Hardware Radeon HD 4890 vs GeForce GTX 275

They used Catalyst 9.4 (beta) for the 4890, according to the Test Set-Up.

 

erm and the gtx275? a slightly less tiny chip with substantially more power consumption, thats worser even surely

 

why on earth didnt AMD put 2GB's on the 4890?

 

yes I was referring to the 4890, put you also have to take into consideration that the RAM does not downclock at all.

 

glad someone pointed that out.. nvidia worked really closely with the devs on crysis- you could say the game was written with a nvidia card in mind.. firing squad usually has the best reviews (that I trust at least- bit-tech seems to be unbiased too, just I'm pretty new here).. something to look at also is the ati dx10.1 performance gains

why dx10.1 isn't used in any nvidia cards is still kind of a mystery.. they claim it isn't needed- but why not put it into your new cards for the hell of it- but then they keep physx proprietary.. they are just- well strange this whole last year.. gtx390 maybe something interesting finally

 

P18 FINAL THOUGHTS: All of the above presents a problem for ATI. To get more speed from its RV770 (HD 4870) GPU, it's had to stretch the design to allow the use of fatter copper interconnects to help maintain signal integrity and place more silicon between transistors to keep power leakage down as it raises the clock speed. But these tweaks have produced a hot and loud card.

As for the wider traces on the board as a failing of the 4890 is moronic at best. If that's what it takes for it to remain reliable so be it. What are the actual trace widths on both cards.

Placed more silicon between transistors. What is the spacing for either card, see what I mean... Empty advertising tripe.

Domination, the HD 4890 crumbles, comprehensively faster, struggling to play the game smoothly at all when the 275 does better.

Touch slower, it's not so much slower that you'd ever care, roughly matching, performance difference is marginal, so you're not losing anything when the 4890 does better.

Come on guys I like OBJECTIVE, not SUBJECTIVE and this reads like a Nvidia ad. Nice charts, poor proofing.

 

fingerprints on the last photo

 

people should allways read more than one review bit-tech.net, techpowerup.com, anandtech.com, techreport.com, pcper.com

 

Thanks for the review Tim.

According to both xbitlabs and anandtech, the 4890 consumes less power at both idle and load, and its cooler than the 4870.

http://www.xbitlabs.com/images/video/radeon-hd4890/4890power.png

http://www.xbitlabs.com/images/video/radeon-hd4890/4890temp.png

What do you think about their measures?

Ohh and another thing TIm, you stated that in the Cod 5 test , you used a 90-second manual run through from the second mission in the game.Can you tell me how you measured the average fps of the cards?
Cuz again at xbitlabs, the 4890 does better than the gtx285 in COD 5.

http://www.xbitlabs.com/images/video/radeon-hd4890/codww.png


Thanx, konstantine.

 

+1

I sadly have to agree. While the review/article is technically great as always on bit-tech it still seems somehow slanted in nVidia's favor. I don't question the benchmarking, but the wording is a bit iffy. It almost seems the reviewer(s) wanted to find flaws in ATi's new card and none in nVidia's new card.

Incidentally, I currently have an nVidia graphics card in my system.

@biebiep: I faintly remember having read somewhere that burst memory read/writes aren't enabled in the current driver.

 

That's because nVidia didn't actually change the chip for the GTX 275. They've simply taken a stock GT200b chip, which would qualify for a GTX 285 (240 shaders) and "blown some fuses" (i.e. limited the memory path and the number of ROPs) and then downclocked it. Implementing DX10.1 would require a redesign of the chip.

 

Author: bit-tech Staff
Is that because no one wanted to put their name to this load?

Biased language? Check.
Claims unsupported by facts? Check.
Dodgy graph reading skills? Check.
Laughable conclusions based on faulty perspective and logic? Check.

Yep, this is a copy/paste from a nVidia press release alright!

I'm a graphics atheist, I like whatever is fastest & affordable, and I come at comparison reviews with no preconceived hopes of what the outcome is going to be.
That is not what the authors of this puff piece did.

Scientific rigour means that science articles are peer reviewed before published. In the absence of this process for computer hardware/software, I rely on using the reviews of several sites I respect. This article has not passed my 4890/GTX275 peer review process.
As a long-time reader and forum lurker since 2004, this is disappointing stuff from Bit-tech. Can we look forward to more of this in the glorious post-Dennis period?

 

I read from somewhere that it was just a typo in the spreadsheet -> no burst memory read in the 4890.

Oh yea, it was from here (Finnish): http://plaza.fi/muropaketti/amdn-dokumentissa-virhe-rv790ssa-ei-purskelukua

 

Interesting you say that because some sections don't paint Nvidia in a particularly great light, especially where drivers are concerned - "we've been had" was the feeling going around the office on Thursday afternoon.

Comparing fan speed percentages across board manufacturers is about as useless as a chocolate teapot, frankly. The 4890 is louder, side by side, than the 4870 and the reason for that is the GPU temperature needs to be kept under control (with a loud fan). We think the 4890's fan speed could actually be set lower by default, because the cooler is pretty effective.

Your 'playable' settings are quite different to ours, then. 1,680 x 1,050 0xAA 16xAF VHQ is playable on the 275, while 1,280 x 1,024 4xAA 16xAF VHQ is the maximum settings we'd be happy playing Crysis at with the 4890. And even then, the minimum fps is six lower than the 275 and the average is five lower. That spells 'firmly in the shade' out to us. With regards to GRID, the frame rates are high and, as you correctly point out, the difference is negligible. The difference in Crysis at 1,920 x 1,200 0xAA is also negligible because, in our opinion, neither delivers a playable frame rate at this resolution.

We are a UK site and therefore our review tends to favour UK pricing over US pricing if it is available (like it is today). We do, however, list the US pricing at the start of the review because we respect that a sizeable portion of our readers come from the US. Note that we disregard the Palit card (which is the price Nvidia has been pushing onto us), because we're unsure about the cooling solution and whether it's any good or not. We could have quite easily used that price if we were Nvidia biased - as the saying goes, if you get flak from both sides, you're pretty close to the mark. We spent time arguing with both ATI and Nvidia in the run up to this article over different things.

Anyway, with the pricing the same (as it is in the UK) and performance very similar - we count three wins each and the margin of victory in Nvidia's wins is larger - you start to look at other things, like noise and maybe folding performance if you're into that kind of thing. If you use these cards in your main PC, idle and/or folding power consumption will be more important than gaming power consumption because you will spend the majority of your time idling or folding.

You're quoting us out of context in the first line - that's something we usually have to watch out for manufacturers doing, not readers. What we actually said was that it was a bit disappointing considering the song and dance ATI made about being able to achieve 1GHz GPU clocks. We didn't get there... some have, but we didn't and we can only go on what we saw.

 

Interesting, but we can only go on what we saw. We saw the 4890's idle power consumption was higher than the 4870's - I didn't run the numbers for that part of the review. In fact, we were counting last night that there have actually been five of us working on this review at some point in the last week. Three identical test rigs, side by side, spitting out numbers in tandem.

We measured the average and minimum FPS with the latest version of FRAPS (2.9.8).

 

No, there was not one person to credit with it because five of us worked on the article for some or all of the last week. It was mentioned earlier in the thread, but I managed to not count Antony, who did a lot of the early benchmarking with Harry and Clive as I was on holiday at the start of this week and joined in when I got back.

 

though I generally respect bit-tech and its staff, i can't help but notice how even Tim glossed over a few points in his latest replies... the first being the claims about the 4890 being a "hot" card and yet no comment on the 275, especially considering the evidence. also, RE: 4890 overclocking to 1GHz, someone else pointed out that HardOCP pushed the voltages up to reach 1GHz, proving it is possible - though it may involve processes outside of Catalyst.
Not a personal attack or anything, but I have to agree that currently much of the comments on the figures themselves reads like nVidia spin.

edit: perhaps the issue on heat isn't being completely conveyed? Throughout the review, the 4890 is constantly criticised for being a "hot" card. However your figures show it is only a couple of degrees hotter than the 275 at idle, and hardly heats up at all at full load (especially compared to the "boil a cup of water" 275 temperature, which is never mentioned outside of that single graph).

 

If you set the 4890's fan speed to the same as the 4870's, the card is hotter. AMD has used a simple trick of turning the fan speed up to make the card cooler (HIS did it here: http://www.bit-tech.net/hardware/graphics/2009/01/19/his-radeon-hd-4870-1gb-iceq4-turbo/14) and it makes the card look great on the page. But there's another side to the story and without a soundproof room we are not willing to present noise data outside of our admittedly subjective (but side-by-side) comparisons.

The ambient noise that our sound meter has picks up a lot of background noise - it's a problem with working in a busy lab, four times the size of our previous lab - with Custom PC and PC Pro. As for the 275, I didn't test it personally - as mentioned above, the reason for 'bit-tech Staff' being listed as the author is that the article was a culmination of five peoples' work over the week - so I can't answer. If I could, I would.

I'd be less blunt if it was a weekday and we hadn't answered all of the questions, but **** me it's the weekend and most of us have a life outside of bit-tech. I, for example, have travelled the best part of 180 miles to visit my 93 year old Gran today, as she's been in hospital for the last week - the last thing I wanted to be doing was answering questions I frankly don't know the answer to (see above).

 

point taken on the 4890 being hotter than the 4870 at the same speed, but that is still only in comparison to the previous design (and no mention of the 275)

none of us were demanding answers within minutes of posting, i for one could live without getting a reply till monday

 

four ATI/AMD cards beeing presented....and none have a individual cooling solution...sad.