4890 Review Thread

[judasmachine]

I'm sorry if this is the wrong thread or if this is threadc**pping. But is there going to be any changes in the more mid range? Something around the HD4850?

Back on topic: This one looks great, just wish it was bumping the prices of the other cards further.

 

[SunnyD]

Sweet, looks like I know where I'm siphoning off some of next week's paycheck to.

 

[Warren21]

Originally posted by: SlowSpyder

Originally posted by: BFG10K
Check out the hugely reduced power consumption in the XBit review, and also the amazing performance in Call of Duty 5 and Fallout 3.

Yea, pretty amazing that it shows less power usage for more performance on the same process, especially considering it's basically a higher voltage version of the RV770. They must haved saved some power on a few components or something.

Check out Creig's post above yours.

The AT review did a crap job at investigating that the HD 4890 is NOT just an OC'ed RV770. Have you seen RV770s clock to 1 GHz on stock cooling? I don't think so.

Check out info about 'RV790' if you want to know more. It's a bigger die, with 959M vs. 956M. It's not even pin compatible with RV770.

 

[SlowSpyder]

Originally posted by: Warren21

Originally posted by: SlowSpyder

Originally posted by: BFG10K
Check out the hugely reduced power consumption in the XBit review, and also the amazing performance in Call of Duty 5 and Fallout 3.

Yea, pretty amazing that it shows less power usage for more performance on the same process, especially considering it's basically a higher voltage version of the RV770. They must haved saved some power on a few components or something.

Check out Creig's post above yours.

The AT review did a crap job at investigating that the HD 4890 is NOT just an OC'ed RV770. Have you seen RV770s clock to 1 GHz on stock cooling? I don't think so.

Check out info about 'RV790' if you want to know more. It's a bigger die, with 959M vs. 956M. It's not even pin compatible with RV770.

Yea, I know. That's what I'm saying, the changes they made must have been fairly significant to have the card as a whole (which has a GPU that last I read uses a slightly higher voltage then the RV770 - unless my information is old and wrong) save power overall. Good job keeping power in check while increasing performance on the same process, that's a pretty good accomplishment.

Nvidia managed to make the GTS250 save a few watts over the older style 9800GTX+, but at the same time they made it into a more 'budget' card. As someone put it in a different thread, the 9800GTX+ used to be a pretty high end part, now it's more of a budget card. So when Nvidia made some changes to help bring the costs down they also made some of the on-board parts lower end... less complex PCB for instance. So they were able to save a few watts in that regards.

AMD essentially made the 4870 into a higher end part and saved a few watts. I think that's a pretty good accomplishment being that they both still 55nm.

 

[AnandThenMan]

Originally posted by: Creig
So what are these "tweaks and modifications" mentioned by DT and where did they disappear to between then and today?

Good question, but I guess consider the entire article. Quite a lot of it seems ripped from the pages of a mindless Nvidia marketing manual. Also consider that the conclusion was changed because it completely contradicted the numbers. A chart was updated because Anand "thinks" the proper ATI drivers were used (seems like if you are reviewing a card, you would know exactly what drivers you were dealing with)

The fact that no overclocking tests were done is odd. Even stranger is this from the last paragraph.

The overclock on the AMD hardware is fairly modest, but does make a difference and the same holds true for the GTX 275 products in early testing.

What? Modest overclocks on the AMD hardware? The thing was made to overclock, several sites are breaking the 1Ghz barrier. And of course overclocking makes a "difference" why wouldn't it? Strange, just strange.

 

[Red Storm]

Originally posted by: AnandThenMan

Originally posted by: Creig
So what are these "tweaks and modifications" mentioned by DT and where did they disappear to between then and today?

Good question, but I guess consider the entire article. Quite a lot of it seems ripped from the pages of a mindless Nvidia marketing manual. Also consider that the conclusion was changed because it completely contradicted the numbers. A chart was updated because Anand "thinks" the proper ATI drivers were used (seems like if you are reviewing a card, you would know exactly what drivers you were dealing with)

The fact that no overclocking tests were done is odd. Even stranger is this from the last paragraph.

The overclock on the AMD hardware is fairly modest, but does make a difference and the same holds true for the GTX 275 products in early testing.

What? Modest overclocks on the AMD hardware? The thing was made to overclock, several sites are breaking the 1Ghz barrier. And of course overclocking makes a "difference" why wouldn't it? Strange, just strange.

Exactly what I was thinking. As soon as I saw those 1GHz clocks, I was praying for benchmark numbers. Sadly I guess everyone was in a rush to get their reviews out. I expect to see some updates fairly soon.

 

[chewietobbacca]

I must say that I am dissapointed in the AT review

They missed out on a lot of the reason why the 4890 has its appeal, and frankly, a lot of it does seem like its copied from the Nvidia PR manual

There *WERE* changes done, mainly in the fact that they expanded the chip a bit to add capacitors to help lower signal noise as well as reclocking the core, so there are definitely differences

 

[WelshBloke]

Originally posted by: chewietobbacca
I must say that I am dissapointed in the AT review

They missed out on a lot of the reason why the 4890 has its appeal, and frankly, a lot of it does seem like its copied from the Nvidia PR manual

There *WERE* changes done, mainly in the fact that they expanded the chip a bit to add capacitors to help lower signal noise as well as reclocking the core, so there are definitely differences

I would think that NV is pretty pissed off at Anandtech at the moment.

They pretty much said that NV has been nagging them to talk about PhysX in every review, and now when they do they give it a big Meh.

 

[ronnn]

Originally posted by: WelshBloke


I would think that NV is pretty pissed off at Anandtech at the moment.

I felt the same. Between the lines was buy ati, or maybe I misread. Seemed quite harsh to nvidia.

 

[bryanW1995]

After reading the AT article I felt that the gtx 275 (cough *paper launch* cough) was equal to the 4890. After reading several others it looks to me that the 4890 is slightly faster. Xbit did a good job of digging a little bit deeper into the card. AT needs to do a better job of video card reviews, the last few that I've seen have been subpar. The lower noise of the gtx 2xx series is a huge plus imho, however.

 

[ronnn]

For me the big plus for the gtx 275 is lower idle power usage and I suppose, maybe one day physx.

Big plus for the 4890 is ati has a better track record in not needing to use your warranty, load power usage, overclocking and maybe one day dx 10.1. Plus you can actually buy it today for $269 canadian.

I always buy after market cooling - so actually I care too much about sound.

 

[CrystalBay]

Here's my take, Anand did an very interesting read ,Xbit did an excellent detailed review. The rest of the web were the same old same ,benches ,watts, etc... Anand showed what usefulness CUDA is at this point in time .They also were asking if CUDA software should cost extra.........To which I replied "If nV wants more sales add the CUDA video software for no charge" That is much more incentive for me to buy a GTX275 now than a new Radeon...;

 

[bryanW1995]

well, cuda actually matters to me since my seti numbers are skyrocketing right now, too. ATI seems to be a little bit behind in this area.

 

[bunnyfubbles]

Originally posted by: chewietobbacca
I must say that I am dissapointed in the AT review

They missed out on a lot of the reason why the 4890 has its appeal, and frankly, a lot of it does seem like its copied from the Nvidia PR manual

There *WERE* changes done, mainly in the fact that they expanded the chip a bit to add capacitors to help lower signal noise as well as reclocking the core, so there are definitely differences

I don't know about you, but this sounds about as far as you could get from sticking to an nVidia PR manual:

At 1680 x 1050 and 1920 x 1200 the 4890 is nearly undefeated. At 2560 x 1600, it seems to be pretty much a wash between the two cards.

I can sense the fanboy is strong with this one

 

[Elfear]

Originally posted by: bunnyfubbles

Originally posted by: chewietobbacca
I must say that I am dissapointed in the AT review

They missed out on a lot of the reason why the 4890 has its appeal, and frankly, a lot of it does seem like its copied from the Nvidia PR manual

There *WERE* changes done, mainly in the fact that they expanded the chip a bit to add capacitors to help lower signal noise as well as reclocking the core, so there are definitely differences

I don't know about you, but this sounds about as far as you could get from sticking to an nVidia PR manual:

At 1680 x 1050 and 1920 x 1200 the 4890 is nearly undefeated. At 2560 x 1600, it seems to be pretty much a wash between the two cards.

I can sense the fanboy is strong with this one

It may be that Anand made the changes to the conclusion (to better reflect their take on 275 vs 4890) after chewietobbacca made his comment. The article in it's original form wasn't very clear and made it seem like the GTX 275 was the faster card.

 

[SunnyD]

Has HardOCP done a review yet? I'd love to go read another one of their reviews which has one card at one set of settings versus another card at another set of settings and comparing those results like they're meaningful in a comparison of the cards.

 

[error8]

I want to see a good review where 4890 is overclocked using ATi tray tools ( to surpass the 1 ghz limit) and a higher rpm on the fan, with eventually increased voltage. Is there any?

 

[Cookie Monster]

Wait for the ones with 3rd party cooling, and since these chips has the capability of changing the voltage for the vGPU and vMem, i think we've yet to see the true overclocking potential of the HD4890.

 

[BRDiger]

As I recall, the mainproblem of the reference 4870s was not keeping the gpu temperature down, but keeping the Synchronus Buck Voltage Regulator by Vitec cool since it often radiatet more heat then the gpu itself... i wonder if they somehow avoided that problem with the new reference coolers or changed the Regulators...

 

[thilanliyan]

Originally posted by: BRDiger
As I recall, the mainproblem of the reference 4870s was not keeping the gpu temperature down, but keeping the Synchronus Buck Voltage Regulator by Vitec cool since it often radiatet more heat then the gpu itself... i wonder if they somehow avoided that problem with the new reference coolers or changed the Regulators...

No apparently that Vitec chip didn't need any heatsink but just some airflow. It was the VRMs that heat up a lot.

 

[BRDiger]

Hmm, it was ht4u who first realized a rather strange behaviour by the regulators...
here they compare temperature behaviour under load between gpu and voltage regulator resulting in the regulators temps always rising faster then the gpu´s indicating the cooling solutions weren´t up 2 the job and proofing that even most aftercoolers were simply cooling the wrong part. (Or ignoring the unusual heat source, resulting in a higher spinning cooler then nessecary)

 

[thilanliyan]

Originally posted by: BRDiger
Hmm, it was ht4u who first realized a rather strange behaviour by the regulators...

Yeah the voltage regulators that weren't cooled properly by aftermarket heatsinks are those 5 small regulators right beside the Vitec chip. On the stock heatsink those are all being cooled by the backplate while the Vitec chip is left unsinked but does have airflow over it.

That's why when I watercooled my card I left the stock backplate on and just used a GPU only waterblock. The regulators get to just above 70C at load now (assuming the VRM temps in Rivatuner are for those 5 chips).

 

[error8]

As seen in this review: http://www.bjorn3d.com/read.php?cID=1537&pageID=6673 Asus 4890 hits 1 ghz on the core, but only after a serious voltage bump to 1.41 V!! So it seems that good overclocking is not granted for these cards after all.

 

[Mana]

Originally posted by: error8
As seen in this review: http://www.bjorn3d.com/read.php?cID=1537&pageID=6673 Asus 4890 hits 1 ghz on the core, but only after a serious voltage bump to 1.41 V!! So it seems that good overclocking is not granted for these cards after all.

I concur. The 4890 I received did not overclock very well at all, and in fact I just have it sitting at stock speeds now. That said, I may have just been "unlucky" and didn't get a card that performs. I'm still very happy with my purchase, playing through Crysis: Warhead on max settings is almost surreal.

 

[soonerproud]

Originally posted by: nib95
Why do ATI cards have such terrible idle power consumption? It really bothers me. Especially since my computer is on so often, and electricity in the UK isn't cheap!

That is what sleep and hibernate are for. Just set the pc up to go to sleep in 10 minutes if it is not used.