first r420 review leak

[elroy]

Completely OT, but:

1) Dave, could you implement something on the boards where we can jump to any page in a thread? About 10 pages has been added to this in two hours, and it is a real pain to go back through all pages to get to one in the middle.

2) Glad to hear the bjorn3d server is alright. It is not cool to f*ck around with someone's website, esp. the night before a big product launch. I'm surprised at a few of the people on these boards that did - I thought they would have known better. And no, I'm not going to name any names.

That is all .

 

[Wunderchu]

night guys...feels good to be posting here after lurking for ages.

night

 

[Fodder]

1) Dave, could you implement something on the boards where we can jump to any page in a thread? About 10 pages has been added to this in two hours, and it is a real pain to go back through all pages to get to one in the middle.

You can do it manually, copy one of the available page links and change the post number to start at (ie. page 2 is post 20, page 16 is post 300).

 

[Evildeus]

Really disapointed with the XT scores if they are true. Too much hope i think
Fortunately the XT and U cards are both great cards.

 

[Typedef Enum]

At this point in time, I would have to look @ the NV40 as the leader of the pack. The most dissapointing aspect of the R420 is the fact that, other than adding the pipelines, ATI really didn't do a heck of a lot in terms of functionality.

This is one of those catch-22 deals because you could definitely apply the ole' "if it ain't broke..." cliche to ATI. In other words, nVidia had a heck of a lot of work to do in order to get back in the race, so you had to believe that they were going to clean house on their NV3x arch. On the other side, ATI has had a pretty darn good run since the 9700, and they didn't really need to do a heck of a lot...

But all in all, I'm just dissapointed in that ATI didn't do more...Heck, I would make the overall argument that you can still improve, even if you're kicking ass and taking names. They could have supported FP32 (even if you think it's unnecessary), supported the latest/greatest shader specs, come up with more interesting AA modes (IE incorporate some SS modes, a-la nVidia), and a couple of niche features that one or two developers might have adopted.

Anyhow, I'm never one to pounce on a 1st gen. product, so I'm going to wait until this Fall to buy something. Actually, IMHO, there's not a single game out there that makes my 9800 Pro feel like a chump, so there's no real compelling reason to upgrade now...but if I were to buy something now, there's no doubt it would be an NV40.

 

[Randell]

I actually put off playing Far Cry for a couple months since I know my machine 2.4Ghz Athlon/9800 PRO/512 RAM can't run it at max detail. You guys playing now....do you really plan on replaying the game when u get new hardware? I barely have time to play each game I have once....far less multiple times. KOTOR is like 70 hours all by itself!!! And the mandatory sessions of BF, UT and Warcraft don't help either

I'm with you there! I know someone with an FX5200 who isn't playing it either..

 

[IgnorancePersonified]

Jeeeeez...
I started reading this thread at 9 am - 8hrs ago and the page count allready was at 12...I don't think I have ever got through more than half of it

 

[jvd]

Jeeeeez...
I started reading this thread at 9 am - 8hrs ago and the page count allready was at 12...I don't think I have ever got through more than half of it

hehe

 

[anaqer]

Back to the THG leak for a sec...

Unless Samsung somehow changed their numbering scheme for GDDR3, this looks like 1.6ns and 2.0ns (obviously) sooo... why such conservative memory clocks? Looks like the XT could go with as much as 1250MHz and the Pro with 1GHz... the Pro might not need that extra 100MHz that bad with the 12 pipes, but I'd expect ATi to crank up XT memory clocks as much as possible to ease Bandwith Blues.


| EDIT : I keep forgetting the damn thing is called GDDR3... |

 

[Nite_Hawk]

At this point in time, I would have to look @ the NV40 as the leader of the pack. The most dissapointing aspect of the R420 is the fact that, other than adding the pipelines, ATI really didn't do a heck of a lot in terms of functionality.

This is one of those catch-22 deals because you could definitely apply the ole' "if it ain't broke..." cliche to ATI. In other words, nVidia had a heck of a lot of work to do in order to get back in the race, so you had to believe that they were going to clean house on their NV3x arch. On the other side, ATI has had a pretty darn good run since the 9700, and they didn't really need to do a heck of a lot...

But all in all, I'm just dissapointed in that ATI didn't do more...Heck, I would make the overall argument that you can still improve, even if you're kicking ass and taking names. They could have supported FP32 (even if you think it's unnecessary), supported the latest/greatest shader specs, come up with more interesting AA modes (IE incorporate some SS modes, a-la nVidia), and a couple of niche features that one or two developers might have adopted.

Anyhow, I'm never one to pounce on a 1st gen. product, so I'm going to wait until this Fall to buy something. Actually, IMHO, there's not a single game out there that makes my 9800 Pro feel like a chump, so there's no real compelling reason to upgrade now...but if I were to buy something now, there's no doubt it would be an NV40.

I'll wait and read Dave's review before making any definite conclusions, but I still think there are upsides and downsides to both. So far it seems that the cards are close enough in speed that it comes down to a couple of factors for me:

1) Linux Support
2) Image Quality
3) Heat/Power Issues

I really dislike the PR tactics nVidia has used for the last year, but I'm also rather upset with ATI over their linux support (or lack thereof). I'll probably end up sitting this round out as I'm still reasonably happy with my 9700pro, but if I had to choose this generation I'm really not sure who I would pick. I don't think there is really enough difference between the two companies offerings that either choice could be claimed to be better than the other.

Nite_Hawk

 

[dr3amz]

some of you really are stupid. the 6800 was a MASSIVE leap over previous hardware - AS IS THE X800.

and if you read properly - the X800 is much faster than the 6800 when enabling AA/AF - not to mention IQ is better!

so stop whining - did you really expect it to be 300% faster than the 6800? No. If you did then yer just plain stupid tbh

check out the shots from the Ruby tech demo on this page (scroll down)

http://www.meristation.com/sc/articulos/articulo.asp?cr=5218&p=&c=GEN&pag=5

 

[hmmm]

I am a fan of ATI cards, and I have complimented ATI in the past. Prove me wrong.

I just thought that was funny. You're telling him to prove that you have never said a nice thing about ATI. That is like trying to prove god doesn't exist. It is logically impossible to prove a negative.

 

[carpediem]

This ruby picture is really nice

 

[Rican]

to me all this iq is better than the other dosent hold ground this time around on the leaked thg pics the iq looks almost the same on both card
people should accept that nvidias card is just ass good as ati's. all the benchies say the same nv wins some and losen some the same with ati for me iam just going to wait and see how much the 6800U and the x800XT are going to cost and go for the lower price

 

[hoom]

So why is ATI using Samsung ram?
I thought they were supposed to have a monopoly on Micron ram?

(sorry if its already been discussed, I couldn't be bothered reading the whole thread)

 

[LeGreg]

That is like trying to prove god doesn't exist. It is logically impossible to prove a negative.

How would it more impossible to prove a negative than a positive ?

I can prove that 6 is not a prime number.

 

[DemoCoder]

Math is deductive. It's a closed system. Proving the non-existence of something via induction is close to impossible.

Prove there is no extraterrestial life.
Prove that there is no dog that can speak english.
Prove that water cannot change into wine via magic.

 

[DemoCoder]

Hmm, what's the max parallel issue per pipe for R420 in ideal case? From the diagram it looks like:

2 vector ops (dual issue)
2 scalar ops (sincos, pow, etc?) dual issued
1 texture op

= 5 ops per pipe

agree, disagree?

 

[g__day]

This thread has grown so fast - it was 6 pages when I started and had to go out, the end has always been about 6 pages ahead of me.

It makes me wonder will there be this much comment when the final, real reviews and previews are delivered.

Personally I still feel its to early to judge much beyond both cards are excellent - especially merely from looking at alot of graphs other websites did.

I can wait for a few more hours until the real Macoy steps into the ring!

 

[Ailuros]

I obviously missed (or not) the train of thought for 17 pages, yet small OT: I honestly hope that the supposed XBox2 document is either ancient or just a fake. I would like to see more than just 16 bilinear pixels per clock in future designs.