1GB 4870 now below $300 at Newegg (and in stock).

[dennilfloss]

http://www.newegg.com/Product/...x?Item=N82E16814103067

It keeps getting more enticing to those whose case won't hold a 4870X2.

 

[WT]

*copy*
450 Watt or greater power supply with 75 Watt 6-pin PCI Express power connector recommended
*paste*



But it states "Power Connector - 2 x 6 Pin"

I assume you need both 6 pins plugged in, but its slightly confusing when reading the PSU requirement that I pasted. I'll assume you need BOTH 6 pin PCIe power plugs plugged in.

 

[apoppin]

Originally posted by: WT
*copy*
450 Watt or greater power supply with 75 Watt 6-pin PCI Express power connector recommended
*paste*



But it states "Power Connector - 2 x 6 Pin"

I assume you need both 6 pins plugged in, but its slightly confusing when reading the PSU requirement that I pasted. I'll assume you need BOTH 6 pin PCIe power plugs plugged in.

all the 4870 require a 2 6-pin connectors -BOTH plugged in. if your PS is capable, you can use the molex to 6-pin connector adapters for your 2nd connector [or both of 'em] .. heck, i am running 4870 + 4870X2 and i am definitely using some molex adapters and even using an alligator clip for the 'extra two pins' the 4870x2 requires.

'mickey mouse' is still alive

 

[WT]

Thanks for the clarification, apoppin. I was 90% sure that's what it needed (they don't put them on there unless you NEED it) but I was remembering possibly the older Radeon line that you could function on one power connex, but needed both plugged in to do any OC'ing of the card.

I was sure I was going with the 4870 1gb cards until the huge GTX260 price cuts, so now I am again undecided. I have an eVGA 750FTW board, so my gut tells me to go Nv and SLI on this one, even tho I prefer to support ATI for a change.

 

[apoppin]

Originally posted by: WT
Thanks for the clarification, apoppin. I was 90% sure that's what it needed (they don't put them on there unless you NEED it) but I was remembering possibly the older Radeon line that you could function on one power connex, but needed both plugged in to do any OC'ing of the card.

I was sure I was going with the 4870 1gb cards until the huge GTX260 price cuts, so now I am again undecided. I have an eVGA 750FTW board, so my gut tells me to go Nv and SLI on this one, even tho I prefer to support ATI for a change.

See .. Nvidia IS smart

their marketing and pricing strategy is working

you have a very difficult choice .. the 260 is very appealing for $80 less
- and the 280 is also a choice for some for $80 more

i told you they would figure it out
- what do you think they will do when they transition to 55nm? Become less aggressive? i doubt it; they are in no danger of losing further market share unless AMD can transition to their own smaller process very quickly

it appears to me the price war is mostly over until the new GPUs arrive ... but the prices are also really good for us

 

[dennilfloss]

Just ordered a 1GB 4870.

http://pics.bbzzdd.com/users/dennilfloss/48701g.png

Newegg doesn't ship to Canada yet but Diamond does so I used the coupon code in the link below and just ordered one straight from Diamond. UPS delivery is $15, coupon takes off $20, no taxes were charged so my total was $314.99.

http://www.rage3d.com/board/showthread.php?t=33932063

My normal reseller is gonna take my factory-overclocked 3870 and give me a nice credit for it for my soon-coming CPU-Mobo-RAM upgrade. I already have some credit left with him from about a month ago (got $90 for my 1900XTX) plus I have some textbooks that will be bought by a new professor.

 

[Extelleron]

Originally posted by: apoppin

Originally posted by: WT
Thanks for the clarification, apoppin. I was 90% sure that's what it needed (they don't put them on there unless you NEED it) but I was remembering possibly the older Radeon line that you could function on one power connex, but needed both plugged in to do any OC'ing of the card.

I was sure I was going with the 4870 1gb cards until the huge GTX260 price cuts, so now I am again undecided. I have an eVGA 750FTW board, so my gut tells me to go Nv and SLI on this one, even tho I prefer to support ATI for a change.

See .. Nvidia IS smart

their marketing and pricing strategy is working

you have a very difficult choice .. the 260 is very appealing for $80 less
- and the 280 is also a choice for some for $80 more

i told you they would figure it out
- what do you think they will do when they transition to 55nm? Become less aggressive? i doubt it; they are in no danger of losing further market share unless AMD can transition to their own smaller process very quickly

it appears to me the price war is mostly over until the new GPUs arrive ... but the prices are also really good for us

nVidia isn't in a bad position in terms of market share, they have competitive products. The GTX 260 is fairly competitive to the HD 4870 for slightly cheaper on average and the new 216SP version establishes pretty much performance parity. For a bit more than the HD 4870 the GTX 280 is still a better card overall. Same thing with the lower end of the market. A 9600GSO is competitive with the HD 4670 and overall superior at the $80 price point.

But the problem is nVidia products at the same price point cost so much more to produce. The GTX 260 has a 576mm^2 chip, 896MB of GDDR3, a complex PCB / w 448-bit interface, and a high end cooler. The HD 4870 is a 260mm^2 chip, 512MB of GDDR5, and a (relatively) simple PCB w/ a 256-bit interface. And the GTX 260 is selling for the same or less. With the 9600GSO vs. HD 4670 it is even worse. The 9600GSO is a 324mm^2 chip and a 192-bit interface, the HD 4670 is a 146mm^2 chip and a 128-bit interface..... at the $80 price point nVidia has to be losing money on the 9600GSO (considering vendors get their share along with retailers).

The strategy of selling much more expensive to produce products at the same price as much cheaper products is not sustainable and it is not going to work for nVidia long term. IMO at least, nVidia needs a new architecture to compete. The GT200 / G80 architecture is not at all efficient/mm^2 and also requires a large amount of memory for decent performance at high res/ AA+AF. Even on the same 55nm process GT200 would likely be in the range of 400-450mm^2, yet it does not provide truly superior performance to a 260mm^2 RV770. With how efficient/mm^2 the R700 architecture is, AMD could likely fit a part with double the resources they have now (1600SP/80TMU) into the same space as a GT200 on the same process.

 

[taltamir]

ah, but nvidia is using GDDR3, which is much cheaper.
nVidia made the decision that it would be cheaper for them to make a huge die and use inexpensive GDDR3, instead of a tiny die and very expensive GDDR5. Unless you can tell me how much nvidia pays for their ram, and how much AMD does, you cannot assume that AMD is making more money per chip sold.

At the end of the day, nvidia's cards are prices very agressively.

 

[Extelleron]

Originally posted by: taltamir
ah, but nvidia is using GDDR3, which is much cheaper.
nVidia made the decision that it would be cheaper for them to make a huge die and use inexpensive GDDR3, instead of a tiny die and very expensive GDDR5. Unless you can tell me how much nvidia pays for their ram, and how much AMD does, you cannot assume that AMD is making more money per chip sold.

At the end of the day, nvidia's cards are prices very agressively.

No way is the price of GDDR3 vs. GDDR5 going to make a difference that would make the HD 4870 anywhere near as expensive as the GTX 260.

Remember the GTX 260 has 896MB of GDDR3, the HD 4870 has 512MB GDDR5. I would expect that 512MB of GDDR5 costs about the same amount as 896MB GDDR3.

nVidia did not design GT200 to be a competitor to RV770, they did not design it to be a cost effective chip. They designed it to be a chip that sold in cards priced at $450-650. Now they are selling those same cards for $250-400. Not very good for business, not just for nVidia but their partners as well. AMD designed the HD 4870 to be sold at the $300 price point, and they are doing just that. Same thing with the HD 4670; it was designed to sell for $80, the 9600GSO/8800GS was designed to sell more in the $120-150 price range.

nVidia doesn't have a choice about big die vs. small die at this point, the G80 architecture is not efficient in using die space so until they release a new architecture they are stuck. Going from G92 -> GT200 doesn't quite double the chip's resources and it is 78% larger. Going from RV670 -> RV770 provides 2.5x the shading/texture resources and improved pixel performance, yet is 33% larger in size. RV770 is a bit of a special case because AMD also removed the ring bus architecture which freed up a good amount of space, but regardless AMD's architecture is more efficient/mm^2.

 

[nitromullet]

Originally posted by: Extelleron

Originally posted by: taltamir
ah, but nvidia is using GDDR3, which is much cheaper.
nVidia made the decision that it would be cheaper for them to make a huge die and use inexpensive GDDR3, instead of a tiny die and very expensive GDDR5. Unless you can tell me how much nvidia pays for their ram, and how much AMD does, you cannot assume that AMD is making more money per chip sold.

At the end of the day, nvidia's cards are prices very agressively.

No way is the price of GDDR3 vs. GDDR5 going to make a difference that would make the HD 4870 anywhere near as expensive as the GTX 260.

Remember the GTX 260 has 896MB of GDDR3, the HD 4870 has 512MB GDDR5. I would expect that 512MB of GDDR5 costs about the same amount as 896MB GDDR3.

nVidia did not design GT200 to be a competitor to RV770, they did not design it to be a cost effective chip. They designed it to be a chip that sold in cards priced at $450-650. Now they are selling those same cards for $250-400. Not very good for business, not just for nVidia but their partners as well. AMD designed the HD 4870 to be sold at the $300 price point, and they are doing just that. Same thing with the HD 4670; it was designed to sell for $80, the 9600GSO/8800GS was designed to sell more in the $120-150 price range.

nVidia doesn't have a choice about big die vs. small die at this point, the G80 architecture is not efficient in using die space so until they release a new architecture they are stuck. Going from G92 -> GT200 doesn't quite double the chip's resources and it is 78% larger. Going from RV670 -> RV770 provides 2.5x the shading/texture resources and improved pixel performance, yet is 33% larger in size. RV770 is a bit of a special case because AMD also removed the ring bus architecture which freed up a good amount of space, but regardless AMD's architecture is more efficient/mm^2.

This tactic seemed to work for ATI with the 2900XT... Either way, the cost of manufacturing, profit margin, or whatever isn't really relevant to the consumer... It's what we pay that matters. I'm also pretty sure NVIDIA will be around just long enough to support any card you purchase today.

The prices on all video cards are really good right now for what you get. I just placed an order for another 280 GTX AND an SLI motherboard for a little bit more than I paid for just an X1800XT a few years a ago.

 

[Bateluer]

That's a tempting price, but I think I'll just sit with my 512MB 4870 until the R870s are released next year. I seriously doubt the minute performance boost is enough to justify the hassle of buying the new card and selling the old card.

 

[apoppin]

nVidia did not design GT200 to be a competitor to RV770, they did not design it to be a cost effective chip. They designed it to be a chip that sold in cards priced at $450-650. Now they are selling those same cards for $250-400. Not very good for business, not just for nVidia but their partners as well. AMD designed the HD 4870 to be sold at the $300 price point, and they are doing just that. Same thing with the HD 4670; it was designed to sell for $80, the 9600GSO/8800GS was designed to sell more in the $120-150 price range.

nVidia doesn't have a choice about big die vs. small die at this point, the G80 architecture is not efficient in using die space so until they release a new architecture they are stuck. Going from G92 -> GT200 doesn't quite double the chip's resources and it is 78% larger. Going from RV670 -> RV770 provides 2.5x the shading/texture resources and improved pixel performance, yet is 33% larger in size. RV770 is a bit of a special case because AMD also removed the ring bus architecture which freed up a good amount of space, but regardless AMD's architecture is more efficient/mm^2.

i can't see ANY reason a large die is a problem
- Nvidia joined SOi and i *bet* they will be stacking vRAM - on die - on Tesla in a year or two!

Open up an image of a GT200 card in an image editor and copy images of a RAM package over it. i found you could fit 12 ram packages into the same die area.

The RAM chips themselves are smaller than the packaging that house them, so i estimate there is enough room on the die for all 16 RAM packages a GTX280 uses, plus the NVIO die.

We know that IBM has been researching ways of stacking multiple dies on top of each other for quite a while now. This (currently) has a heat disadvantage, but many other advantages - shorter wiring, higher speeds possible as a result, etc.

Since Nvidia recently joined the SOI Consortium, this is important since SOI is a major technique employed with die stacking. By putting all this together, we see that the Tesla GPU die size is large enough now for die stacking the GPU + RAM to be feasible for Nvidia with future chip variations.
:Q


---i bet they are thinking outside AMD's box

Tesla is the future for Nvidia .. and that future just "includes" gaming

 

[angry hampster]

Are there any reviews comparing the 1GB and 512mb versions of the 4870?