ATI Removes NDA on Performance Previews for 4850 (#'s Inside)

Ok, so my understanding is that getting 2 HD4850 and running crossfire is near equivalent to a single GTX280. At least that's what I'm seeing from benchmarks. The HD 4850 is going for $200 (approx.) and the GTX280 is going for $660 (approx). The HD4850 does not really compare to a GTX280 because it's lowest end of a series vs. highest end of a series. So technically speaking, buying 2 HD4850 is the best bang for your buck.

Now all we have to do is wait for the HD4870. Assuming it's in half-way between a HD4850 and a GTX280 in regards to benchmarks, depending on the price, you could even pick up one of these and be completely happy with it--If you're not you could grab a second one and have better performance than a GTX280 for about $50-100 more than a single 280.

In the end, my assumption is as follows:

9800GT < HD4850 < GTX280
HD4850X2 == GTX280 (more or less depending on the resolution)
HD4850 < HD4870 < GTX280
HD4870X2 > GTX280
HD4870X2 < GTX280X2 (The former one cost less money then if you were to SLI 2 GTX280s)

Please correct me if I maybe be wrong somewhere. I'm trying to figure out whether to grab an intel board with crossfire or an intel board with SLI (don't have the money for one that has both :p)
 
One thing to consider, my 8800GT overclocks 20% whereas the 4850 gets less than 10% so overall the 8800GT is quicker.
Its easy to get a 20% overclock on the GT with a 1.1V Bios mod, some can achieve it anyway :)

I'm waiting for the 4870x2 results!

The OC to 700 is only because CCC is the only thing that currently works for OC'ing the 4850 and its limited to 700. Most users / reviewers have said they reached 700 without breaking a sweat. Until RivaTuner or ATITool are around to OC the card, we won't see > 700 scores yet, but I've heard getting close to 750 is possible.
 
One thing to consider, my 8800GT overclocks 20% whereas the 4850 gets less than 10% so overall the 8800GT is quicker.
Its easy to get a 20% overclock on the GT with a 1.1V Bios mod, some can achieve it anyway :)

I'm waiting for the 4870x2 results!

Indeed. I'm not impressed by the HD 4850.
 
No but it is not the be all and end all of benchmarks. Just because the card may perform poorly at Crysis or some other game does not mean it will suck at all others. Not all future games will run off the Crysis engine. You can't take home much from the PCPer preview because they test two games. The reviews that test 7 or 8 games, I would say that the 4850 beats the 9800GTX in the majority of benchmarks.


cosign back in in the day when HL2 dropped and doom3

ati and nvidia each ran on engine better than the other and vice versa

i am tired of Crysis being the end all for game testing
 
i am tired of Crysis being the end all for game testing

And I'm tired of umpteen cards in the market performing roughly the same or with unnoticeable differences just competing on price. We'll know when the manufacturers actually gave us something new when we get better Crysis performance. The GTX 280 isn't it. HD 48xx in Crossfire isn't it either.
 
I am glad that ATI is back in the game.
Competition is good for the consumer and we can now get a $200 card that performs very well.
 
And I'm tired of umpteen cards in the market performing roughly the same or with unnoticeable differences just competing on price. We'll know when the manufacturers actually gave us something new when we get better Crysis performance. The GTX 280 isn't it. HD 48xx in Crossfire isn't it either.
Then you're looking for something that will pull a 9700 Pro (or an 8800GTX, I suppose) and just utterly crush everything that came before it...and if the GTX 280 doesn't even meet your criteria, then those days are probably long behind us. Might as well pack it in...

Huh. We are now in the third stage of tech fandom, I guess. :D
 
And I'm tired of umpteen cards in the market performing roughly the same or with unnoticeable differences just competing on price. We'll know when the manufacturers actually gave us something new when we get better Crysis performance. The GTX 280 isn't it. HD 48xx in Crossfire isn't it either.
I'm sorry, but unless you just have more money than you know what to do with, price is arguably the biggest concern for the vast majority (more accurately, price/performance) and that has steadily gotten better and better. And this is only the HD4850 for pete's sake, we still have 2 more cards coming up from the guys in red that are even better.
 
Sooooo....nobody testing in DX9/WinXP?

Why would they? In a month you won't be able to buy XP any more. Not to mention with 1GB graphics cards in SLI or crossfire your 32 bit OS isn't going to be to pleased if a game tries to take more than 2 gigs of ram.

I'm sorry, it's not nice to hear, but MS has outdated your OS.
 
It'd be kind of nice if someone had bucked the trend entirely and busted out XP/x64 at some point...but I know that's a pipe dream (poor redheaded stepchild OS iteration).
 
same could be said about you, any time someone disagrees with a product assumption they get called fanboys.

Most of this started with some people posting things we don't even know for sure yet as fact. Or what seemed to be meant as fact anyway. I don't go for any 'one' company. As I have stated before, when I do a fresh build I get whatever is best at the time, I don't care if it comes from AMD/ATi, Intel, or nVidia. FACT is, we don't know until these other cards get here. Also, these prices are nothing new, every time a top end PC part comes along with a high price tag people get shocked, like they have never seen it before. Something tells me it is going to be a while before we see another 9700/9800Pro or 8800GTX. Sometimes I wonder if people just have unrealistic expectations. Whenever something big like the 9700/9800 Pro's or 8800GTX's come along and crush everything for over a year, we get spoiled and expect the next thing to do the same, doesn't always happen.
 
iBuyPower Testing and Overclocking Team Presents:
Radeon HD 4850 Single Card/Crossfire Benchmark scores

Hi guys, we have benchmarks for the ATI Radeon HD 4850 on crossfire. All tests performed in Windows Vista. We're still testing and will update these benchmarks when available.

ATI Radeon Test rig:
Intel Qx9770 @ 4.4Ghz
2x 1Gb Mushkin DDR3 XP3-12800 @ 7-7-6-18 1600Mhz
Asus P5E3 Premium X48
2x ATI Radeon HD4850


---

3DMark Vantage Score (Single Card): 7271
3DMark06 Score (Single Card): 13388

3DMark Vantage Score (Crossfire): 12223
3DMark06 Score (Crossfire): 21535

---

World in Conflict Benchmark (Crossfire)
Average FPS on Very High settings

59 FPS @ 1024x768
55 FPS @ 1280x1024
51 FPS @ 1600x1200

---

Crysis Benchmark (Single) [Stock Clocks]
41.395 FPS @ 1280 x 1024 (High)
31.305 FPS @ 1600 x 1200 (High)
24.37 FPS @ 1280 x 1024 (Very High)
17.735 FPS @ 1600 x 1200 (Very High)

Crysis Benchmark (Single) [665/1100]
44.285 FPS @ 1280 x 1024 (High)
33.495 FPS @ 1600 x 1200 (High)
26.14 FPS @ 1280 x 1024 (Very High)
19.26 FPS @ 1600 x 1200 (Very High)
(OC Notes: Card runs pretty hot, idles at around 55C and loads around 78C :eek:, the OC added 2 frames, which in Crysis is pretty much bread crumbs for us starving plebs. :p)

Crysis Benchmark (Crossfire)
35.62 FPS @ 1280 x 1024 (Very High)
26.94 FPS @ 1600 x 1200 (Very High)

Compare to the 3870X2 (Crossfire) on the same rig:
24.09 FPS @ 1280 x 1024 (Very High)
19.085 FPS @ 1600 x 1200 (Very High)

---

ATI Radeon HD4850 Pictures:

A very finely setup rig if I do say so myself. :)
p6190016bi7.jpg

p6190017uf0.jpg

p6190020wi1.jpg


Enjoy. :D
-From the iBuyPower.com Testing and Overclock Team-
 
Honestly, that "review" is worthless in my eyes as you don't compare them to anything or tell us how much you OC'ed the cards. Which I can only assume you did based on the dumb fan photos.
 
Honestly, that "review" is worthless in my eyes as you don't compare them to anything or tell us how much you OC'ed the cards. Which I can only assume you did based on the dumb fan photos.

The processor was OC'd and the video cards were at 675/1000, which isn't much of an OC since these cards run hot to begin with in idle. :(

Overall, I saw very little difference in 3DMark scores when bumping up the clocks.
 
13.4k with a quad at 4.4?
The quad padding on that score should be much higher.
 
Please remember that the processor is at OC'd @ 4.4GHz on a QX9770 Yorkfield. We'll post more benchmarks and stats as we go through the day. So far, we're all pretty impressed at the HD4850s, which are performing comparably to a 9800GTX rig we also have set up and testing. :)
 
Please remember that the processor is at OC'd @ 4.4GHz on a QX9770 Yorkfield. We'll post more benchmarks and stats as we go through the day. So far, we're all pretty impressed at the HD4850s, which are performing comparably to a 9800GTX rig we also have set up and testing. :)
sorry but Im not impressed with the temps Im seeing in reviews because high 80s to mid 90s is ridiculous. hell the 9800gtx is 25 degrees cooler under load. also the 4850 cant overclock worth a shit. the highest I have seen so far is right around 675 which was the freaking stock speeds for the 3850. :rolleyes:
 
sorry but Im not impressed with the temps Im seeing in reviews because high 80s to mid 90s is ridiculous. hell the 9800gtx is 25 degrees cooler under load. also the 4850 cant overclock worth a shit. the highest I have seen so far is right around 675 which was the freaking stock speeds for the 3850.
Stock 4850 fan is 14%.
ATI purposely forces a slow fan speed @ stock to reduce the noise as much as possible.

We already went through this with the 3850/3870 last year. Unfortunately everyone was so obsessed with the 8800GT, no one was paying attention. Old news.

I'd like to see the iBP person (or anyone who owns a 4850) play around with fan speeds and temps in RivaTuner and post the results. I don't think any reviewer has touched the fan speed, but so many seem interested in the horrible temps... Perhaps someone should investigate further.
;)

right around 675 which was the freaking stock speeds for the 3850.
What does the 3850's clock have to do with anything?
 
sorry but Im not impressed with the temps Im seeing in reviews because high 80s to mid 90s is ridiculous. hell the 9800gtx is 25 degrees cooler under load. also the 4850 cant overclock worth a shit. the highest I have seen so far is right around 675 which was the freaking stock speeds for the 3850. :rolleyes:

Christ, you people are now grasping for straws.

It can easily get to 700, and that's the limit imposed by CCC.

What does the 3850's clock have to do with anything?

Exactly. At this point they're just trolling.
 
iBuyPower Testing and Overclocking Team Presents:
Radeon HD 4850 Single Card/Crossfire Benchmark scores

Hi guys, we have benchmarks for the ATI Radeon HD 4850 on crossfire. All tests performed in Windows Vista. We're still testing and will update these benchmarks when available.

ATI Radeon Test rig:
Intel Qx9770 @ 4.4Ghz
2x 1Gb Mushkin DDR3 XP3-12800 @ 7-7-6-18 1600Mhz
Asus P5E3 Premium X48
2x ATI Radeon HD4850


---

3DMark Vantage Score (Single Card): 7271
3DMark06 Score (Single Card): 13388

3DMark Vantage Score (Crossfire): 12223
3DMark06 Score (Crossfire): 21535

---

World in Conflict Benchmark (Crossfire)
Average FPS on Very High settings

59 FPS @ 1024x768
55 FPS @ 1280x1024
51 FPS @ 1600x1200

---

Crysis Benchmark (Single) [Stock Clocks]
41.395 FPS @ 1280 x 1024 (High)
31.305 FPS @ 1600 x 1200 (High)
24.37 FPS @ 1280 x 1024 (Very High)
17.735 FPS @ 1600 x 1200 (Very High)

Crysis Benchmark (Single) [665/1100]
44.285 FPS @ 1280 x 1024 (High)
33.495 FPS @ 1600 x 1200 (High)
26.14 FPS @ 1280 x 1024 (Very High)
19.26 FPS @ 1600 x 1200 (Very High)
(OC Notes: Card runs pretty hot, idles at around 55C and loads around 78C :eek:, the OC added 2 frames, which in Crysis is pretty much bread crumbs for us starving plebs. :p)

Crysis Benchmark (Crossfire)
35.62 FPS @ 1280 x 1024 (Very High)
26.94 FPS @ 1600 x 1200 (Very High)

Compare to the 3870X2 (Crossfire) on the same rig:
24.09 FPS @ 1280 x 1024 (Very High)
19.085 FPS @ 1600 x 1200 (Very High)

---

ATI Radeon HD4850 Pictures:

A very finely setup rig if I do say so myself. :)
p6190016bi7.jpg

p6190017uf0.jpg

p6190020wi1.jpg


Enjoy. :D
-From the iBuyPower.com Testing and Overclock Team-
looking very good
 
Nice job IBP james, those temps do seem high for being benched out of case and with fans blowing on them. I was wondering if anyone so far has managed to re-apply new thermal compound and see if that helped. Some people were saying that it dropped temps by 10-20c.
 
I'd like to see the iBP person (or anyone who owns a 4850) play around with fan speeds and temps in RivaTuner and post the results. I don't think any reviewer has touched the fan speed, but so many seem interested in the horrible temps... Perhaps someone should investigate further.
;)

From what I've heard RivaTuner doesn't recognize/support the 4850s yet, so that's why reviewers haven't been messing with the fan speed. I agree it would be nice to see what temps they run at with the fan running faster. It might help the overclocking as well, maybe ATi doesn't wan't people overclocking them to achieve near 4870 performance (really low fan speed + locked settings in CCC) ;)
 
I'd like to see the iBP person (or anyone who owns a 4850) play around with fan speeds and temps in RivaTuner and post the results. I don't think any reviewer has touched the fan speed, but so many seem interested in the horrible temps... Perhaps someone should investigate further.
;)
Unfortunately, we haven't been able to get RivaTuner to work with the 4850 drivers that the AMD reps have given us, so fan speed modifications haven't been done yet. They did promise us special developers' tools to work on that, but for now, the best we can do here at iBuyPower is to blast air into the cards via 133CFM UltraKaze fans. :D
 
Christ, you people are now grasping for straws.

It can easily get to 700, and that's the limit imposed by CCC.



Exactly. At this point they're just trolling.
trolling? grow the fuck up. the 4850 can only oc to about 675 in the reviews I have seen. If its the CCC then fine but that still seems odd. I just thought it would oc much higher then the 3850 it replaced yet it can only oc to 675 which was stock speed on the 3850. thats not trolling you idiot, its wondering why the 4850 doesnt do better or why they limit it. maybe its because of heat because the card does run hot as hell.
 
trolling? grow the fuck up. the 4850 can only oc to about 675 in the reviews I have seen. If its the CCC then fine but that still seems odd. I just thought it would oc much higher then the 3850 it replaced yet it can only oc to 675 which was stock speed on the 3850. thats not trolling you idiot, its wondering why the 4850 doesnt do better or why they limit it. maybe its because of heat because the card does run hot as hell.

It doesn't matter that the 4850 won't clock to the same clock as the 3850 because the architecture is more efficient.

If you're really not trolling and you're serious, then you must've also been wondering about the 2.4GHz C2D E6600 compared to the 4GHz Pentium D. Heat isn't an issue in this case for enthusiast who overclock because they'll simply watercool or put on an aftermarket air cooler.
 
It doesn't matter that the 4850 won't clock to the same clock as the 3850 because the architecture is more efficient.

If you're really not trolling and you're serious, then you must've also been wondering about the 2.4GHz C2D E6600 compared to the 4GHz Pentium D. Heat isn't an issue in this case for enthusiast who overclock because they'll simply watercool or put on an aftermarket air cooler.
no thats a pretty crappy analogy. the 4850 and 3850 are built on the same basic architecture and process. I just figured that it would oc as good as the 3850. to be honest it looks like heat is the biggest culprit here.
 
no thats a pretty crappy analogy. the 4850 and 3850 are built on the same basic architecture and process. I just figured that it would oc as good as the 3850. to be honest it looks like heat is the biggest culprit here.

That's interesting, I hadn't seen this in any article yet. Can you provide a link?

I thought the architecture design of the 4800 series was still under NDA and the only things we know for sure about it architecture wise it is has 800 SPs, is on a 55nm process, has 16 ROPs and performs better clock for clock than the 3800 series.
 
That's interesting, I hadn't seen this in any article yet. Can you provide a link?

I thought the architecture design of the 4800 series was still under NDA and the only things we know for sure about it architecture wise it is has 800 SPs, is on a 55nm process, has 16 ROPs and performs better clock for clock than the 3800 series.
how about looking at all the reviews on the internet. also I thought it was pretty common knowledge that 4850 has a lot in common with the 3850.
 
how about looking at all the reviews on the internet. also I thought it was pretty common knowledge that 4850 has a lot in common with the 3850.

Anand's review states it's under NDA

http://www.anandtech.com/video/showdoc.aspx?i=3338

I can't find another review that specifically states information with reference to the architecture.

I'm just going to unsubscribe from this thread now because you're not interested in actually defending your posts with any real evidence. Take it easy.
 
Back
Top