Hd5870 - $299

[spicypixel]

This should be interesting. I might have to buy ATi for my next gen.

I too may get a 5870 on launch, 8800GTS needs retiring and having 3x DVI while not something I will be using immediately it will encourage me too... chicken and the egg situation in this case ATI comes to the rescue and Samsung and other TFT companies probably cried when it was announced, cried tears of joy.

ps sigged

[ajaidev]

I don't think it is a monolithic core, but rather a MCM and is a server product, not being targetted for the desktop.

He is talking about 2011 not 2010. By that time frame a 12 core bulldozer and sandy bridge is plausible. Other than that Intel cheap westmere with 6 cores aka i9.

The 16 core amd will also be launched in 2011 for servers.

[Hell Hound]

A solid eight-core is what's needed not twelve core's that will get 40% usage.

[Origin_Unknown]

cant wait for this the ol' 8800gts (640) is in need of replacing - 1.5/2 months is not going to go fast enough!

[Helloworld_98]

Your fine, PCIe 1.0 16x is plenty, even with lower ram amounts (amounts you wouldn't want in the first place ). The 2.0 standard really was for enhanced 1-4x slot throughput and multi gpu setups. Gotta tell you I'm much more looking forward to this card as time passes by, really would love to build and uber Deneb system with my MSI board.

I disagree, the PCIe 1.0 16x might have a bandwidth limitation as it only has 4GB/s of bandwidth, the 4890 uses 3.9GB/s of bandwidth so the 58xx's will probably use more.

[Oberon]

I disagree, the PCIe 1.0 16x might have a bandwidth limitation as it only has 4GB/s of bandwidth, the 4890 uses 3.9GB/s of bandwidth so the 58xx's will probably use more.

Does it really? I've never seen that figure before. What exactly would we use to measure throughput over the PCI-E bus?

[Helloworld_98]

Does it really? I've never seen that figure before. What exactly would we use to measure throughput over the PCI-E bus?

I'll try and dig it up, but I saw those figure's about a week ago.

edit: here we go, link

[largon]

edit: here we go, link

Those numbers are worthless and the guy that originally presented them is mighty confused.
Memory bandwidth, or rather the throughput-per-pin he's talking about has absolutely nothing to do with the amount of data going through PCIe link.
Ignore those numbers.

[Tim]

RV870 is 1.6 times faster than RV770

DirectX 11 40nm leap

We've heard that ATI's soon to lunch DirectX 11 performance card, something that we know as RV870 and something that AMD calls Evergreen, should be roughly 1.6 times faster than the RV770 chip.

RV770 is the chip behind ATI's Radeon HD 4870 card and it showed quite good performance, so this 1.6 times performance increase can give you a general idea that the new chip should be quite fast.

Of course 1.6 times is the best case scenario and it should be less than that in some cases, but overall this 40nm chip should be a good step forward for ATI.

The big issue is that by late November Nvidia might launch its Geforce GT300 chip, that might end up even faster, but at the same time Nvidia's DirectX 11, 40nm chip will also be bigger and probably hotter as well.


Source : Fudzilla



((Warning : Fudzilla))



Soon to lunch ey? Yum.

[Oberon]

Hrmm, (alleged) doubling of shaders and (allegedly) only 1.6 times the performance? I either call shenanigans on one or both of those numbers or TSMC still has some issues with the 40nm process.

[Helloworld_98]

Hrmm, (alleged) doubling of shaders and (allegedly) only 1.6 times the performance? I either call shenanigans on one or both of those numbers or TSMC still has some issues with the 40nm process.

it's Fudzilla, so he probably just made it up.

[Manicdan]

his math is wrong, any of us idiots know, the shirnk to 40nm gave them 60% more for the same area, and this chip is a good 40% bigger than the 770. and my math is so uber i know 40% + 60% = 100% (yes this was half a joke, but if ATI gives us just a 60% increase for being so much better and on a smaller process, they fail big time)

[koei]

That math is uberly wrong in too many ways to list. I wouldn't be surprised by 60% increase in performance although it would be much more upgrade-worthy if it was more.

[kromosto]

hey there are many people waiting for this to replace his 8800gts. like me

[Magic Chief]

hey there are many people waiting for this to replace his 8800gts. like me

I'm with you on that. I think its time to upgrade from the 8800GT. I just hope I can play Crysis 2.

[jam2k]

Either it's BS or he's talking about 5850 (Based on rv870)

[russian boy]

He probably meant only number of sps in these chips. 800*1.6 = 1280 quite possible

[spursindonesia]

IMHO, 1280 SP for RV870 = failure, even if it has a pretty nice clock in it, IF the efficiency of the mArch stays about the same.

[clonez]

imo the chip is too large for having only 1280sp...

[pellen]

hey there are many people waiting for this to replace his 8800gts. like me

Yeees, it's time for the 8800GTS to go

[~CS~]

1.6x times better then rv770 is essentially better then 4870x2 , perfect .

[Hell Hound]

For the 5850.

[Olivon]

I got a 8800 GTS 512 and it's a pretty good card (1680X1050) but I expect an upgrade near february 2010.

One question : I got a good old P5B Deluxe with PCIe X16 1.0.
This board is enough for the RV870 or GT300 at bandwidth's level or not ?

[ajaidev]

I got a 8800 GTS 512 and it's a pretty good card (1680X1050) but I expect an upgrade near february 2010.

One question : I got a good old P5B Deluxe with PCIe X16 1.0.
This board is enough for the RV870 or GT300 at bandwidth's level or not ?

Its okkk, we were talking about it a few posts up.

[Jamesrt2004]

+1, should be fine, Dual cards WILL take a hit this time around though