The Official Graphics Card and PC gaming Thread

Only if you use what I call 'crazy', i.e. beyond 8x, and I'm an image quality fanatic and even I can barely tell the difference. Also consider that in some of these extreme cases the GTX295's memory limit can be broken even at 1920x1200, and when that happens the 4870X2 will fly past it. The GTX295, if it had proper memory management and 2GB of RAM rather than 1.7, would be unstoppable, but as it is, it's got such a big flaw there's never really any situation where it's worth having over a 4870X2.

 

WOW, I've just noticed how cheap the HD4890 is. That's less than I was expecting, basically the same price the HD4870 1GB used to be. This is excellent news.
XFX have already got a pre-overclocked version out (lol) for $265. Two of those ($530) will strip the pants off the cheapest GTX295 ($510)

 

Yes, but unfortunately I have the 'we live in australia' tax and two radeon 4890s is $900, a gtx295 is $850, and a radeon 4870x2 is $750

 

That's just weak currency, there's not much unfairness going on here. Think of the South Africans, they do literally pay 50% extra above the equivalent value of their currency to get their hardware.
$750AU is about £360 here, and 4870X2s are usually around the £340-£360 mark, though there are some £320 and £330 ones to find from specific brands.

450AUD for an HD4890 works out at £217, we pay guess what? £218 for them here.

The difference is, these prices are after 15% VAT is applied. I'm guessing in Australia there's no such tax, or there's less of it.

 

yeah we have a GST, 10%, but all prices always include gst, they arent allowed to advertise prices ex. gst as far as I know. Ok so its not as bad as I thought, thats good I remember paying $900 for my 7800gtx back in the day lol.

 

15% is included in all our items (except food and kids clothes, and it's 5% for utilities like Gas and Electricity), but some sites do advertise ex-VAT, some don't. Either way, the After-VAT price ALWAYS has to be printed somewhere on the page. If you bought the 7800GTX when it first came out, I'm not surprised you paid that much, they used to cost that much in pounds here as well, back in the 2005 glory days... haha.

 

A proper benchmark for the HD4890 has been upped. I'll post some detailed results in this post later this afternoon. One of the things I notice is that in some games the HD4890 vs HD4870 is utterly ineffectual (meaning the GTX275 wins) but in others I'm seeing bigger differences than preliminary reports indicate, for instance the minimum frame rate in HAWX when the detail is maxed is 27 for an HD4870, 35 for an HD4890.

Also, this will please hardware geeks no end:
http://techreport.com/discussions.x/16670




OK, Here's the full scoop on the GTX260 vs HD4870 VS GTX275 vs HD4890, courtesy of DriverHeaven.net!

Minimum Frame Rate Graph:
http://img14.imageshack.us/img14/7636/minimum4890.gif
Average Frame Rate Graph:
http://img27.imageshack.us/img27/681/average4890.gif

Comments to note:

The HD4890 architecture does not change the fact that ATI cards perform poorly in Call of Duty: World at Warcompared to their nvidia rivals. What's interesting though is they respond relatively well to clock speed increases as the XFX card shows. Since the newer architecture offers a 100mhz overclock over the 4870, less than the increase that XFX added on top, this must mean the newer Radeon architecture actually performs worse overall at the game. Perhaps the cards are being optimised in a way COD:WAW does not work well with. Crossfire performance for this game is, after all, mediocre at best with numerous graphical glitches as well as unsatisfactory performance.

In GTA4, on average frame rate ATI sit marginally ahead by around 2-3%. However, when the going gets tough, once again the Geforces win out, as we know already. On a more positive note, 19 to 22fps minimum is a relatively distinct improvement in minimum frame rate, even if it does mean you have to buy the overclocked card to reach it.

Left 4 Dead mirrors the same problems COD:WAW encounters, the performance gain from going 4870 to 4890 is negligible. The HD4890 competes with the GTX275 solely on the basis that the HD4870 solidly beats the GTX260 at this game. Again, some overclocking does as much good as the change in card.Intriguingly though, none of these results are what you would consider unplayable. There are far smoother experiences to be had with Dual or Quad graphics, but it shows you that if you want to play Left 4 Dead on high with a 30" monitor and only one GPU, you can.

Empire Total War carries an SSAO feature which at the time of the test does not function in DirectX9 with ATI cards, so won't be used for the comparison. However, compared to the other settings the GTX260 drops from 22-28 to 12-16, and the GTX275 from 29-33 to 16-20. A shocking decrease in performance it must be said. Perhaps this is one for the SLI configs, and with luck, in future drivers to come, Crossfire as well. Overall what's most important in this game is not how the ATI vs nVidia comparisons run at each end, but the huge gap between the GTX260 and GTX275 -nearly 20% at average, and nearly 25% at minimum. The HD4870 is definitely a better choice than the GTX260 here, but the GT275 can hold its own against the HD4890 rather well - higher minimum, lower average as we see so often from nVidia (Realistically this is what I prefer to see in games). It's worth noting that at this resolution performance is not exactly stellar on these cards. To run Empire on a 30" screen with AA, Dual graphics are mandatory.

World in Conflict: Soviet Assault is an interesting title. At 1920x1200 the HD4870 and GTX260 are on almost dead even ground, and the GTX275 and HD4870 aren't much apart either. Up the res, and the situation remains the same for the 4870 and GTX260, but the GTX275 pulls ahead of the HD4890, being a better fit for the XFX overclocked card rather than the stock product. Overall, advantage nVidia, but it's not by very much.

The trend continues with HAWX, ATI have a relatively significant lead here with average frame rates but the minimum frame rates are still very much in nvidia's favour especially with the older generation hardware. In addition to these test results, HAWX offers a 'Very High' SSAO mode utilising DirectX 10.1. Since nVidia do not produce DX10.1 cards, this is strictly ATI territory. Performance drops from 27-53 to 27-46 for the HD4870, 35-57 to 35-52 for the HD4890 and 35-61 to 37-54 for the overclocked HD4890. On the whole then, no impact on minimum performance, but a slight drop in average frame rates. Interestingly, the XFX HD4890 card is bundled with this game. A truly fitting demonstration of the card's performance.

Lastly, the review features a 'Crossfire' section. It is not made entirely clear whether the HD4890 has been combined with the HD4870, or whether it's two HD4890s, but I'm almost certain it i two HD4890s as three different brands were available on test for the review. It is assumed they are using the two stock-clocked cards for these tests. In lack of having two GTX275s on hand, there are no SLI results available as of yet until another site like Bit-Tech posts a review. However, since we all know at aD that SLI is only worth bothering with for i7, there's no huge benefit to them being here anyway yet, other than for interest purposes.
HD4890 CF results as follows:

Crysis Warhead 1680x1050 Enthusiast
HD4870: 14-21
HD4890: 15-25
HD4890CF: 26-41, but with Display Corruption. Expected fixed in Catalyst 9.4. 65% scaling is relatively impressive though, given the difficult circumstances.

Lost Planet Colonies 1920x1200 4xAA
HD4870: 16-28
HD4890: 16-31
HD4890CF: 32-58
About as close to 100% scaling as you get, but noticeably the performance improvement between the generations is not enough to up the minimum frame rate by the 6% required to see 17 on a single card.

As per usual, ATI are ahead in HD video playback performance, but by now the difference is never going to be enough to save you many watts or cure any lag, there shouldn't be any with either card.
The one difference is that CUDA saves a lot of processor activity when playing MKVs, so HD video users who have the CoreCodec will see benefits there, but only if they have a weak CPU, as even though the nVidia setup has less than half the CPU usage of the ATI in this instance, the maximum CPU usage ever reached was 10.5% in the latter case.


My final word on the speed of the cards

1st. XFX Radeon HD4890 XXX
2nd. GeForce GTX 275
3rd. Radeon HD4890
4th. Radeon HD4870
5th. GeForce GTX 260

 

After all that work, the HardOCP benchmarks completely conflict with what I just wrote. How fortunate.... :S


Check these results out:
Main chart details maximum playable settings. Comparisons shown underneath main chart.
(Minimum / Average fps)

Crysis Warhead
HD4870: 1920x1200 Gamer 1/29
GTX275: 1920x1200 Gamer (Enthusiast Water & Textures) 7/29
HD4890: 1920x1200 Gamer (Enthusiast Water & Textures) 8/32
GTX280: 1920x1200 Gamer (Enthusiast Textures, Water, Volumetric Lighting, PostProcessing) 7/29

GTX280: 1920x1200 Gamer 2xAA 16/25
GTX275: 1920x1200 Gamer 2xAA 17/26
HD4890: 1920x1200 Gamer 2xAA 17/28
HD4890 OCMax: 1920x1200 Gamer 2xAA 20/32


Fallout 3 Max
HD4870: 2560x1600 4xAA 27/44
GTX275: 2560x1600 4xAA 34/49
GTX280: 2560x1600 8xAA 28/44
HD4890: 2560x1600 4xAA 33/50
HD4890: 2560x1600 8xAA 28/48
HD4890 OCMax: 2560x1600 8xAA 33/53

Far Cry 2 Max
HD4870: 2560x1600 30/40
HD4890: 2560x1600 2xAA 30/41
GTX275: 2560x1600 4xAA 34/41
GTX280: 2560x1600 8xAA 32/44

HD4870: 1920x1200 4xAA 32/46
HD4890: 1920x1200 4xAA 37/49

HD4890: 2560x1600 4xAA 13/35
GTX275: 2560x1600 4xAA 15/38 (Test Variance due to Game bug)
HD4890 OCMax: 2560x1600 4xAA 21/40

FEAR 2 Max
HD4870: 2560x1600 4xAA 29/50
GTX275: 2560x1600 8xAA 23/48
HD4890: 2560x1600 8xAA 29/52
GTX280: 2560x1600 8xAA 25/47

HD4870: 2560x1600 4xAA 29/50
GTX280: 2560x1600 4xAA 30/57
HD4890: 2560x1600 4xAA 33/55

HD4890: 2560x1600 8xAA 25/52
GTX280: 2560x1600 8xAA 12/37 (Test Variance to note difference in other game areas)

HD4890: 2560x1600 8xAA 15/37 (Test Variance to note most demanding section)
HD4890 OCMax: 2560x1600 8xAA 25/42

 

I remember paying $550 USD for my 8800GTX back in the day ;P

 

Wow guys! Is bleeding edge performance really worth that price? I guess to each his/her own. Im not a huge gamer. However certain games have tried pushing me in that direction LOL! Grand theft auto for instance. Quite possibly the best game ive ever played. I must admit though, The legend of Zelda, the Ocarina of Time was the big one in its day. I still like it a lot to this very day! However, ive never played Far Cry, Crysis, etc. Heck, ive only played Gears of War briefly. I was quite impressed with that one! Probably the sharpest thing ive seen on my screen.

What would a person like me need, to be wowed, and be CLOSE to the times for at least a year. What I mean is, play any released game for the next year, and have an agreeable picture, little to no stutters, without spending more than 150$. Is that possible? Keep in mind what im now running, as well as you guys being substantially more critical than myself

 

You can get an HD4870 for about that these days in the US (or do you mean in your own currency?) and that falls into that category perfectly.

 

Ok, now how about a comparison. Ive been reading a little about PhysX. And from what ive read, its mainly Nvida based.
Lets just say I find PhysX intriguing! Now, what do you think between the two of these. I know, I know, im a cheap SOB right? LOL! As I said, im not TOO picky. There's totally nothing wrong with your being picky though. Im picky about other things

Will either of these cards get me by for a year or 2, and be substantially better than the 8600GT? And how much better?
http://www.newegg.com/Product/Produ...mpareItemList=N82E16814102822,N82E16814130445

 

You can ignore the PhysX really, it's just a fad that only really works in Mirrors Edge, and even then only minor benefits...
The HD4830 is miles better than the 9600GT.

 

What makes you come to that conclusion? FORGET the PhysX! I read a similar comment.

 

Because the HD4830 is the competitor to the 9800GT, not the 9600GT. The HD4830 is just a much faster card than the 9600GT.

 

Something tells me im just gonna be better off spending a little more money! Here's an interesting question. How much better in your opinion is the 4870 compared to the 4830? Disregarding the X2 model obviously LOL!

To the naked eye KEEP IN MIND

 

Maybe 50% faster...

 

Hmmm...thats rather considerable! Im gonna grit my teeth on this one, and just do it here in a week or so. But one last question. LOL, I think ive asked this one once before, not certain. One more time, Is the 1gb version worth the few extra pennies? Is it really only beneficial to dual monitors?
http://www.newegg.com/Product/Produ...mpareItemList=N82E16814102801,N82E16814102810

 

Not really dual monitors, it's beneficial if you use Anti-Aliasing though (though only at 1920x1200 and above)

 

My monitors max is 1920 X 1200(Samsung 2433BW). Do you think its worth it?