HD4850 Vs. 9800GTX w/AA, review inside .

[wolf2009]

http://www.hardware.fr/articles/724-1/preview-ati-radeon-hd-4850.html


looking at the AA charts, it looks like the AA hit has been fixed on the 4800's finally and it seems to solidly beat the 3870X2 and even beats the GTX260 w/ AA in some games. All at $199 wow.

Plz leave a thx if u found this post helpful .

EDIT : if anyone wants to buy this card below MSRP for $189 , look here Sapphire 4850 (http://fxvideocards.com/Sapphire-HD-Radeon-4850-p-16386.html)

The vendor is reputable.

[Darkmag]

Average FPS without AA. With AA the 4850 out paces the 9800GTX with a larger margin.

HD 4850 = 70.5
HD 3870 = 53.5
9800GTX = 69.7
GTX 260 = 80.2
GTX 280 = 95.1

[Darknova]

The more benches I see from the 4850, the more I fall in love with the card, and the more I want to see what the 4870 can do.

[wolf2009]

i think this card is going to do wat 8800gt did in last generation.

[mandelore]

The more benches I see from the 4850, the more I fall in love with the card, and the more I want to see what the 4870 can do.

then the more we cant quite believe what may be instore from the 4870x2 :D

[Darknova]

then the more we cant quite believe what may be instore from the 4870x2 :D

I hear that!

I think ATi have really done it this time. They don't have the highest performing single card out by any means, the GTX 280 still takes that crown, but the price difference, and the fact you can pick up 2!!! 4850s for the price of a single 280 that outperforms it...well that's just magical :)

[[I.R.A]_FBi]

when is this card due for release again?

my hd3870 is for sale!

[panchoman]

look like the new arch. and extra shaders really give it an edge when it comes to processing data

[Darknova]

_FBi;844349']when is this card due for release again?

my hd3870 is for sale!

23rd I think.

[lepra24]

Nvidea:( Ati :)

[oli_ramsay]

I've heard from vr-zone that they'll be released on the 25th. And wow they're really sorted out the AA on these beauties. I would love to see how the 4870s perform with their nearlly twice as fast GDDR5!!!

Ati=win!

[From_Nowhere]

ATI Radeon HD 4850 :), ATI Radeon HD 4870 :D, ATI Radeon HD 4870X2 :cool:

That is a good card for only ~ $200.

[[I.R.A]_FBi]

well .. im waiting with bated breath.

[[I.R.A]_FBi]

babelfish'd

http://66.196.80.202/babelfish/translate_url_content?.intl=us&lp=fr_en&trurl=http%3A%2F%2Fwww.hardware.fr%2Farticles%2F72 4-1%2Fpreview-ati-radeon-hd-4850.html

[imperialreign]

I knew ATI still had it in 'em! Nice to see that they're coming back full-force with this generation, and if the 4850s are performing this well, that bodes great news for the rest of the series.


<sigh> it's like seeing the release of the X1900 series all over again :D

[wolf2009]

if anyone wants to buy this card below MSRP for $189 , look here Sapphire 4850 (http://fxvideocards.com/Sapphire-HD-Radeon-4850-p-16386.html)

The vendor is reputable.

[Megasty]

To put the 4850 in the same league as those NV monsters just goes to show you what mid-range has become. Sure 2 of them can eat a 280 but when 1 can hold its own so well...I have to get my hands dirty too :D

[[I.R.A]_FBi]

if anyone wants to buy this card below MSRP for $189 , look here Sapphire 4850 (http://fxvideocards.com/Sapphire-HD-Radeon-4850-p-16386.html)

The vendor is reputable.

Are you sure? My 3870 goes for sale 2mrw.

[dark2099]

Now that the new Cat 8.6 drivers are out and are really increasing performance it seems, these cards should definitely do quite well.

[wolf2009]

_FBi;844466']Are you sure? My 3870 goes for sale 2mrw.

i'm sorry i didn't get what u want to ask ?

if u r asking about the vendor, they r completely legit. they have been coming up with awesome deals on 8800gtx and gts in the past few months .

[[I.R.A]_FBi]

well ima try and get it asap

[trt740]

http://www.hardware.fr/articles/724-1/preview-ati-radeon-hd-4850.html


looking at the AA charts, it looks like the AA hit has been fixed on the 4800's finally and it seems to solidly beat the 3870X2 and even beats the GTX260 w/ AA in some games. All at $199 wow.

Plz leave a thx if u found this post helpful .

EDIT : if anyone wants to buy this card below MSRP for $189 , look here Sapphire 4850 (http://fxvideocards.com/Sapphire-HD-Radeon-4850-p-16386.html)

The vendor is reputable.

not thats it's ment to but it really doesn't beat a 3870x2 as you say only in half of two bench marks and it really about the same as a 9800gtx. It will do well the chip thats gonna fly will be the 4870x2 and it looks to me the 4870 will beat a 3870x2.

[lemonadesoda]

Interesting test report. Here's a summary:

http://img.techpowerup.org/080619/Capture072.png

Observations,

1./ 4850 is on average 31.6% faster than 3870. I think we were expecting moew out of 800 shaders and 32 TMU. A bit disappointing IMO.
2./ Cannot match 3870x2 or GTX260
3./ STILL cannot play crysis at 1920x1600. OMG what's wrong with that game? LOL :roll:

[OEGUSAndy]

Interesting test report. Here's a summary:

http://img.techpowerup.org/080619/Capture072.png

Observations,

1./ 4850 is on average 31.6% faster than 3870. I think we were expecting moew out of 800 shaders and 32 TMU. A bit disappointing IMO.
2./ Cannot match 3870x2 or GTX260
3./ STILL cannot play crysis at 1920x1600. OMG what's wrong with that game? LOL :roll:

The average and index statisthically speacking are not acurate. You have to raport it everything to a base like raport it to the 3870 and then do the index. Else if lets say in one game the 280 has 300 fps and 3870 only 100 this will blow your entire calculation away.

My 2 cent

[lemonadesoda]

^^ your wish is my command

http://img.techpowerup.org/080619/Capture073.png
WOW. Big change in stats. LOL

nvidia must be enjoying this http://motownmuscle.com/forums/images/smilies/flip.gif

[OEGUSAndy]

280 gtx don´t really seems to have that big muscles :) only 28% faster than 4850 in gaming...

[lemonadesoda]

^^ if you take the table above as the "muscle test" benchmark, then the GTX280 is 39% faster (not 28%) than the 4850. That's not insignificant, since it is a bigger improvement over the 4850 (38.7%) than the 4850 is over the 3870 (31.9%)!

GTX280 >> 4850 than 4850 > 3870

[OEGUSAndy]

ok lets do it then.. we have to transform it in 182.9 base so:

100/182.9 *100 = 54.6 => 280 has 45.4% more muscle than 3870
next is 4850:
131.9/182.9 * 100 = 72.1 => 280 has 27.9% more muscle than 4850

comparing 4850 and 3870 in base 131.9(4850 base)
100/131.9 * 100 = 75.8 => 4850 has 24.2% more muscle that 3870

GTX280 >> 4850 than 4850 > 3870 your point is correct but not really acurate because we have to utilise or 260 GTX or 4870 :) that woult be only fair.

Andy

[DarkMatter]

ok lets do it then.. we have to transform it in 182.9 base so:

100/182.9 *100 = 54.6 => 280 has 45.4% more muscle than 3870
next is 4850:
131.9/182.9 * 100 = 72.1 => 280 has 27.9% more muscle than 4850

comparing 4850 and 3870 in base 131.9(4850 base)
100/131.9 * 100 = 75.8 => 4850 has 24.2% more muscle that 3870

GTX280 >> 4850 than 4850 > 3870 your point is correct but not really acurate because we have to utilise or 260 GTX or 4870 :) that woult be only fair.

Andy

You did it wrong or must I say you interpreted it wrong. By your calculations (i.e 54.6) the HD3870 has 45.4% LESS muscle than the 280 and not the other way. Just compare 75 and 100 numbers to a baseline of 50 annd then the other way around.

100/50 * 100 = 200% >> 100 is double of 50, right.

75/50 * 100 = 150% >> and 75 is 50% more than 50, right again

YOUR way:

50/100 * 100 = 50% >> 100 is 50% more than 50, WRONG

50/75 * 100 = 67% >> 75 is 33 % more than 50, wrong again.

EDIT: Anyway IMO both RV770 and GT200 (as well as faster G92 cards) are performing less in games than what their actual performance is. The higher you go in the stack the bigger the impact is, lesser the performance compared to it's full potential. This is because most games are using engines 5 years old!! Almost all engines used in the games are based on Doom3, Source and UE2. Even Unreal Engine 3 is very little more than a revamped UE2 really. Take into account those engines and games were created with consoles in mind that use technology almost 8 years old!! With more horsepower, with unified shaders in the case of the Xbox but rather old nonetheless.

EDIT2: Uff I exagerated a bit. I thought Radeon 9700 was of around year 2000. Sometimes the time doesn't pass as fast as we could first think, but there's been so many cards in between...

[DarkMatter]

^^ your wish is my command

http://img.techpowerup.org/080619/Capture073.png
WOW. Big change in stats. LOL

nvidia must be enjoying this http://motownmuscle.com/forums/images/smilies/flip.gif

After a second thought, the results in fact should be the same. The difference is probably in the rounding, but WOW the difference is big. Almost 5% in the case of the GTX280!

[wolf2009]

guys this is not a stats class, everything is going over my head . LOL .

[OEGUSAndy]

DarkMatter i think both are correct. It depends of the viewer. In the chart that lemonadesoda did, i would use as a baseline 280GTX as 100% then you could really see the diferencte in %. But using 3870 is the same thing only diferent perception.

Andy

Edit:
guys this is not a stats class, everything is going over my head . LOL .

Youre tottaly right. We should wait one week or so and with all the data we should do it again. Then will se. Wee could do it with price performance and everyone will se it. The best worst buy :)

[Megasty]

guys this is not a stats class, everything is going over my head . LOL .

They're just doing what NV & AMD did with their respective propaganda. Taking a baseline as the 3870 & using the other cards as a 100+ percentile over that. It doesn't work unless you use each card as a baseline & then work a percentile scale for the other cards. Trying to compare the 4850 & GTX280 from the 3870 only works in dunt-da-dunt land. I didn't spend half my life in a classroom for nothing :p

[EastCoasthandle]

http://i11.photobucket.com/albums/a154/eastcoasthandle/4800%20Series/4850numbersgi3.jpg
I took the data from the 9800 GTX, Ultra, and 260 to allow for a better comparison.

[DarkMatter]

They're just doing what NV & AMD did with their respective propaganda. Taking a baseline as the 3870 & using the other cards as a 100+ percentile over that. It doesn't work unless you use each card as a baseline & then work a percentile scale for the other cards. Trying to compare the 4850 & GTX280 from the 3870 only works in dunt-da-dunt land. I didn't spend half my life in a classroom for nothing :p

I am not doing anything. I'm just correcting a conception that is inevitably wrong. As I said in the above 100 is 100% more or double of 50, 50 is 50% of 100 and/or 50% less than 100 too, 100 is NOT 50% more than 50, that's 75. Period.

[lemonadesoda]

I didn't spend half my life in a classroom for nothing :pI think you just did. ROFL

[Megasty]

I am not doing anything. I'm just correcting a conception that is inevitably wrong. As I said in the above 100 is 100% more or double of 50, 50 is 50% of 100 and/or 50% less than 100 too, 100 is NOT 50% more than 50, that's 75. Period.

I'm just saying if you want to compare the 4850 & GTX280 by indexing them then use the 4850 as 100%. The first chart w/o the indexing has nothing wrong but the indexed chart is just comparing the other cards with the 3870. I'm only saying this because I made the same mistake on a research paper back in the day & the GD professor tore it up in my face :D

I think you just did. ROFL

My 4hr work days & my bank account says differently :D

[Azazel]

man..i cant wait for the 4870x2

[lemonadesoda]

http://img.techpowerup.org/080619/Capture075.png

1./ The 4850 "is a" 9800GTX overall, with a small +/- in any ONE game
2./ The 4850 is 31.9% faster than the 3870
3./ The GTX280 is 39.1% faster than the 4850

[lemonadesoda]

My 4hr work days & my bank account says differently :D
I believe in job sharing ;) LOL Gimme gimme
P.S. So long as you arent in finance... that's OK... or there will probably be another banking crisis coming if you're at the helm. LOL
P.P.S. I cant believe you guys are making me recut these numbers. Nothing has changed by more than a few marginal % (less than a driver update IMO). But anyway, there you go.

[Megasty]

http://img.techpowerup.org/080619/Capture075.png

1./ The 4850 "is a" 9800GTX overall, with a small +/- in any ONE game
2./ The 4850 is 31.9% faster than the 3870
3./ The GTX280 is 39.1% faster than the 4850

...& that's the answer :respect: That 4850 is a fighter. Even if its a small sample size, the overall difference between it & the 9800GTX are nil. Increasing the number of games will only yield the same results. Everyone is going nuts over how CF is beating the GTX280 when it actually should do much better than that. It also means that CFx still scales like garbage :ohwell:

[Ravenas]

Those cards are cheap enough. Go ahead and buy 2 or 3 guys :)

[Disruptor4]

Interesting test report. Here's a summary:

http://img.techpowerup.org/080619/Capture072.png

3./ STILL cannot play crysis at 1920x1600. OMG what's wrong with that game? LOL :roll:
Performance and scaling in Crysis is poor. Don't expect much to run it.
This is a post on another forum:
If they had done what they said they were going to do, Crysis would have been fine.

Crysis does not run on computers 3 years old. It barely runs on computers that are 3 days old.

I could put my settings on 800x600, no AA, everything low, and i would still get lag in multiplayer, with an 8800GTX and an overclocked Quad core. They failed to deliver on performance scaling and that was it's biggest downfall, because hardly anyone could play it, or would put up with the terrible frame rates with no second option.

I made this sarcastic picture to send to a friend back when the BETA was on. It still holds true today:
http://img443.imageshack.us/img443/8550/crysisscaleqh3.th.jpg (http://img443.imageshack.us/my.php?image=crysisscaleqh3.jpg)

[OEGUSAndy]

I wonder what image will have with some mature drivers from Ati. :) Time will tell.
Those cards are cheap enough. Go ahead and buy 2 or 3 guys :)
Not yet. I would say that the smarthest thing is to wait till August till the 4870x2 gets out and then decide.

[Megasty]

I wonder what image will have with some mature drivers from Ati. :) Time will tell.

Not yet. I would say that the smarthest thing is to wait till August till the 4870x2 gets out and then decide.

Never :cry: Hell its been 4 friggin months for the 3870x2 & its still getting noticeable performance increases. That PLX chip is bottlenecting the hell out of that thing. The subtle increases are coming from the drivers allowing data to move move efficiently through the PLX :rolleyes:

The 4870x2 won't have that problem but its architecture is still an experiment which will see stupid performance increases from 2 yrs down the line :wtf:

[lemonadesoda]

I'm updating the table to include the 1600x1200 results from w1z. If anyone spots any mistakes in the table, let me know.

(UPDATED)

http://img.techpowerup.org/080619/Capture079.png

GTX280 averaging 49.2% 38.7% faster than 4850. Big drop due to bizarre Quake4 results

[DarkMatter]

I'm updating the table to include the 1600x1200 results from w1z. If anyone spots any mistakes in the table, let me know.

http://img.techpowerup.org/080619/Capture077.png

Hmm Wizz's results are a lot more favorable to both new cards than the other one.

EDIT: It's a lot more favorable to soposedly faster cards, not only new ones. Kind of fits with my previous claim of system bottleneck?

[yogurt_21]

it seems like this will definetly be the cad to buy at 200$ we'll see how the 4870 come in on price/performance but thus far from what I'm seeing we finally have something that beats the 8800gt in that category.

[DarkMatter]

it seems like this will definetly be the cad to buy at 200$ we'll see how the 4870 come in on price/performance but thus far from what I'm seeing we finally have something that beats the 8800gt in that category.

8800 GT is a lot cheaper AFAIK. The price is more on the line of the GTS, this one being overpriced due to the lack of compentence. They will come down, while I would expect Radeon prices to go up. It's a better buy at $200 anyway if only because of the power consumption and noise. Temps are way too high on the other hand, hopefully fan speeds can be turned up easily.

[yogurt_21]

8800 GT is a lot cheaper AFAIK. The price is more on the line of the GTS, this one being overpriced due to the lack of compentence. They will come down, while I would expect Radeon prices to go up. It's a better buy at $200 anyway if only because of the power consumption and noise. Temps are way too high on the other hand, hopefully fan speeds can be turned up easily.

price/performance ie 200$/performance index vs the 8800gt's 170$/performance index. thus far the 4850 easily has a better ratio.

[Ninkobwi]

wonder how well these 4850s will overclock...gah I cannot wait

[flyin15sec]

if anyone wants to buy this card below MSRP for $189 , look here Sapphire 4850 (http://fxvideocards.com/Sapphire-HD-Radeon-4850-p-16386.html)

The vendor is reputable.

Nice link, I checked this morning, they had 0 pieces. Just checked again, about 10:40CST and they have 22 pieces now.

Still debating, can't decide between these cards:

9800GTX - XfX $244.99 /w rebate $214.99 Tigerdirect
8800GTS - ECS $209 /w rebate $159 Newegg
HD4850 - Sapphire $189 FXcards

I'm only going from a 9600GT, but damn I am tired of using an underclocked 9600GT because of the Blackscreen OCP.

Might have to wait a bit longer for all the cards to be out then see if there's any good deals.

[DarkMatter]

price/performance ie 200$/performance index vs the 8800gt's 170$/performance index. thus far the 4850 easily has a better ratio.

I have seen plenty of GTs below $150, a pair or three as low as $120 after rebate. That's what I was saying not $170.

EDIT: And actually, it seems you can have a HD4850 for $189. 8800 GT below $150 has still a quite better perf/price ratio. That being said, I would take the Radeon any day.

[yogurt_21]

yeah just took a look at the egg and they're averaging 150-160 prior to mail in. so it seems the 8800gt remains the price/performance king lol. wow a full gen later and still kicking.

[HTC]

http://www.hardware.fr/articles/724-1/preview-ati-radeon-hd-4850.html


looking at the AA charts, it looks like the AA hit has been fixed on the 4800's finally and it seems to solidly beat the 3870X2 and even beats the GTX260 w/ AA in some games. All at $199 wow.

Plz leave a thx if u found this post helpful .

EDIT : if anyone wants to buy this card below MSRP for $189 , look here Sapphire 4850 (http://fxvideocards.com/Sapphire-HD-Radeon-4850-p-16386.html)

The vendor is reputable.

http://img.techpowerup.org/080619/2008-06-19_165641.png

4850 and 9800 GTX are really close, with the nVidia card very slightly ahead.

Now, notice what happens when you increase the details @ the same resolution:

http://img.techpowerup.org/080619/2008-06-19_165654.png

The 9800 GTX loses by nearly 14%: quite significant, IMO.


And now, in an even higher resolution:


http://img.techpowerup.org/080619/2008-06-19_165713.png

In here, the GTX is losing by just 5.4%: a small lead for the 4850 card.

Now, notice what happens when you increase the details @ the same resolution:

http://img.techpowerup.org/080619/2008-06-19_165722.png

The 9800 GTX loses by a whopping 41.7%: WOW doesn't quite cover it, really!



I had posted this (modified it slightly, for this thread) in here (http://forums.techpowerup.com/showthread.php?t=63344&page=2) (post #39).

[DarkMatter]

Anyway after seeing the later reviews I can't help but think that this generation (at least right now) is better to Crossfire/SLI a pair of 3850s / 3870s / 9600GTs / 8800GTs than buying any of the new cards, from the pure perf/price point. A direct consecuence of the great improvements on multi-GPU configuration by both camps for sure. Funny how a new generation of cards has to come out for us to be able to see that multi-GPU scaling was NOT the problem after all.

[Ninkobwi]

up next: OC'd 4850 VS OC'd 9800GTX?

[p_o_s_pc]

I wanty the 4850. I think i am going to wait till i get the money for one of them... looks to be a sick card

[DarkMatter]

http://img.techpowerup.org/080619/2008-06-19_165641.png

4850 and 9800 GTX are really close, with the nVidia card very slightly ahead.

Now, notice what happens when you increase the details @ the same resolution:

http://img.techpowerup.org/080619/2008-06-19_165654.png

The 9800 GTX loses by nearly 14%: quite significant, IMO.


And now, in an even higher resolution:


http://img.techpowerup.org/080619/2008-06-19_165713.png

In here, the GTX is losing by just 5.4%: a small lead for the 4850 card.

Now, notice what happens when you increase the details @ the same resolution:

http://img.techpowerup.org/080619/2008-06-19_165722.png

The 9800 GTX loses by a whopping 41.7%: WOW doesn't quite cover it, really!



I had posted this (modified it slightly, for this thread) in here (http://forums.techpowerup.com/showthread.php?t=63344&page=2) (post #39).

That already happened with HD3000 cards and frankly is a good achievement. But going above 4x is pointless. At 1600/1920 x 1200 even 4x is kind of innecesary, 2x is enough to make dissapear the more pronounced edges when they change their position slightly (when they are more noticeable). I know there's people who would claim they can see the difference up to 16x @2560x1600 while they are playing COD4 MP and that they find annoying anything below that when they are gaming. The point is 99% of people won't, and it's those people who will pay for the cards. I just prepare the ground before the possible responses. :D

[HTC]

up next: OC'd 4850 VS OC'd 9800GTX?

That would be of interest. Wonder if / when such bench will emerge ...

[magibeg]

Well looking at wizards review the 4850 only overclocked 9% which he believes may be because the higher binned chips are going into the 4870.

[DarkMatter]

That would be of interest. Wonder if / when such bench will emerge ...

Hmm that scenario doesn't look very promising for the Radeon card I fear. Wizz couldn't take more than 675 Mhz (8% overclock) out of the card.

EDIT: magibeg beat me to it.:mad:

[HTC]

Hmm that scenario doesn't look very promising for the Radeon card I fear. Wizz couldn't take more than 675 Mhz (8% overclock) out of the card.

EDIT: magibeg beat me to it.:mad:

Maybe a better cooler can help?

[DarkMatter]

Maybe a better cooler can help?

Sure. But low OC on stock cooler is not a good symptom to begin with. And Ati stock coolers are usually very good ones. I dunno, I hope non reference cards OC better for the sake of competition.

[HTC]

Sure. But low OC on stock cooler is not a good symptom to begin with. And Ati stock coolers are usually very good ones. I dunno, I hope non reference cards OC better for the sake of competition.

Signed!

[eidairaman1]

OC is a hobby, not a must, they provide the cooler as a stock unit to keep prices down, but from looks of it the 4850 is far more powerful yet probably occupies the same Thermal Dynamics of the Older Part.

[lemonadesoda]

UPDATED FIGURES

OK, final results of summary scores from 34 sources:

http://img.techpowerup.org/080619/Capture086.png

Results of ATI vs. ATI

1./ 4850 is 31.9% faster than 3870 when scaling to a high resolution with low shader demand (Hardware.fr at 1920x1200)
2./ About 45.5% faster than 3870 at über resolutions like Apple Cinema 30" (Erenumerique.fr at 2560x1600)
3./ About 48.1% faster than 3870 at common resolutions with low shader demand (TPU at 1600x1200)
4./ About 64.8% faster than 3870 at common resolutions with high shader demand (Computerbase.de 1600x1200).

Conclusion

A. 4870 SOLVES the performance brick wall that hit 3870 at 1600x1200 when scaling shaders
B. 4870 improves but doesnt solve performance scaling to very high resolutions e.g. 1920x1200 and above

Results of nVidia vs. nVidia

1./ Across all benchmarks 8800GTX and 9800GTX are identical (although some differences in any one test)
2./ GTX280 is 60.0% faster on average than 8800GTX/9800GTX

Results of ATI vs. nVidia

1./ 4850 is faster than 8800/9800 by 4.9% on average
2./ GTX280 remains king by a lead of 50.8% over 4850
3./ At über resolutions 2560x1600 GTX280 is 70% faster than 4850

Conclusion ATI vs. nVidia
A. With GTX280>4850 by 50.8% and 4850>3870 by 45.5% then GTX280 is approximately "one generation ahead" in performance terms over 4850
B. With price point of $199 for 4850, nVidia need to get 8800/9800 down to $199 pretty fast or everyone (in the know) will move over to ATI
C. The GTX280 is simply an amazing performer in every case, even more so at über resolutions like Cinema 30"

***
.xls attached for anyone that wants to play with the figures

[yogurt_21]

1./ 4850 is faster than 8800/9800 by 7.2% on average
2./ GTX280 remains king by a lead of 44.4% over 4850

Conclusion ATI vs. nVidia
A. With GTX280>4850 by 44.4% and 4850>3870 by 48.2% then GTX280 is approximately "one generation ahead" in performance terms over 4850
.xls attached for anyone that wants to play with the figures

nota bad synopsis but you forget one thing in your generational comparison. the 4850 is the replacement for the 3850 the 4870 will be the replacemnt for the 3870

thus in order to make a generational comparison you have to compare the card against it's replacement. so in reality the 4850 is more like 60% faster than the 3850 meaning the gtx280 is not a fiull generation out from the 4850. we'll see how the 4870/gtx260 battle goes next week. that to me seems to be the more important one.

[MrHydes]

Preview : Radeon HD 4850 & GeForce 9800 GTX v2

http://img154.imageshack.us/img154/3537/img0023593qu6.gif (http://imageshack.us)http://img46.imageshack.us/img46/8293/img0023586ny4.gif (http://imageshack.us)

http://img209.imageshack.us/img209/1193/img0023585lz3.gif (http://imageshack.us)http://img209.imageshack.us/img209/7024/img0023584qv9.gif (http://imageshack.us)

http://img46.imageshack.us/img46/442/img0023587eb1.gif (http://imageshack.us)http://img76.imageshack.us/img76/3408/img0023582xq5.gif (http://imageshack.us)

http://img515.imageshack.us/img515/6809/img0023581ua1.gif (http://imageshack.us)http://img154.imageshack.us/img154/4485/img0023588hq1.gif (http://imageshack.us)

http://img154.imageshack.us/img154/9010/img0023589mm0.gif (http://imageshack.us)http://img154.imageshack.us/img154/8122/img0023590zh7.gif (http://imageshack.us)

Source: hardware.fr (http://www.c/articles/724-1/b-maj-b-preview-radeon-hd-4850-geforce-9800-gtx-v2.html)

gtx280 better n' better

[yogurt_21]

That already happened with HD3000 cards and frankly is a good achievement. But going above 4x is pointless. At 1600/1920 x 1200 even 4x is kind of innecesary, 2x is enough to make dissapear the more pronounced edges when they change their position slightly (when they are more noticeable). I know there's people who would claim they can see the difference up to 16x @2560x1600 while they are playing COD4 MP and that they find annoying anything below that when they are gaming. The point is 99% of people won't, and it's those people who will pay for the cards. I just prepare the ground before the possible responses. :D

lol see the funny thing is I remember an ati fanboy saying the same thing when the 2900xt couldn't process aa so well. imo it's a dumb argument. just because you assume *most* people wont use aa, doesn't mean they don't and turning up the pretties is the whole point of buying a new graphics card. I can play any game I want in dx9 with aa off with my 2900, if I wasn't wanting dx10 with aa I wouldn't upgrade. ATI not only fixed their aa issue, they are now beating nvidia in it and throwing that away just because you don't think the 4850 is worth it isn't a good argument, it sounds alot like the ati fanboy. and 1280x1024 with 8xaa is a popular gaming setting at which the 9800gtx loses by 14% 1600x1200 with 8x aa is less likely for the guy getting low frames, but the instant the 4850 user sees that he can game at those settings, you can bet he's going to use it. it's not a matter of what most people do because most people don't have the luxury of a card/monitor that can game at those settings and you can bet that if they did have the hardware to do it, they would.

[yogurt_21]

gtx280 better n' better

is that an actual v2 or just a 9800gtx clocked to v2 levels?

[DarkMatter]

lol see the funny thing is I remember an ati fanboy saying the same thing when the 2900xt couldn't process aa so well. imo it's a dumb argument. just because you assume *most* people wont use aa, doesn't mean they don't and turning up the pretties is the whole point of buying a new graphics card. I can play any game I want in dx9 with aa off with my 2900, if I wasn't wanting dx10 with aa I wouldn't upgrade. ATI not only fixed their aa issue, they are now beating nvidia in it and throwing that away just because you don't think the 4850 is worth it isn't a good argument, it sounds alot like the ati fanboy. and 1280x1024 with 8xaa is a popular gaming setting at which the 9800gtx loses by 14% 1600x1200 with 8x aa is less likely for the guy getting low frames, but the instant the 4850 user sees that he can game at those settings, you can bet he's going to use it. it's not a matter of what most people do because most people don't have the luxury of a card/monitor that can game at those settings and you can bet that if they did have the hardware to do it, they would.

I knew it will happen. 8x vs. 4x AA is not the same as 4x vs. 0x AA. 14% is not the same as 50%. Not going to say anything more.

[rockfella]

has someone considered the load temps of HD4850 compared to 9800gtx??????????????????????????????????????????? ?????????????????????????????????????????????????? ??????????????????

[Ninkobwi]

has someone considered the load temps of HD4850 compared to 9800gtx??????????????????????????????????????????? ?????????????????????????????????????????????????? ??????????????????

yes. 4850 runs hotter. whats the problem? its probably higher quality materials otherwise ATI wouldnt consider running it that high.

[rockfella]

Not really, we you are "comparing cards" keeping in mind generations. Compare price VS value. I dont care if i am buying a "generation card" for more or less, all i would care is erformance/price.

nota bad synopsis but you forget one thing in your generational comparison. the 4850 is the replacement for the 3850 the 4870 will be the replacemnt for the 3870

thus in order to make a generational comparison you have to compare the card against it's replacement. so in reality the 4850 is more like 60% faster than the 3850 meaning the gtx280 is not a fiull generation out from the 4850. we'll see how the 4870/gtx260 battle goes next week. that to me seems to be the more important one.

[erocker]

Not really, we you are "comparing cards" keeping in mind generations. Compare price VS value. I dont care if i am buying a "generation card" for more or less, all i would care is erformance/price.

Yes sir you are completely correct in my book! There are "the fringe" that care only about performance vs. performance, but when it comes down to the average consumer it's all about price vs. performance.

[yogurt_21]

Not really, we you are "comparing cards" keeping in mind generations. Compare price VS value. I dont care if i am buying a "generation card" for more or less, all i would care is erformance/price.

the 3850 didn't launch at 120$ it launched at closer to 200$ taking a peek around the 4850 can be had as low as 170$ so yest it really is the replacement for the 3850 the 4870 will launch closer to 300$ and will likely be found as cheap as 260$ at some places which puts it close to where the 3870 launched. you can't compare launch price to year old price. lol

[yogurt_21]

I knew it will happen. 8x vs. 4x AA is not the same as 4x vs. 0x AA. 14% is not the same as 50%. Not going to say anything more.

you're not making much sense here. both cards were ran with the same settings. so I really don't see where you're going with this.

[DarkMatter]

you're not making much sense here. both cards were ran with the same settings. so I really don't see where you're going with this.

My claims that going above 4x AA on high resolutions doesn't make much sense are a lot more legitimate than those to which you were comparing. Same with the performance differences, now we are talking about much smaller differences..

[rockfella]

It is strange and also nice how HD4850 scales so well as screen size increases!

[btarunr]

Ok, the OC does help the GTX+ out.

http://www.hardware.fr/medias/photos_news/00/23/IMG0023591.gif

Do you still need more power input (2x 6-pin)?

[yogurt_21]

My claims that going above 4x AA on high resolutions doesn't make much sense are a lot more legitimate than those to which you were comparing. Same with the performance differences, now we are talking about much smaller differences..

gonna have to ask for some screenies, I listed 1280x1024 with 8x aa, wanna show me 1280x1024 with 4x aa that looks as good ? 1680x1050 I notice a difference between 4x and 8x, perhaps you're screen is in need of updating. and I definetly notice a difference between 0aa and 4x aa. I mean seriously this is exactly what the ati fanboys were saying when the 2900 didn't fare so well, yet I seem to recall the normal people showing them that IT DOES matter. we dont buy these cards to not turn on the pretties. that'd be like getting an xbox 360 to play pong. it's ridiculous to say "ooo I don't like ati so I'm going to find any victory they have and say it doesn't matter because nvidia rules." the 4870x2 will likely beat the gtx280 and the nvidia fanboys say "ooo but it takes 2 core to do it" wtf does that have to do with the rice in china? the 4850 beats the 9800gtx with aa on and the nvidia fanboys say "oooo but aa doesn't matter" and thats the same nvidia fanboys who were saying aa was everything when the 2900 couldn't run it well. the 4870 will compete against the gtx260 and if it beats it the nvidia fanboys will say "ooo but the gtx 260 is cooler"

I mean seriously it's like you're hanging on to an argument that was disproven just so you can discredit one company over another. imo it's dumb and childish. if aa didn't matter, aa reviews wouldn't be included, no games would support it and no cards would use it. IT DOES MATTER. which is why it is included and it's why the 2900 was such a bad buy. not because the 2900 lagged so far behing the 8800's in normal detail settings, but because it absolutely failed with the details turned up. the 9800gtx now sits in that same situation and just as the 2900 was it's a bad buy right now. before the 4850 launch it wasn't, with the 4850 out it is. It happens get over it.

[DarkMatter]

gonna have to ask for some screenies, I listed 1280x1024 with 8x aa, wanna show me 1280x1024 with 4x aa that looks as good ? 1680x1050 I notice a difference between 4x and 8x, perhaps you're screen is in need of updating. and I definetly notice a difference between 0aa and 4x aa. I mean seriously this is exactly what the ati fanboys were saying when the 2900 didn't fare so well, yet I seem to recall the normal people showing them that IT DOES matter. we dont buy these cards to not turn on the pretties. that'd be like getting an xbox 360 to play pong. it's ridiculous to say "ooo I don't like ati so I'm going to find any victory they have and say it doesn't matter because nvidia rules." the 4870x2 will likely beat the gtx280 and the nvidia fanboys say "ooo but it takes 2 core to do it" wtf does that have to do with the rice in china? the 4850 beats the 9800gtx with aa on and the nvidia fanboys say "oooo but aa doesn't matter" and thats the same nvidia fanboys who were saying aa was everything when the 2900 couldn't run it well. the 4870 will compete against the gtx260 and if it beats it the nvidia fanboys will say "ooo but the gtx 260 is cooler"

I mean seriously it's like you're hanging on to an argument that was disproven just so you can discredit one company over another. imo it's dumb and childish. if aa didn't matter, aa reviews wouldn't be included, no games would support it and no cards would use it. IT DOES MATTER. which is why it is included and it's why the 2900 was such a bad buy. not because the 2900 lagged so far behing the 8800's in normal detail settings, but because it absolutely failed with the details turned up. the 9800gtx now sits in that same situation and just as the 2900 was it's a bad buy right now. before the 4850 launch it wasn't, with the 4850 out it is. It happens get over it.

Blah blah blah.

4x antialiasing is relevant, more than that no. No at high resolutions and definately not in mainstream cards (chances are someone looking into sub $200 segment doesn't ave the best screen out there). Even reviews don't use more than that even when 6x and 8x has been available for many years. You can argue as much as you want, and bring in the same old and FALSE argument of fanboism. The truth is that both cards perform well, it is not me the one jumping to any oportunity to claim any superiority of one company over the other. The release of this card doesn't make G92 cards obsolete and that is what I'm saying. Ati is back on the game again and that is good, but in way is owning as every Ati fan is so fast to try to point out, that's all. It's funny how I didn't saw you saying so cathegorically that AA matters. On the other hand I have not changed of opinion, AA at higher resolutions is not very important for me (more than 2x) and is not for many people, in fact every Ati R600/670 owner shouldn't be so concerned if they chose that cards. What has changed in the last moth that makes AA so important? Ah, yes...

BTW GT200 can do AA better, did you brought the argument then, no? Well...

[lemonadesoda]

You have to agree with DarkM on this. While 16x AA is, in theory, better than 4x AA, the difference is very hard to notice unless you "investigate" a static screen. In action MP you really will not notice any quality difference at all. Other than a performance hit. And quality gains become even less noticable as the resolution of the TFT scales up, and pixel density increases. In theory, if we had 3200x2400 19" screens we wouldnt need 4x AA at all.

4x AA has killed most of the jaggies. It's hard to quantify the improvement using any numbers. Only a subjective interpretation of what you see. But the result is that 16x AA has a huge processing requirement for a very very marginal gain. And I mean very very very marginal.

In theory, we could demand 64x AA and 64xAF. But I think we would all agree that was nonsense. So we've established the principle. Decreasing "quality returns" on exponentially increasing AA and AF. So where is the best bang for the buck? Somewhere around 4x AA.

[yogurt_21]

Blah blah blah.

4x antialiasing is relevant, more than that no. No at high resolutions and definately not in mainstream cards (chances are someone looking into sub $200 segment doesn't ave the best screen out there). Even reviews don't use more than that even when 6x and 8x has been available for many years. You can argue as much as you want, and bring in the same old and FALSE argument of fanboism. The truth is that both cards perform well, it is not me the one jumping to any oportunity to claim any superiority of one company over the other. The release of this card doesn't make G92 cards obsolete and that is what I'm saying. Ati is back on the game again and that is good, but in way is owning as every Ati fan is so fast to try to point out, that's all. It's funny how I didn't saw you saying so cathegorically that AA matters. On the other hand I have not changed of opinion, AA at higher resolutions is not very important for me (more than 2x) and is not for many people, in fact every Ati R600/670 owner shouldn't be so concerned if they chose that cards. What has changed in the last moth that makes AA so important? Ah, yes...

BTW GT200 can do AA better, did you brought the argument then, no? Well...


bla bla bla all I see here is someone failing to point out one card's advantage over another which seems to indicate a bias. the 9800gtx is a decent card, but once the details are turned up it fails to compete with the 4850 which btw is cheaper. and no not everyone has a big screen, which is exactly my point with 1280x1024 in which aa matter more. I'll post up some pics when I get home. but at 1280x1024 aa matter just as I see a differentce at 1680x1050. you can argue all you want. but I intent to bring proof. you can do the same for all ic are, maybe then you'll actually take alook at the issue rather than pointlessly disagreeing based off of your own whim.

[KainXS]

bla bla bla all I see here is someone failing to point out one card's advantage over another which seems to indicate a bias. the 9800gtx is a decent card, but once the details are turned up it fails to compete with the 4850 which btw is cheaper. and no not everyone has a big screen, which is exactly my point with 1280x1024 in which aa matter more. I'll post up some pics when I get home. but at 1280x1024 aa matter just as I see a differentce at 1680x1050. you can argue all you want. but I intent to bring proof. you can do the same for all ic are, maybe then you'll actually take alook at the issue rather than pointlessly disagreeing based off of your own whim.

your right in the fact that aa seems to like the 4850 more but when I try to go from 4x to 8x, i really fail to notice a significant difference on my monitor.

[DarkMatter]

bla bla bla all I see here is someone failing to point out one card's advantage over another which seems to indicate a bias. the 9800gtx is a decent card, but once the details are turned up it fails to compete with the 4850 which btw is cheaper. and no not everyone has a big screen, which is exactly my point with 1280x1024 in which aa matter more. I'll post up some pics when I get home. but at 1280x1024 aa matter just as I see a differentce at 1680x1050. you can argue all you want. but I intent to bring proof. you can do the same for all ic are, maybe then you'll actually take alook at the issue rather than pointlessly disagreeing based off of your own whim.

All I see here is a person unable and unwilling to undersand a simple thing, with a big desire (only God knows why) o start a flame war. Others have undertood the point. Don't be silly continuing with the nonsense.

BTW screenshots mean NOTHING. Yeah let's take some screenies, open them up on PS and zoom in 8X. Then get our face 2 inches away from the monitor and "Oh yeah definately 64x AA is INDISPENSABLE.

You didn't catch my simple point in the first place so here it is: HD4850 is A LOT better once we use 8x AA, FACT(I could say it's even 4x times faster if that makes you sleep better), point is tht even if it was 100 times faster at 8x AA, since at 4x AA is barely faster it doesn't matter for the mayority of people buying a card in that price segment. It will only depend on price/performance/reasonable quality settings, period.

[yogurt_21]

uh huh, I figured you'd opt out of real research. and it's not just 8x aa in question here the card is 14% faster in 4x aa at 1280x1024 menaing there are going to be situations where a 9800gtx user has to run without aa to get reasonable frames and the 4850 user can run with it. there is no zoom on I'm merely looking a building, some cars, and a powerline in hl2 episode 2.

1280x1024 no aa menu
http://img.techpowerup.org/080621/1280x1024 no aa menu422.jpg

1280x1024 no aa shot
http://img.techpowerup.org/080621/1280x1024 no aa shot.jpg

1280x1024 4x aa menu
http://img.techpowerup.org/080621/1280x1024 4x aa menu.jpg

1280x1024 4x aa shot
http://img.techpowerup.org/080621/1280x1024 4x aa shot.jpg

1280x1024 8x aa menu
http://img.techpowerup.org/080621/1280x1024 8x aa menu.jpg

1280x1024 8x aa shot
http://img.techpowerup.org/080621/1280x1024 8x aa shot.jpg

I have 1650x1080 ones will will upload them on request. note that there are jpegs and quality will drop from the originals. jaggies on cars buildings and powerline may not amtter to you, but to me it matters so much so that I'll overclock my card before I'll drop details. there are many other like me. it's the whole reason most poeple overclock (other than the 3dmark score hounds) to be able to play at a higher detail level without buying a new card. the 4850 can handle higher levels of aa at higher frames than the 9800gtx can. take the screenies as you will, but there's no way you can tell me that aa doesn't matter. not that I gave any weight to your opinion in the first place. lol

[DarkMatter]

OMG I'm not going to spend even 1 more minute to read your crap above. WE ALL KNOW in this forum what AA is and how it affects the picture. We all can use those settings on our PC, we don't need some screens posted here to see how it looks. :shadedshu

[yogurt_21]

lmao and yet you seem to think it pointless. somehow I'm seeing a disconnect between your logic on aa and your understanding of it.

[snapper]

darkmatter is right. I own a hd 4850 and a 8800gts 512 and at 1600x1200 there is no diff in benchmark scores. Both are overclocked ,but the 4850 only overclocks to 645/1013, while the 8800 overclocks to 810/ 1030 mem.The 4850 blow out hot air like a 1200w hairdryer inside the case,(no kidding). I love the fact it is only a single slot,but i'm going with the 8800gts,(44 degree idle, compaired with the 4850's 75 degree idle).SO if you want to buy a one day old 4850,I'll sell it for $170

[btarunr]

darkmatter is right. I own a hd 4850 and a 8800gts 512 and at 1600x1200 there is no diff in benchmark scores. Both are overclocked ,but the 4850 only overclocks to 645/1013, while the 8800 overclocks to 810/ 1030 mem.The 4850 blow out hot air like a 1200w hairdryer inside the case,(no kidding). I love the fact it is only a single slot,but i'm going with the 8800gts,(44 degree idle, compaired with the 4850's 75 degree idle).SO if you want to buy a one day old 4850,I'll sell it for $170

Mind giving us GPU-Z validations of the 645/1013 for HD4000 and 810/1030 for the 8800GTS?

[snapper]

I've got the 4850 card in my computer now,and gpu-z only reads 500/750 unless i have a load on it.Any idea how to display it without a game playing?

[From_Nowhere]

Disable ATI PowerPlay. It will prevent it from down-clocking.

[btarunr]

I've got the 4850 card in my computer now,and gpu-z only reads 500/750 unless i have a load on it.Any idea how to display it without a game playing?

Download the latest GPU-Z, validate it.

[snapper]

I have a GA-MA78GM-S2H ati based motherboard and was hoping the 4850 was going to make a difference because i need a slot for my tv tuner. But for me the two cards are a toss up.The 8800throws out less heat(in the case), but consumes more power.I would love to keep the hd4850, but when i overclock it just a little, it shuts down in the first 5 min of the game.(I returned one 4850 already because i thought it was defective).I have not tested both at stock, but I would say the 4850 would beat out the 8800 easy.I do overclock ,so that is why I would go with the 8800(plus the heat the 4850 throws out) It's HOT!

[btarunr]

Validate your HD4800 at any setting, even if it runs at 1 MHz.

[snapper]

Btarunr,thanks for your help it worked.Now just let me know how to paste gpu z here

[btarunr]

Btarunr,thanks for your help it worked.Now just let me know how to paste gpu z here

Don't paste. Don't need to. Use the validation tab, submit, you get a validation number. Put up that number.

http://img.techpowerup.org/080622/bta027.jpg

Here's mine: http://www.techpowerup.com/gpuz/a38uu/

[snapper]

9k4e8

[btarunr]

Ouch. I didn't realise 641 was just a 3% OC.

[snapper]

Pretty good overclock ,I thought you were an ati man.
When i put my 8800gts in I'll send you my gpu z stats.

[snapper]

I can overclock the 4850 to 690,but will system will shut down.Needs a lot better cooler!!

[btarunr]

Like I said, the reference ATI cooler just about keeps the card operational, with a 9% overall OC at the most. W1zzard was able to take it to 9%. But yes, when the likes of GeCube, HIS, come up with their double-slot coolers and non-reference PCB's, you can expect better OC's. Add to that, any HD3800 series compatible water-block / cooler should work for a HD4850, but it's also that since it's the same core as HD4870, it might not OC well with the initial BIOS. Remember how the initial BIOS of HD3850 didn't allow you to take the core beyond 730MHz?

[snapper]

BTARUNR,Put my 800gts in, I don't know if you can use the same number to check it out

[btarunr]

Can't. Validate it again.

[snapper]

v2zhy

[snapper]

or try gxwk9

[rnxw]

we'll have to wait for 25th for the 4870 to show, i seriously hope its better than gtx260 so Nv would not put such ridiculous high price..

[Haz197]

yes, deffo, roll on the 25th !!!

My g92 is getting a bit old by the looks of things :( ........

Only hope someone near Manc has them in stock so I can go get mine after work, lol.

[yogurt_21]

yes, deffo, roll on the 25th !!!

My g92 is getting a bit old by the looks of things :( ........

Only hope someone near Manc has them in stock so I can go get mine after work, lol.

you know you might be able to pull a fast one on amazon or ebay and manage to sell it for 250$ u.s. and then be able to pick up a 9800gtx+ for a few dollar profit. gotta remmeber amazon and ebay are slow catching up to the market and the price drop was just announced.