Radeon 4800 series chips ahead of schedule- shipping to partners

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

here you go
image3.php

http://www.nordichardware.com/news,7659.html
It's fake though :lol: and I'd bet we'll see a couple more fake ones before anything reliable shows up.
 


Nordic Hardware's original February report might be erroneous, but they reported the 4850 as 850 megahertz clock, and the 4870 and 4870x2 at 1050:

http://www.nordichardware.com/index.php?news=1&action=more&id=7356

But they recently reported they were off on some of the specifications, that the stock clocks for 4870 will be 900. Perhaps cherry picked factory overclocked GPU's will be higher?

http://www.nordichardware.com/news,7575.html

I can't see 4850 at 850 now. A 4870 might be a good match for a 3870x2 in CrossfireX after all. The only difference is the memory clocks and IMU.
 


Well they're already at 13.5W on the HD3850 according to Xbit labs, and 18.7W on the HD3870;
http://www.xbitlabs.com/articles/video/display/gainward-bliss9600gt-512gs_7.html#sect0

9600gt_power.gif


So depends on whether they disable unused portions of the chip and also reduce the lane useage (which is something the do with powerplay onthe laptops since the original mobile X series).
10W seems very doable at idle, but I would expect that to probably double at peak 2D and not change that portion much from the previous generation.
 


I could pay less (8c/ KWH in the land of oil) but I voluntarily pay a slight premium for wind generated 'Greenmax' power so it's about 9.5C/KWh for me.
 


Yeah it will be interesting to see how that works and if you look at the command and addressing schemes as outlined by that Qimonda link you posted the other day, it may offer some interesting insight into another potential for such a design.

all that is still ahead of the game, but it's nice to think of potential R700 configurations simply from from a design standpoint, not as anything to compare A vs B.
 


Not yet.
Not until I know the composition of the TMUs and what addressing and filtering unit sub numbers.
Also ROPs aren't as necessary for the HD series as they are currently composed, and how they will be used in DX10.1+ apps, the major concern with ROPs were AA and AF performance, and AF speed is easy, and AA speed is more shader bound, for the TMUs I would just prefer they improve ratios and also the number of FP32 pixels processed by each per clock, so raw ROP and TMU #s I don't care about so much as their composition.

1GHz core will run very hot BTW, and consume lots of energy.

Not necessarily, it depends on leakage and how the power is used, alot of things make up power consumption. But a high transistor count plus speed almost guarantees it won't be sipping power without some serious good conducting chips, but whether or not that's more/less heat & power than the R600/G80 depends on alot of factors, not just speed.

It'll be a while before we move onto 45nm process

It won't be that long, we're already at 45nm in the processors, and the GPUs have been at 55nm for a while. It depends mainly on how well TSMC does theri 45nm transition as to when the major VPU vendors go for that. Intel is definitely going 45nm for Larrabee.
 
Indeed I also hope they also improve on the shader processor architechture as r600 (and variants) didnt fair too well because of the fact that only a few from each cluster did heavy duty work whereas in the G80( and variants), all the SPs were useful. I cant wait for benchmarks. Its almost May, the expected release month, so I hope to see some solid specs and benchmarks.
 
Technology advances so fast. There has to be something to fix the fact that so much hardware ends up laying around doing nothing. Manufacturing companies should buy it from its consumers and recycle the parts or something. Im sick of having graphic cards lying around.
 



looks like im not alone switching from intel to amd.im looking forward to get my spider platform ready and rolling in 3 months time.BTW you just gave a head shot to the nV fB :lol: .LukeBird what mobo would you probably choose to run the 9850BE?
 

Aaah, but i'll be switching from AMD to errrrrr... AMD! :lol:
I want one of the 790FX boards definitely, because if the 4xxx's are as good as rumoured, I think a 4870X2 or two may be in order! The 790FX is a fantastic chipset, expensive, yes, but it supports all I could need and more! 😀
 


hahaha..........even though i own intel proc, ive been an AMD fan all along.i took a long decision to get intel.but quadiees,for the pooriees,hell who wont want one :ange:
 

:lol:
Well I have upgradeitis, so a 9850BE and my first delve into OCing looks like fun.
The 790FX is rumoured to be one of the best chipsets out there and is my choice as I really don't want to go AM2+ & SLi!
 


lol 790 is the new king of chipsets.id stick to spider platform to be future proof.that way i can hold upgrading for about 3 to 4 years! are you going to get spider as well?
 

Almost certainly, but not because of the excellent marketing!
I'm going that way, because I like AMD processors and the ATi cards are looking nice which ties in nicely with my desire to get away from an SLi board.
Never had an ATi card either, so that'll be something new :)
Although I have a 3870 in my newest work machine and it seems nice enough 😀
 

Aaah well I was late to the 'building PCs myself' party, my first build was the one before this and was a 6800GS 256mb. 😀
But yeah, next build is looking like a Phenom prob 9850BE, Asus 790FX board, 8Gb 1066 RAM and a 4870(s)/4870X2(s)
 


Yeah but I'm talking about shipping parts, 40nm is still in the early phase where it's targeted mostly at things like memory and low frequency high density chips
45nm is expected sometime this year (but IMO for the lower end and mobile parts.
40nm is just starting testing, so product built on the process likely won't arrive until the end of the year around Xmas, or depending on yields and process success may even be 2009.

Here's Xbit's recent article on the two 40nm processes they're testing 40LP (for small devices) and 40G (for CPUs / GPUs);
http://www.xbitlabs.com/news/other/display/20080324215024_TSMC_to_Make_Chips_Using_40nm_Fabrication_Process.html

Looks promising, but I would expect 40LP to get useable yields first and then they would tweak 40G to give the high frequencies for processors.

But if it was just launched now it's usually 9-18months until we see any of the first fast processor products depending alot on if there are any issues (similar to the 80nm HS problems). Usually we would se something like a chipset or small part first before something as complex as a VPU, they may come quick, but usually the very very first are memory and simpl ICs.
 
Thanks for that. I like the half node approach, to me its like fine tuning. Theres so many rumors on this card : 850 or 1Ghz, 32 TMUs or 16, 200Mhz or no new design/code etc. Im just hoping its worth all the hype when all is said and done. Im still predicting 10%+ over the Ultra, but who knows
 


The confusion is really the two memory types that will be used on the card. Do believe DDR3 and DDR5 (forgot actual names, too tired to look it up). One is a slow speed, the other extremely high.
I'm looking forward to it, hoping it will be much faster. It's been a LONG time (years) since waiting for a video card to actually perform much better then my 8800 card.
 
GDDR5 has several nice aspects to it, besides just higher clocks. Im looking forward to that implementation. If theyve corrected the "problems" with the r600, and upped the clock speeds, and the shaders and etc Zx2 etc, all thses things, then maybe we will actually see what the potential of this arch can do. Still hoping for the best
 
We all are jaydeejohn, cuz WE all wondered why the HD2900XT didn't massacre the G80. Maybe this is it? Nvidia is on pace to freefall, the bigger they are the harder they fall.