Gigabyte 6850 vs 5870 - performance vs price analysis

Gigabyte 6850 vs 5870 - performance vs price comparison

http://i200.photobucket.com/albums/a...Untitled-1.jpg

http://i200.photobucket.com/albums/a...0/IMG_0742.jpg

These GPU's are not even in the same class, why compare them? New generation card, should be capable of higher FPS.. But the 6850 isn't exactually new generation card, in fact it probably fits in between the 5770 and 5850 somewhere, it is BUDGET GAMING, where as the 5870 is straight up power.

AMD has released the 6850 for one reason, it is this budget gaming sector. For a while now the 460 has been dominating them and we have already seen all the reviews show in anything but benchmarks/games that have heavy tesselation the 6850 is the new king of budget gaming.

I want to take you down a slightly different route, I want to look at what the right card is for that gaming sector. Obviously the 5870 is much more powerful and is soon to be related, but it also comes at a higher cost.. Is that cost worth it? Or are you better to go for something like the 6850? Well of course the FPS and your budget are going to decide that for you... But maybe we can shed a little light on performance along the way.

Taking a quick look at the card, this is a reference design of 6850 with a non-reference cooler. At this stage this is what alot of vendors are doing with the 6000 series, I am sure later we will see alot more non-reference cards with beefed up PWM and features coming onto the market. The 6850 only features one 6 pin PCI-E connector, it is a budget gaming card remember. The Gigabyte non-reference fan is both quiet and keeps the card cool, rarely did it get above 50 degrees celcius under long heavy loads. This fan is quiet but it does some damage also, fiddling around with the card while it was on resulted in this brutal injury.

http://i200.photobucket.com/albums/a...Untitled-2.jpg



Test Setup
Gigabyte 6850
AMD 5870
Gigabyte X58A-UD3R
Gulftown 980X @ 4GHz cpu, 3.2Ghz uncore
ADATA 1866 8-8-8-24 @ 1600 9-9-9-24

This is my standard comparison configuration





Benchmarks
* Vantage Performance - something most overclockers are familiar with and a good comparitive points against all cards
* Vantage Xtreme - Showing off the raw GPU power, taking the CPU out of the equation
* 3DMark03 - A good blend of CPU and GPU
* Crysis AVG FPS @ Veryhigh - Can it play crysis? Always need to include this one in modern review
* Heaven HWBOT DX11- Another raw GPU based benchmark, get an idea of how this GPU goes in a DX11 environment with very high settings




Results


Vantage Performance

http://i200.photobucket.com/albums/a...0/vantagep.jpg

stock 6850
http://i200.photobucket.com/albums/a...k_vantagep.jpg

6850 @ 850/1150
http://i200.photobucket.com/albums/a..._vantapgep.jpg

stock 5870
http://i200.photobucket.com/albums/a...0_vantagep.jpg


Vantage Xtreme

http://i200.photobucket.com/albums/a...0/vantagex.jpg

stock 6850
http://i200.photobucket.com/albums/a...k_vantagex.jpg

6850 @ 850/1150
http://i200.photobucket.com/albums/a...c_vantagex.jpg

stock 5870
http://i200.photobucket.com/albums/a...0_vantagex.jpg


Crysis Avg FPS

http://i200.photobucket.com/albums/a...0/crysis-2.jpg

stock 6850
http://i200.photobucket.com/albums/a...ock_crysis.jpg

6850 @ 850/1150
http://i200.photobucket.com/albums/a...0oc_crysis.jpg

stock 5870
http://i200.photobucket.com/albums/a...870_crysis.jpg


Heaven

http://i200.photobucket.com/albums/a...heavendx11.jpg

stock 6850
http://i200.photobucket.com/albums/a...eaven_dx11.jpg

6850 @ 850/1150
http://i200.photobucket.com/albums/a...0oc_heaven.jpg

stock 5870
http://i200.photobucket.com/albums/a...870_heaven.jpg


3DMark03

http://i200.photobucket.com/albums/a...0/3dmark03.jpg

stock 6850
http://i200.photobucket.com/albums/a...stock_3d03.jpg

6850 @ 850/1150
http://i200.photobucket.com/albums/a...c_3dmark03.jpg

stock 5870
http://i200.photobucket.com/albums/a...870_heaven.jpg




Cost per Dollar analysis

Costing was done at cheapest price I could find to purchase each unit from staticice.com.au (Australia's pricing search engine)

5870 - $359
6850 - $219

http://i200.photobucket.com/albums/a..._perdollar.jpg

http://i200.photobucket.com/albums/a..._perdollar.jpg

http://i200.photobucket.com/albums/a...rdollarjpg.jpg




Max Overclocking

3DMark03 full pass was used as a test of max clock stability.

http://i200.photobucket.com/albums/a...850/core_2.jpg

http://i200.photobucket.com/albums/a...0/memory_2.jpg




Clarkdale Gaming Comparison

Just for kicks I put this card into my clarkdale system, set the same clocks and punched out a Crysis and Heaven.. To see is these games gain from a massive 980X or maybe they are fine with a little clarkdale and gigabyte H55N-USB3


http://i200.photobucket.com/albums/a...0/crysis-1.jpg

http://i200.photobucket.com/albums/a...850/heaven.jpg




Brief Analysis

It is clear looking across all the benchmark results that the 6850 is no 5870 and in most tests the 5870 is 20% of above the 6850. This was to be expected, they are different specifications and not in the same price range.

Things get interesting when we start to look at price vs performance results.

The 6850 is a winner in all these tests by a massive percentage. When overclocking the 6850 we see nearly a double price/performance ratio advantage in Crysis and about a 50% advantage in Heaven. This is what "budget gaming" is right here, a card that doesn't break the bank but can deliver a large percentage of the performance of the higher end cards.

Moving onto the Max Overclocking, this is all done on the stock cooler of course and without moving the stock voltages too far.

We can see the core scales up nicely and consistantly with voltage bumps and we were passing 3DMark03 and Vantage Xtreme with 1025 core, 1.225v and reaching a maximum temperature of 55 degrees. This is a fairly impressive clock.

The memory overclocks nicely, but doesn't seem to be impacted at all with volts. I was able to bench at 1200MHz on stock volts but bumping to 1.65 didn't increase these clocks at all. Perhaps with some more time and playing with these settings we could get the voltage to help.

The Clarkdale gaming comparison is fairly interesting.

Crysis has always been known as a game that scales reasonable with CPU power and cores. It seems there is a heavy GPU bottleneck even at 4GHz, as the clarkdale system actually comes out with the advance, be it so small it is within a tolerance of difference between runs.

As Heaven is heavily GPU bound, again there is very little difference, again within a tolerance of difference between runs.

These results give a clear picture, at least with these two game benchmarks. If you are running single card gaming chances are you are going to be heavily GPU bound with the newest games. If you are purchasing an expensive X58 system, with a 980X or bloomsfield CPU it might go to waste with one GPU. Multiple GPU configurations may be difference for these benchmarks, but that is testing for another day.




No subzero?

Not yet, card is prepped though. Just waiting on the ability to push past 1025MHz with software and I'm away!

http://i200.photobucket.com/albums/a...0/IMG_0793.jpg

http://i200.photobucket.com/albums/a...0/IMG_0794.jpg

http://i200.photobucket.com/albums/a...0/IMG_0799.jpg

http://i200.photobucket.com/albums/a...0/IMG_0798.jpg

http://i200.photobucket.com/albums/a...0/IMG_0795.jpg

http://i200.photobucket.com/albums/a...0/IMG_0796.jpg

Why testing 16xQ in Crysis? Doesn't make sense to me.

PS: in Crysis, 6850 @ 1000 performs more or less like a GTX470 @ 800.

why did you compare crysis max fps?
to me that doent matter at all when gaming, its all about reaching a comfort zone for av fps, which is different for everybody and every game, and then having min fps at or above the min fps comfort zone...

actually if i had to chose id even go for av fps below my comfort zone if it gives me min fps at or above my min fps comfort zone.

40av/30min>50av/20min

i dont understand the volt/oc scaling, there are two entries for 1.16 and 1.225?

Quote:

---

When overclocking the 6850 we see nearly a 100% price/performance ratio advantage in Crysis and about a 50% advantage in Heaven.

---

what do you mean? the 5870 has a 10-20% higher perf/dollar score than the oced 6850 :confused:

i think nobody is going to get a 5870 now, 6850 vs 460 or 6850 vs 5850 or 6850 vs 470 would have been more interesting imo... cause those are alternatives for people who might buy a 6850. and i have a feeling that especially with overclocking taking into account, a 5850 would be the best option right now...

atm 5850s cost only 10% (20-30$) more than 6850s/460s(1gb) but are faster at stock and from what ive seen so far overclock about the same as 5850s and 460s...

perf average@1920x1080 (tpu) - default clock - air oc with vgpu boost - oc %
GTX470 250$ - 100% - 607 - 800+ - 32%
HD5850 230$ - 90% - 725 - 950+ - 31%
HD6850 200$ - 86% - 775 - 1000+ - 29%
GTX460 210$ - 83% - 675 - 900+ - 33%

correct me if im wrong... but those are the numbers you can except from these cards when oced, and despite all the ocing hype for the 460s, they dont really oc more than any other card ;) nvidia just seeded the message and everybody followed... it DOES oc well, but so do all the other cards :D

i really think it makes more sense getting a 470 or 5850 than a 6850 or 6870 or 460 right now... to me the 6800 series looks like a complete joke... worse performance than 5800, same overclocking, same price... slightly better tesselation perfomance, slightly lower power consumption, but who cares, honestly... :D

now that prices of the 5850 and 470 have dropped, FINALLY, they are the cards to get... at least right now.
i dont think the 6900 series will change a lot, its aiming at 480/580 and the prices in that segment tend to not affect prices in the mainstream segment at all these days. if a 5850 or 470 is fast enough for you, now is just as good of a time buying it as when the 6900 series is out i think.

saving 20-50$ and going for a 6850 over a 5850/470 doesnt make sense to me... if you spend 210$ already you might as well spend 230-250, a mere 10-25% more, and you get 10-15% more performance. the only reason to go for a 6850 is if its fast enough for you and you dont NEED any extra performance... now whether or not that is the case depends on the person...

so far i havent seen a single good 6800 series review out there that hows what the max res and aa and quality settings are that you can play games at with this card *hint hint :D* but i guess your looking at it from a benching pov anyways :D

id love to see what setting and res and aa bonus you get for spending more and getting a better card.
nowadays all reviews are just about pure fps-bar p3nis show offs, sigh...
all reviews do is say A>B and A=good, but none of them answer the questions people have when visiting their sites and reading the reviews
should i get one?
do i need one?
what benefits do i get from upgrading to this?

heh, sorry for the long post man :D
thx for the benches and im looking forward to your ln2 numbers m8 :D

to me, a 6850 would be a great card at ~150$...
its better than a 460 already at 200, but its too close to a 5850 and 470, price wise, to really shine...
at 150-175 it would be almost 100$ cheaper and deliver 85%+ of the performance... THAT would be a great deal!

Quote:

---

Originally Posted by prava

Why testing 16xQ in Crysis? Doesn't make sense to me.

PS: in Crysis, 6850 @ 1000 performs more or less like a GTX470 @ 800.

---

this is the settings I always use for this testing, so i will continue to use a setting for comparison sake, if you can suggest settings that is more likely to be used by gamer, please let me know i can include in future testing

Quote:

---

Originally Posted by saaya

why did you compare crysis max fps?

---

i titled the graph wrong, you can see now it is AVG fps

Quote:

---

Originally Posted by saaya

i dont understand the volt/oc scaling, there are two entries for 1.16 and 1.225?

---

i have changed it just for you sacha, now it should be more clear

Quote:

---

Originally Posted by saaya

i think nobody is going to get a 5870 now, 6850 vs 460 or 6850 vs 5850 would have been more interesting imo...

---

i dont have those cards, i can only compare against what i have, so i take a slightly different approach :)

Quote:

---

Originally Posted by saaya

correct me if im wrong... but those are the numbers you can except from these cards when oced, and despite all the ocing hype for the 460s, they dont oc a lot more than any other card they compete with.

---

i am not sure what you mean, here i am showing 6850 stock, 6850 oc and 5870 only.. yes of course you can expect 900/950+ on any 5870 and it will kill 6850 @ 1000/1200

dammit, I was hopin' to see you crank that soulya boy past 900Mhz and run the benches, that non-ref cooler looks like it could take it

crank that soulja boy LOL

http://www.discobelle.net/wp-content.../07/soulja.jpg

okay i will run some benches at 1000/1200 in the next couple days

Quote:

---

Originally Posted by saaya

snip

---

check crossfire and compare ;)

6850cf is sometimes fast as 5870cf ;)

Quote:

---

Originally Posted by youngpro

crank that soulja boy LOL

http://www.discobelle.net/wp-content.../07/soulja.jpg

okay i will run some benches at 1000/1200 in the next couple days

---

awesome, be good to see you "watch me crank that robocop" :p

@saaya: comparing min fps in Crysis is totally stupid...because at the beginning of the bench everybody has a HUGE stutter for a really short period of time that jeopardizes the whole benchmark. Its so bad that you can improve the min fps up to a 20% just running the bench enough times (cause the stutter is kinda random), that is why I never take into account the min fps value for that benchmark.

Quote:

---

Originally Posted by youngpro

this is the settings I always use for this testing, so i will continue to use a setting for comparison sake, if you can suggest settings that is more likely to be used by gamer, please let me know i can include in future testing

---

It is pointless IMO because you will never use that setting. You will use AAx4 and thats it, so its better to test with realistic settings.

Quote:

---

Originally Posted by youngpro

i have changed it just for you sacha, now it should be more clear

---

cheers :D

did you run some benchmarks with the 5870 oced in the past by any chance? im curious how both cards scale when oced... is the gap the same when they are maxed out or maybe smaller? or even bigger?

and yeah, im curious how far you can push it on air with high volt, the heatsink looks good... but def do some ln2 runs first, just in case it dies with high volts on air somehow... :D

Quote:

---

Originally Posted by Tomasis

check crossfire and compare ;)
6850cf is sometimes fast as 5870cf ;)

---

didnt know that... interesting... even when the cpu isnt limiting? any links?
for gaming this still doent change anything though, sli works better than xfire in my experience and even sli sucks for gaming... some people dont notice it, but i feel a difference...

btw, how are prices in down under m8?
here in taiwan nvidia cards are way expensive for some reason...
ati cards 6xxx cards cost the same as in the us, but 470s still cost 10.000NT+ which is 320$+ and 460 1gb cards go for 8000NT+ which is 250$+
in the us ive seen 470s for 250$ and 460s for 210... weird huh?
and 5850s and 5870s still cost as much as they did a couple of weeks ago before the 6000 series launched... weird...

Quote:

---

Originally Posted by saaya

didnt know that... interesting... even when the cpu isnt limiting? any links?
for gaming this still doent change anything though, sli works better than xfire in my experience and even sli sucks for gaming... some people dont notice it, but i feel a difference.

---

stuttering eh?

I mixed up 5870cf with 5970 when I saw 6850cf is equal with 5970 in techpowerup review. 6870cf is about 10% faster than 5970. There in Anandtech review, 68xx is faster than 58xx at Cf, you can look that as single card 58xx is faster by many fps but the charts are reversed up or gap is closed much once it runs at CF. Too bad I dont see any data to compare as 5870cf in techpowerup. I know 5970 is downclocked. Anyway, I think in newer games with tessellation support, 6850cf can catch up to 5870cf.

It is interesting with another 6850 OC article in Anandtech and it shows how much mem clocking makes close gap between 6850 and 6870. Core clocking makes less difference, 850mhz can be enough when 950 is easily available at stock volt for 6850.

I think you know better sources for comparing cf charts.

Quote:

---

Originally Posted by prava

@saaya: comparing min fps in Crysis is totally stupid...because at the beginning of the bench everybody has a HUGE stutter for a really short period of time that jeopardizes the whole benchmark. Its so bad that you can improve the min fps up to a 20% just running the bench enough times (cause the stutter is kinda random), that is why I never take into account the min fps value for that benchmark.

---

The first run doesnt count towards your score, and the lower frame rate is because its not cached to the HDD yet.

Very nice and detailed review.

btw from where did you get this test bench?

http://i200.photobucket.com/albums/a...0/IMG_0742.jpg

Very nice review with lots of info youngpro - thank you very much:up:

Is that supposed to be the eye of Optimus Prime on the box? BTW, nice review!

lol nerd injury

"bandwidth exceeded" pics broken :(

photobucket sucks i dont know why ppl use that :banana::banana::banana::banana:

mate this is a MSI test bench, i got it for a christmas present a couple years ago,

IMO our good friend Andrea "dimastech", makes the best OC benches!

fLucking photobucket, i had to pay some money for my images to display, oh wells!

are you serious
use imageshack man and direct link so there are no popups for ppl
the images stay for years

Cool review youngpro :)

Notice how GPU-z read the 6850 correctly on the Clarkdale system ? Did you reset the mobo's bios after each card swap ?

nah i dont recall doing that mate, is there some issue there?