Head to Head: Nvidia 9800GTX+ vs ATI HD4850

Mike D. 2 Feb, 2009 at 3:55pm ET Article published in Tech

3DMark 2006

Futuremark’s 3DMark 2006 is arguably the most popular 3D benchmark of all time. We benchmarked using the default settings, as that is the most widely recognized configuration.

There is no question that 3DMark 2006 favors the 9800GTX+ over the HD4850 by roughly 1500 points. There is no PhysX support in 3DMark’06 to skew the benchmark, so this is a fairly accurate picture.

When looking at the individual tests, we see that the HD4850 is not far behind the 9800GTX+ in shader model 2.0 tests, but lags behind by a significant margin in the shader model 3.0 tests.

« Previous Next page »

Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Comments

2 Feb 2009 ~ 4:14pm Sledgehammer70 The 9800GTX+ is a great card you just can't beat the price performance on the thing.
2 Feb 2009 ~ 6:36pm Thrax Great article, Mike. It provided me with some answers I was seriously looking for. Cheers!
3 Feb 2009 ~ 5:37am inf3rno The article was good but not good enough. There was some smell of biasness of green camp fan. The HIS card's specification were not told. And what about the price performance ratio?
3 Feb 2009 ~ 8:10am lemonlime Thank you for the comments, inf3rno,

We didn't get into too much detail about the ATI card, because we were reviewing the Zotac 9800GTX+, not the HD4850. I agree that we should have listed the specifications for the HIS card, which was an oversight on my part. We'll have it added into the testing configuration section. Believe me, I am not biased toward either side of the fence. I own both ATI and Nvidia products and believe both the HD4850 and 9800GTX+ are excellent cards (as stated in the conclusion).

For anyone interested, the card we used for testing was the HIS HD4850 IceQ 4 Turbo 512MB. The full specifications can be found at the following URL: http://www.hisdigital.com/un/product2-28.shtml. The HIS card and the Zotac card were overclocked out of the box, but we also tested both the HD4850 and 9800GTX+ reference configurations to ensure an equal playing field.

Also, I think price performance ratio was stated pretty clearly in the conclusion:

From the review wrote:

Price also plays a major factor in our ability to recommend one card over the other and we believe they are priced appropriately for their capabilities. The 9800GTX+ retails in the $150-210 range depending on the card and bundle, and the HD4850 can be found for as low as $125. To be perfectly honest, we cannot name a clear winner as both offer competitive performance catering to slightly different value propositions.

The bottom line: The 9800GTX+ performs slightly better overall, but costs more. The HD4850 is a better price, but is not quite as fast as the 9800GTX+. I think we were very fair in both our testing and our evaluation of both products.
3 Feb 2009 ~ 12:16pm stas This article was biased - and i own only nVidia cards. Here's why:

1) Catalyst 8.8?
2) 353 pointers fold MUCH faster than 511 pointers... I fold with 2 gtx 260's and the 260 gets 8,000+ points with 353 and only 5000+ points with the 511. The fact that the 4850 is slower at folding is not the issue, its that the test itself is heavily biased.
3 Feb 2009 ~ 12:35pm lemonlime Thanks for the comments. Our apologies. The catalyst version was a typo. We used the latest (at the time of writing) catalyst drivers which was 8.12.

F@H in general is very heavily biased towards nvidia cards, unfortunately. We included the tests as it is a very popular application but it is definitely not a fair test and we agree with you wholeheartedly. As you can see in the Sisoft Sandra benchmarks, the HD4850 does have a lot of synthetic number crunching potential. Applications simply aren't taking advantage of it. This was the only point we were trying to make.
3 Feb 2009 ~ 12:41pm Thrax Correction: The original run of this article erroneously reported that our Radeon HD 4850 employed ATI Catalyst version 8.8 drivers. The article has been corrected to read Catalyst 8.12, the newest available during product testing. We apologize for any confusion this may have caused.
3 Feb 2009 ~ 12:56pm BuddyJ Stas, it's not that the test is biased, it's that Pande Group has given the ATI cards the GPU2 Folding version of a golf handicap. And when it comes down to average PPD, all that matters is which card pushes the most points for the buck. How they get those points is irrelevant.

Units that are assigned to NVIDIA cards aren't assigned to ATI cards, and vice versa, as far as I understand it.
3 Feb 2009 ~ 4:14pm mas0n Correct, they do not share WUs.

I believe that what Stas is referring to is that when analyzing the average PPD for a given card we should take into consideration that performance as expressed in PPD varies greatly between groups of WUs. For example, my 4870 has a performance range of 3000-5300 PPD and my 9800GT gets 3500-5200. I could pick a random pair of WUs showing an impressive performance lead in either way.

Nice article. Two great cards; competitively priced.
3 Feb 2009 ~ 9:03pm Leonardo

HIS HD4850 IceQ 4 Turbo 512MB

Where do they find the crackheads to come up with these silly names? Hey Mr. Marketingdude, we aren't all 12 year olds.

Now, if I were one of the marketing dolts, I'd prefer a name like the "Illudium Q36 Explosive Space Modulator 512MB +/-/=/!."

(ok, I won't quit my day job)
5 Feb 2009 ~ 11:53pm Tushon So, can someone point me to a guide on how to overclock video card? I'm kinda mucking around right now just raising a bit then testing in games. I can't imagine that is best way but real-world effective, I suppose. Anyways, I look forward to the link.
26 Mar 2009 ~ 1:33pm nuke7 hey all,
i just wanted to point out an obvius fact, which was not mentoined in the article - or i just didnt see it.
the ATI card has a more modern architecture!
it supports newer direct x versions, and its performance will get better with - or stay on the same level - games evolving, while the gtx+ card will loose its edge in the upcoming games... (i think this article is pretty biased towards the green zone.. )