display without V-Sync needing to rein in the GPUs as much..
m
0
l
August 19, 2012 5:30:39 PM

blazorthon said:
Around 1150-1250MHz should be where you can go safely on the GPU clock and the memory should easily hit well over 1500MHz. 1600-1800MHz is common for the memory, but it also depends on if your cooler covers the VRAM properly. Some don't and can't push the memory far at all. However, I suggest that you look into reviews that cover the 7870 overclocking (preferably with your model of 7870) for better details and tips.


I have the reference cooler one. Like the basic red and black one.
m
0
l
a c 87 U Graphics card
August 19, 2012 5:32:07 PM

voiidwulf said:
I have the reference cooler one. Like the basic red and black one.


In that case, I wouldn't push the memory far at all. I don't think that the reference cooler covers it well. I also wouldn't go too far on the GPU either with that cooler. AMD's reference coolers kinda suck. A lot. 100MHz GPU and 1400-1500MHz VRAM should be fine, but again, check with reviews of it. I don't have personal experience with a reference 7870.
m
0
l
August 19, 2012 5:38:54 PM

Yeah, OK. I have it set to 1100MHz and 1350MHz now.

Also, my PSU is only 460W. Will overclocking cause problems?
m
0
l
a c 87 U Graphics card
August 19, 2012 5:42:26 PM

voiidwulf said:
Yeah, OK. I have it set to 1100MHz and 1350MHz now.

Also, my PSU is only 460W. Will overclocking cause problems?


Yes that might. I'd use a Corsair 500w with a single 7870 if I was overclocking. I make no guarantees that you won't be fine with that 460w PSU, but I make no guarantees that you will. What brand and model is it?
m
0
l
a b U Graphics card
August 19, 2012 5:49:02 PM

blazorthon said:
Around 1150-1250MHz should be where you can go safely on the GPU clock and the memory should easily hit well over 1500MHz. 1600-1800MHz is common for the memory, but it also depends on if your cooler covers the VRAM properly. Some don't and can't push the memory far at all. However, I suggest that you look into reviews that cover the 7870 overclocking (preferably with your model of 7870) for better details and tips.


Well way to ruin my day!

I thought I had a really good 7870 hitting 1300/1525 stable.
I just tested it at 1300/1600, 1275/1600, 1275/1575, and 1275/1550 and all were unstable :cry:  :cry: 

Oh well, 1300/1525 still yields good performance. I've only been playing Skyrim recently and I keep it at 1000/1200 because I still max the game out with 60% GPU usage. I only crank up the clocks for benchmarks. Stock clocks are more than enough for my gaming needs.
m
0
l
a c 87 U Graphics card
August 19, 2012 5:53:22 PM

stickg1 said:
Well way to ruin my day!

I thought I had a really good 7870 hitting 1300/1525 stable.
I just tested it at 1300/1600, 1275/1600, 1275/1575, and 1275/1550 and all were unstable :cry:  :cry: 

Oh well, 1300/1525 still yields good performance. I've only been playing Skyrim recently and I keep it at 1000/1200 because I still max the game out with 60% GPU usage. I only crank up the clocks for benchmarks. Stock clocks are more than enough for my gaming needs.


1300MHz GPU is pretty good, but a 7870 that has the memory cooled rather than uncovered can hit some pretty high memory frequencies. I would, however, worry about you running it at 1.3GHz unless you have a good GPU that doesn't need a high voltage. What voltage do you set it at for that?

Also, what model 7870 and PSU do you have and what overclocking too did you use?
m
0
l
a b U Graphics card
August 19, 2012 6:01:28 PM

blazorthon said:
1300MHz GPU is pretty good, but a 7870 that has the memory cooled rather than uncovered can hit some pretty high memory frequencies. I would, however, worry about you running it at 1.3GHz unless you have a good GPU that doesn't need a high voltage. What voltage do you set it at for that?

Also, what model 7870 and PSU do you have and what overclocking too did you use?


MSI Afterburner, had to add the EULA to get that memory clock. Voltage, the lowest stable I could use was 1275mV for those clocks. The temps never get over 65C in benching and gaming but I still like the keep it below 1250mV so sometimes I just do stock and slight overclocks unless I want to max it for benching.

Right now I'm using an Enermax Liberty 500w Modular. http://www.newegg.com/Product/Product.aspx?Item=N82E168...
Sapphire HD7870: http://www.newegg.com/Product/Product.aspx?Item=N82E168...
m
0
l
August 19, 2012 6:04:22 PM

blazorthon said:
Yes that might. I'd use a Corsair 500w with a single 7870 if I was overclocking. I make no guarantees that you won't be fine with that 460w PSU, but I make no guarantees that you will. What brand and model is it?


Not sure. My computer is a Dell XPS 8500. People say Dell puts good PSU's in there PC's now though.

Should I remove the OC? I'm planning on upgrading the PSU soon. I don't trust it.
m
0
l
a c 87 U Graphics card
August 19, 2012 6:12:13 PM

stickg1 said:
MSI Afterburner, had to add the EULA to get that memory clock. Voltage, the lowest stable I could use was 1275mV for those clocks. The temps never get over 65C in benching and gaming but I still like the keep it below 1250mV so sometimes I just do stock and slight overclocks unless I want to max it for benching.

Right now I'm using an Enermax Liberty 500w Modular. http://www.newegg.com/Product/Product.aspx?Item=N82E168...
Sapphire HD7870: http://www.newegg.com/Product/Product.aspx?Item=N82E168...


I'd recommend keeping the voltage at or below 1.25mV too. I've heard bad things about getting too close to 1.3v and higher lately, especially with cards that don't have well-cooled VRM and VRAM.
m
0
l
a c 87 U Graphics card
August 19, 2012 6:13:21 PM

voiidwulf said:
Not sure. My computer is a Dell XPS 8500. People say Dell puts good PSU's in there PC's now though.

Should I remove the OC? I'm planning on upgrading the PSU soon. I don't trust it.


I wouldn't trust any PSU from an OEM computer, especially one that might have insufficient wattage. Do you know its rated 12v amperage tolerance?
m
0
l
August 19, 2012 6:33:31 PM

blazorthon said:
I wouldn't trust any PSU from an OEM computer, especially one that might have insufficient wattage. Do you know its rated 12v amperage tolerance?


No. How can I find that out?
m
0
l
a c 87 U Graphics card
August 19, 2012 6:38:21 PM

voiidwulf said:
No. How can I find that out?


Even OEM PSUs should have a sticker on a side of the PSU with this information as well as the PSU's brand and some other characteristics of the PSU. It can also tell you the model, but I've seen a few that don't state the model on it so that's not a guarantee. It might also be listed in Dell's specifications for the machine.
m
0
l
August 19, 2012 6:41:09 PM

blazorthon said:
Even OEM PSUs should have a sticker on a side of the PSU with this information as well as the PSU's brand and some other characteristics of the PSU. It can also tell you the model, but I've seen a few that don't state the model on it so that's not a guarantee. It might also be listed in Dell's specifications for the machine.


Alright, I'll open it up and take a look.
m
0
l
August 19, 2012 7:00:35 PM



That's the sticker I assume?
m
0
l
a c 87 U Graphics card
August 19, 2012 7:00:40 PM

Correct, that is the sticker. Yeah, don't try overclocking much, if at all, with it. It looks like it can handle at least some overclocking from the specs, but I don't recommend pushing your luck with it. Many PSUs (especially those used by OEMs) either can't hit near their rated maximums, or can't do it for long periods of time. Trying to do so in both cases can leave you with a dead PSU or worse. You could give it a shot and probably be fine, but I'm not sure of it. I don't have an XPS 8500 to have tested to be more sure.
m
0
l
a c 1363 U Graphics card
August 19, 2012 7:01:26 PM

It looks to be their Delta unit with 385watts available on +12 volts!
m
0
l
August 19, 2012 7:10:11 PM

OK, I won't overclock until I upgrade the PSU.
m
0
l
August 20, 2012 12:04:48 AM

Thanks for the info guys. my specs are just
i5-3470
Asrock H77 Pro/mvp
8gb G-skill ripjaws
500gb WD digital Black
Corsair Carbide 500r
Seasonic 560w gold Power Supply..
24" Full HD BenQ RL2450H..

These are newly bought and Video Card is the only thing left in my puzzle.. Im not a fanboy of either two companies just want to have the best suited card for my specs and for the taste of my games that I would be playing on a Full HD resolution High to max settings with great FPS.. I really waited to see what the Gtx 660ti can offer to the table against 7870 in particular, because 7950 is already out of my budget.. so just between these two cards I'm choosing on..
m
0
l
a c 1363 U Graphics card
August 20, 2012 12:09:18 AM

Like I said in the beginning of this thread both cards are good with the GTX660Ti having the performance edge everywhere except in Tom's review.
m
0
l
a c 87 U Graphics card
August 20, 2012 12:56:37 AM

You mean everywhere except when people try to use serious AA, some tessellation, DirectC features, and it has poor minimum frame rates. Oh, and that the 660 TI needs a very fast CPU to mask its low minimum frame rates with ridiculously high maximum frame rates from having a GPU that is extremely memory bandwidth bottle-necked.
m
0
l
August 20, 2012 1:05:10 AM

rolli59 said:
Like I said in the beginning of this thread both cards are good with the GTX660Ti having the performance edge everywhere except in Tom's review.
+1 something weird going on there
m
0
l

Best solution

a c 87 U Graphics card
August 20, 2012 1:27:12 AM

regina_49 said:
+1 something weird going on there


I've had enough of this.
http://media.bestofmicro.com/U/P/336769/original/tessel...
There, undeniable proof from none other than Tom's themselves. Kepler has inferior tessellation performance.

All of the reviews, not just the Tom's that list minimum frame rates, show the 660 TI's inferior minimum frame rates. All of the reviews that have comparable AA on/off between tests show the 660 TI doing much worse with AA than AMD's cards, not just Tom's. Tom's simply had a slower CPU, so the 660 TI's maximum frame rates weren't able to stretch as far. They can go almost as far as the GTX 670's can because the 670 and the 660 TI have the exact same GPU, but the 660 TI's weaker minimum frame rates are a result of its weaker memory bus.

It performs inconsistently at best, especially with differently performing CPUs and levels of AA and tessellation. I don't even need to get benchmarks out to tell you how consumer Kepler GPUs are crap for compute compared to GCN GPUs simply because they aren't designed for it whereas the GCN architecture (especially the Tahiti implementation) is designed specifically for compute. Its great tessellation performance is also because that's what it is designed for. I don't even want to guess as to why Nvidia thought it was a good idea to make the horribly unbalanced Kepler GPUs.

Maybe they thought that they could trick people into upgrading more often than they should have to upgrade to retain high quality graphics for the time, maybe they actually had a good reason and decided against telling anyone or even making it obvious. Nvidia sacrificed everything that they could for raw pixel pushing and it obviously backfires when you actually put the GPUs through more severe settings rather than merely giving them more pixels to churn through. For what they can do, Kepler GPUs truly are the best. Not even Tahiti can come close to the considerably smaller GK104 even when it is cut-down by one SMX in raw gaming performance without increasing the settings.

Whatever their reasons, Nvidia abandoned compute performance, tessellation performance, AA performance, and maybe even more that I'm forgetting all for sheer standard performance without such great enhancements. Heck, the Radeon 7970 is about 5 and a half times faster than the GTX 680 for dual-precision compute performance. That's a more than 600% advantage! Any dual-precision benchmark (such as the several such done by Tom's) will show this.

You know what? I can simply look at the very exact hardware specs and I could have told you all this stuff. I'm that well-versed with the technology. I'm not bragging about it, simply stating I fact. I'm good when it comes to doing the performance math. I'm not perfect, but I'm good at it. I said how it is and explained what some bench-marking sites failed to do. Take this as you will. Considering that many (perhaps most or even all) of these bench-marking *experts* don't know crap about the hardware that they aren't told, that's saying something. They tell you what benchmarks say. Heck, you people probably have no idea how many times I've explained to people exactly what is wrong with the Bulldozer based CPUs on a design and even architectural level.

I tell you what they mean and when there are problems. Take it as you will. Ignore me, even ridicule me, if you want, but this is how it is. I told people exactly how the 660 TI would perform in specific tests before I even saw them and they didn't even believe me after they saw the tests, so I don't expect any of you people to believe what I'm saying at this point. The same was true with the 7950 in overclocking, with the 670 when it first came out, and more. Few people believe me and they often try to not believe me even after the benchmarks tell otherwise.

It's that simple. The 660 TI can't do AA as well as, it can't do tessellation as well as, it can hardly do DirectC at all, it can't overclock as well as, and it has lower minimum frame rates than a Radeon 7950 or overclocked RAdeon 7850 with the memory pushed a little. The tests all show this when they are isolated and analyzed. Heck, I could even do the math on how Tom's slower CPU affected their results as well as their slightly higher focus on AA, overclocking, and such.

But why even go on like this? Chances are that none of you will read this anyway. Whatever, I'm done here.
Share
August 20, 2012 2:07:08 AM

blazorthon said:
I've had enough of this.
http://media.bestofmicro.com/U/P/336769/original/tessel...
There, undeniable proof from none other than Tom's themselves. Kepler has inferior tessellation performance.

All of the reviews, not just the Tom's that list minimum frame rates, show the 660 TI's inferior minimum frame rates. All of the reviews that have comparable AA on/off between tests show the 660 TI doing much worse with AA than AMD's cards, not just Tom's. Tom's simply had a slower CPU, so the 660 TI's maximum frame rates weren't able to stretch as far. They can go almost as far as the GTX 670's can because the 670 and the 660 TI have the exact same GPU, but the 660 TI's weaker minimum frame rates are a result of its weaker memory bus.

It performs inconsistently at best, especially with differently performing CPUs and levels of AA and tessellation. I don't even need to get benchmarks out to tell you how consumer Kepler GPUs are crap for compute compared to GCN GPUs simply because they aren't designed for it whereas the GCN architecture (especially the Tahiti implementation) is designed specifically for compute. Its great tessellation performance is also because that's what it is designed for. I don't even want to guess as to why Nvidia thought it was a good idea to make the horribly unbalanced Kepler GPUs.

Maybe they thought that they could trick people into upgrading more often than they should have to upgrade to retain high quality graphics for the time, maybe they actually had a good reason and decided against telling anyone or even making it obvious. Nvidia sacrificed everything that they could for raw pixel pushing and it obviously backfires when you actually put the GPUs through more severe settings rather than merely giving them more pixels to churn through. For what they can do, Kepler GPUs truly are the best. Not even Tahiti can come close to the considerably smaller GK104 even when it is cut-down by one SMX in raw gaming performance without increasing the settings.

Whatever their reasons, Nvidia abandoned compute performance, tessellation performance, AA performance, and maybe even more that I'm forgetting all for sheer standard performance without such great enhancements. Heck, the Radeon 7970 is about 5 and a half times faster than the GTX 680 for dual-precision compute performance. That's a more than 600% advantage! Any dual-precision benchmark (such as the several such done by Tom's) will show this.

You know what? I can simply look at the very exact hardware specs and I could have told you all this stuff. I'm that well-versed with the technology. I'm not bragging about it, simply stating I fact. I'm good when it comes to doing the performance math. I'm not perfect, but I'm good at it. I said how it is and explained what some bench-marking sites failed to do. Take this as you will. Considering that many (perhaps most or even all) of these bench-marking *experts* don't know crap about the hardware that they aren't told, that's saying something. They tell you what benchmarks say. Heck, you people probably have no idea how many times I've explained to people exactly what is wrong with the Bulldozer based CPUs on a design and even architectural level.

I tell you what they mean and when there are problems. Take it as you will. Ignore me, even ridicule me, if you want, but this is how it is. I told people exactly how the 660 TI would perform in specific tests before I even saw them and they didn't even believe me after they saw the tests, so I don't expect any of you people to believe what I'm saying at this point. The same was true with the 7950 in overclocking, with the 670 when it first came out, and more. Few people believe me and they often try to not believe me even after the benchmarks tell otherwise.

It's that simple. The 660 TI can't do AA as well as, it can't do tessellation as well as, it can hardly do DirectC at all, it can't overclock as well as, and it has lower minimum frame rates than a Radeon 7950 or overclocked RAdeon 7850 with the memory pushed a little. The tests all show this when they are isolated and analyzed. Heck, I could even do the math on how Tom's slower CPU affected their results as well as their slightly higher focus on AA, overclocking, and such.

But why even go on like this? Chances are that none of you will read this anyway. Whatever, I'm done here.


I read it and agree with you.
m
0
l
August 20, 2012 4:15:35 AM

blazorthon said:
I've had enough of this.
http://media.bestofmicro.com/U/P/336769/original/tessel...
There, undeniable proof from none other than Tom's themselves. Kepler has inferior tessellation performance.

All of the reviews, not just the Tom's that list minimum frame rates, show the 660 TI's inferior minimum frame rates. All of the reviews that have comparable AA on/off between tests show the 660 TI doing much worse with AA than AMD's cards, not just Tom's. Tom's simply had a slower CPU, so the 660 TI's maximum frame rates weren't able to stretch as far. They can go almost as far as the GTX 670's can because the 670 and the 660 TI have the exact same GPU, but the 660 TI's weaker minimum frame rates are a result of its weaker memory bus.

It performs inconsistently at best, especially with differently performing CPUs and levels of AA and tessellation. I don't even need to get benchmarks out to tell you how consumer Kepler GPUs are crap for compute compared to GCN GPUs simply because they aren't designed for it whereas the GCN architecture (especially the Tahiti implementation) is designed specifically for compute. Its great tessellation performance is also because that's what it is designed for. I don't even want to guess as to why Nvidia thought it was a good idea to make the horribly unbalanced Kepler GPUs.

Maybe they thought that they could trick people into upgrading more often than they should have to upgrade to retain high quality graphics for the time, maybe they actually had a good reason and decided against telling anyone or even making it obvious. Nvidia sacrificed everything that they could for raw pixel pushing and it obviously backfires when you actually put the GPUs through more severe settings rather than merely giving them more pixels to churn through. For what they can do, Kepler GPUs truly are the best. Not even Tahiti can come close to the considerably smaller GK104 even when it is cut-down by one SMX in raw gaming performance without increasing the settings.

Whatever their reasons, Nvidia abandoned compute performance, tessellation performance, AA performance, and maybe even more that I'm forgetting all for sheer standard performance without such great enhancements. Heck, the Radeon 7970 is about 5 and a half times faster than the GTX 680 for dual-precision compute performance. That's a more than 600% advantage! Any dual-precision benchmark (such as the several such done by Tom's) will show this.

You know what? I can simply look at the very exact hardware specs and I could have told you all this stuff. I'm that well-versed with the technology. I'm not bragging about it, simply stating I fact. I'm good when it comes to doing the performance math. I'm not perfect, but I'm good at it. I said how it is and explained what some bench-marking sites failed to do. Take this as you will. Considering that many (perhaps most or even all) of these bench-marking *experts* don't know crap about the hardware that they aren't told, that's saying something. They tell you what benchmarks say. Heck, you people probably have no idea how many times I've explained to people exactly what is wrong with the Bulldozer based CPUs on a design and even architectural level.

I tell you what they mean and when there are problems. Take it as you will. Ignore me, even ridicule me, if you want, but this is how it is. I told people exactly how the 660 TI would perform in specific tests before I even saw them and they didn't even believe me after they saw the tests, so I don't expect any of you people to believe what I'm saying at this point. The same was true with the 7950 in overclocking, with the 670 when it first came out, and more. Few people believe me and they often try to not believe me even after the benchmarks tell otherwise.

It's that simple. The 660 TI can't do AA as well as, it can't do tessellation as well as, it can hardly do DirectC at all, it can't overclock as well as, and it has lower minimum frame rates than a Radeon 7950 or overclocked RAdeon 7850 with the memory pushed a little. The tests all show this when they are isolated and analyzed. Heck, I could even do the math on how Tom's slower CPU affected their results as well as their slightly higher focus on AA, overclocking, and such.

But why even go on like this? Chances are that none of you will read this anyway. Whatever, I'm done here.



okay sir that was a very great information there.. I was hoping that gtx 660ti can easily beat out the 7950, now i know it can not even tackle the 7870 :) 
m
0
l
August 20, 2012 4:19:04 AM

tukbol7 said:
okay sir that was a very great information there.. I was hoping that gtx 660ti can easily beat out the 7950, now i know it can not even tackle the 7870 :) 
Look into it more don't rely on public forums ;) 
m
0
l
August 20, 2012 12:02:47 PM

now im really confused on what to buy hehehe
m
0
l
a c 87 U Graphics card
August 20, 2012 2:08:48 PM

Are you interested in overclocking?
m
0
l
August 21, 2012 5:16:54 AM

nope, dont know how to overclock a gpu, that,s why im getting the OCed. version.. with the recent price drop of AMD 7870 fot 249 i think its a no brainer? right over the 660ti?
m
0
l
a c 92 U Graphics card
August 21, 2012 7:35:56 AM

Yes. 7870 at $250 is a much much better deal imo.
m
0
l
a c 87 U Graphics card
August 21, 2012 7:46:07 AM

If you're not overclocking, then the 7870 is the best overall option out of the 7870, 7950, and 660 TI. If you would overclock, then the 7950 becomes the best option, although a 7870 can still overclock better than a 660 TI can by a large margin, especially a well-cooled 7870.
m
0
l
a b U Graphics card
August 21, 2012 10:05:03 AM

blazorthon said:
If you're not overclocking, then the 7870 is the best overall option out of the 7870, 7950, and 660 TI. If you would overclock, then the 7950 becomes the best option, although a 7870 can still overclock better than a 660 TI can by a large margin, especially a well-cooled 7870.


After our talk I tinkered with my HD7870. I got the memory to 1550 and took the core down to 1250. So I'm doing 1250/1550 but the kicker is, I can do this at 1.24v instead of the 1.28v when I was at 1300/1500. The performance is about 1 frame less in Unigine (only test I've done so far) so I think I'll keep this for an everyday OC.
m
0
l
a c 87 U Graphics card
August 21, 2012 6:43:37 PM

stickg1 said:
After our talk I tinkered with my HD7870. I got the memory to 1550 and took the core down to 1250. So I'm doing 1250/1550 but the kicker is, I can do this at 1.24v instead of the 1.28v when I was at 1300/1500. The performance is about 1 frame less in Unigine (only test I've done so far) so I think I'll keep this for an everyday OC.


Good job on that. It just goes to show that the memory can be far more important than some people seem to think that it is, not that we needed more proof of that at this point.
m
0
l
August 21, 2012 11:50:45 PM

now that another price drop from AMD in their 7800 cards, I'll just go for 7870 for $249 now than a little pricier 660 ti for $299.. thanks blazorthon :) 
m
0
l
!