• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

384-bit Memory Interface for HD5000-series?

Status
Not open for further replies.
dccmadams said:
1st page of info, then the p*ssing contest begins.
Click to expand...
I thought it was a fun read. :p

necrophile said:
I don't think ATI have too much choice in regards to pricing. If they have high prices then they will look as bad as nvidia and I don't think they will make the same mistake.
Click to expand...

That is one good thing that AMD has done in the past. I mean look at their cpus, they've always had a good price/performance ratio. Even when they were owning intel with their socket A chips.
 
if they have 40nm on lock by release and cards can be stock piled ahead of time, the i'm sure the prices will be very nice at launch like the 4800 series, otherwise we get 4770 all over again..
 
And in the after life we can jigg with 1024 bit. Comon some company wants to boast! COMON ATI bellow :eek: "burps"
 
Zero82z said:
Fuck it, I'm talking apples and you're talking oranges. I'm not going to bother arguing with you any more since what you're saying has very little to do with what I'm talking about.
Click to expand...

necrophile said:
No, the only reason they are a multiple of each other is because they are both binary numbers!!!!!!!

There is no direct correlation between bit width and memory amount, sorry if I have said something to indicate that or seemed to be in agreement with other posters.

Memory bus and the speed of the memory is what is actually of importance, these two determine the bandwidth of the memory system.

If you require the gpu to be fed with data at a certain rate then you need a certain bandwidth, ie you need the data to be pushed odwn a certain number of paths at a certain speed.

Memory size is nothing to do necessarily with the bus width, it is to do with the amount need to store locally all the information needed for the gpu to process a frame quickly. Again, large amounts of memory which does not have the bandwidth will create a bottleneck so that is why you either have to increase the speed of the ram or the bandwidth when using high amounf of memory other wise it is a gimmick like all those 1GB 9500gt's with 128 bit bus's.

I ain't lying here, just think about this logically.
Click to expand...


While memory bandwidth is not dependent on memory size in the way we see it, it is from a hardware point of view, this also depends on the memory chips that are available.

the way it works is this.

Samsung produces 1gbit 16bit chips, 1gbit being 1gbits / 8 = 128mbyte 16 bit chips, to use a 64 bit bus in conjunction with these chips and have a total of 512mbytes of memory(aka 9500gt) you would need 4 of these chips, 4 x 128mb = 512mb, each being 16bit, now you have a total of 64bits. to get 1GB cards you would need to use either 8 x 128mb/1gbit 8bit chips, or 4 256mb/2gbit 16bit chips. So there is a dependency on the availability of different chips that will limit the bandwidth and size.

The same goes for the GTX280 vs the GTX260, the only difference is that the GTX280 has 2 more 64mbyte 32bit chips.


Now I didn't read the rest your argument so I'll assume this is what you're talking about.

here is a picture of a GTX260, you can see the missing 2 pieces of ram, the memory controller on the 2 could even be identical, but the 260 is programmed to use 2 less 32bit channels, technically anything that can use more then 1 piece of ram at the same time, as in access them both and write / read to or from both at the same time(either a stick for a pc, or a chip on a vcard) has multiple memory channels.

http://www.ixbt.com/video3/images/ref/gtx260-scan-front.jpg
http://www.ixbt.com/video3/images/ref/gtx260-scan-back.jpg
 
Last edited:
Since it is easier and less expensive to keep all the chips on a card at the same density value, I have yet to see a video card where the amount of RAM was not a multiple of the bus width that doubled the amount of RAM.

A 256-bit bus card was always configured with 256MB, 512MB (256 * 2) or 1GB (256 * 4). nVidia's 8xxx series were the same way. The 320-bit bus cards were available with 320MB or 640MB of RAM.

If this card has a 384-bit bus it is likely to come configured with 768MB of RAM, or 1536MB, simply because of the density values in which RAM modules are created, and the desire not to mix RAM types on the card.
 
Un4given said:
Since it is easier and less expensive to keep all the chips on a card at the same density value, I have yet to see a video card where the amount of RAM was not a multiple of the bus width that doubled the amount of RAM.

A 256-bit bus card was always configured with 256MB, 512MB (256 * 2) or 1GB (256 * 4). nVidia's 8xxx series were the same way. The 320-bit bus cards were available with 320MB or 640MB of RAM.

If this card has a 384-bit bus it is likely to come configured with 768MB of RAM, or 1536MB, simply because of the density values in which RAM modules are created, and the desire not to mix RAM types on the card.
Click to expand...

or they could use higher density chips and leave some of the bits disconnected to acheive 384-bits and effectively limit RAM size to 1152.
 
ryan_975 said:
or they could use higher density chips and leave some of the bits disconnected to acheive 384-bits and effectively limit RAM size to 1152.
Click to expand...

Again, this reduces the cost effectiveness of the manufacturing process by simply connecting RAM modules of the same density (size) and bus width.
 
ryan_975 said:
How so? They purchase more of the same chips and thus get a bulk discount?
Click to expand...

They already buy enough chips that they are getting as much of a discount as they are going to get. Also, using more chips is still more expensive that just not needing them at all.

Just look at the RAM in your computer. The bus is only 64-bits wide, so in theory memory modules could increase in increments of 64, but they don't. Modules have always doubled in size to keep module density and bit values constant. From as far back as using SIMMs, let alone DIMMs, modules size have increased at a size double that of the previous sized modules.

1MB, 2MB, 4MB, 8MB, 16MB, 32MB, 64MB, 128MB, 256MB, 512MB, 1GB, 2GB, 4GB, etc.

From a cost and quality standpoint, as well as historic standpoint, you can just about say with assurance that a 384-bit card will have 768MB or 1536MB.
 
Un4given said:
They already buy enough chips that they are getting as much of a discount as they are going to get. Also, using more chips is still more expensive that just not needing them at all.

Just look at the RAM in your computer. The bus is only 64-bits wide, so in theory memory modules could increase in increments of 64, but they don't. Modules have always doubled in size to keep module density and bit values constant. From as far back as using SIMMs, let alone DIMMs, modules size have increased at a size double that of the previous sized modules.

1MB, 2MB, 4MB, 8MB, 16MB, 32MB, 64MB, 128MB, 256MB, 512MB, 1GB, 2GB, 4GB, etc.

From a cost and quality standpoint, as well as historic standpoint, you can just about say with assurance that a 384-bit card will have 768MB or 1536MB.
Click to expand...

I understand that. But history doesn't necessarily always predict what the manufacturer is going to do. There is no reason (other than cost) that they can't Take the GTX and 8000 series nvidia cards. History would indicate that they would have gone from 256bit to 512, but they didn't. They went to 384, 320, 448, etc. ATI may decide to something similar with individual chips. They probably won't though.
 
you guys understand that the data bus width has absolutely nothing to do with the amt of ram that a video card can address, right?

ATI and Nvidia could access up to 4GB of video ram with a 32bit address bus....... the reason you get the "wierd" memory amts on cards has to due with how they arrange the video ram so that they can access x # of bits at a time.......Ati chooses to keep it more traditional as they have progressed from 16 to 32 to 64 to 128 to 256 and even 512bit bus widths.............
Nvidia went another route but the math still works out fine
 
If they did have a 5770 and it followed the same ideals envolved with the creation of the orginal 4770, it would be a 32nm chip just to try out the new process and wouldn't happen for a year or so.
 
Here are my guessstimates for pricing on release.

5890 $499
5870 $349
5850 $199

5690 $175
5670 $125
5650 $99

4890 $150
4870 $100
4850 $75
 
vjcsmoke said:
Here are my guessstimates for pricing on release.

5890 $499
5870 $349
5850 $199

5690 $175
5670 $125
5650 $99

4890 $150
4870 $100
4850 $75
Click to expand...
so no 5870x2? also those prices dont make a whole lot of sense. I certainly wouldnt try to guess at the prices especially considering we dont even know what cards will actually come out.
 
Why the hell would there be a 5890, that makes no sense. The 4890 like the 770 were oddities, not part of normal practice.
 
According to that chart a 5870x2 would cost $699.
It's just fun guessing ahead of time.
If ATI continues their naming conventions then this list is probably pretty accurate on what cards will come out.
Can't tell for sure of course.

necrophile said:
Why the hell would there be a 5890, that makes no sense. The 4890 like the 770 were oddities, not part of normal practice.
Click to expand...

Why? Because a higher end card nearly always comes out later after the main card has shipped and been proven.
We've seen this pattern with both ATI and Nvidia over the years. It's just common sense, even if you don't think there should be a 'higher' end.

Observe the following chart:
http://en.wikipedia.org/wiki/Comparison_of_ATI_graphics_processing_units

2002 radeon 9700
Jul 2002 radeon 9700 pro

2003 radeon 9800
Mar 2003 radeon 9800 pro

Oct 2003 Radeon 9800 XT

Etc. etc. ad absurdium. The 4890 can be considered the 'pro' or 'xt' version of the 4870. It's not very hard to follow.
Therefore it would not be surprising at all when, not if a 5890 model followed the release of the 5870.
 
Last edited:
So why would it be coming out at the same time and higher priced?

Most later cards come out and replace something, the most recent I can think of was the x1950 series although there might be more recent but still, no reason to predict something that probably won't exist and certaintly won't be priced like that.
 
at $699 they must not plan to sell many of those. I cant wait to see all the people that will be ridiculously cpu bottlenecked by one of those let alone two of them. really if you dont have an overclocked i7 then dont even waste your money.
 
necrophile said:
So why would it be coming out at the same time and higher priced?

Most later cards come out and replace something, the most recent I can think of was the x1950 series although there might be more recent but still, no reason to predict something that probably won't exist and certaintly won't be priced like that.
Click to expand...

Who made you the "nobody can guess what might happen in the future police?" It's already been proven that a 5890 is likely to follow the release of the 5870. The 4890 is NOT an anomaly. It's the pro/xt version of the 4870. You were WRONG on that count.

Who cares if the prices are right, you don't know what it's going to be either. Like I said it's a guess, and you haven't even made one, nor offered a better one, so you have no way to say if my prices are wrong until the cards actually come out.
 
The 5870 at $250 seems more realistic to me. ATI probably wants to keep using descent price points to attract customers, and keep gaining market share I would guess.
 
Proven, what are you talking about. The 4890 was just an updated 4870 which had to have a new name for clarity.

It does not mean they are going to do the same for the next series. The 4890 was a reworked 4870, the 770 was a 40nm test.

Their is no certainty that either situations will be needed again. Also they sure as hell ain't bringing any single gpu card out at that price, the market would turn on them IMO.
 
PS3 said:
The 5870 at $250 seems more realistic to me. ATI probably wants to keep using descent price points to attract customers, and keep gaining market share I would guess.
Click to expand...
well considering the 4870 came out at $299 thats likely to be the minimum msrp we will see for a 5870.
 
Also, most of the xt's pro's of yesteryear were mostly due to binning. THe 4830 and 4850 are like the xt and pro's with the 70 being the xtx.

It would be rare that a 90 situation occurs. If memory serves me right, which it may not things like the 1950 series was due to the shift of production process like the 4770, again nothing like the recent sitution.
 
My guess is that a "5870" model would be no more than $229.99 MSRP at release.
 
Here are my guesses for the pricing, pretty much the same as the 48xx series.

5850 - $199
5870 - $299
5870X2 - $549
 
The reason I think pricing will be less is that with a smaller 40nm GPU ATI can save money on production costs and possibly pass the savings to their growing market for domination over NVIDIA and all other competitors.
 
SonDa5 said:
The reason I think pricing will be less is that with a smaller 40nm GPU ATI can save money on production costs and possibly pass the savings to their growing market for domination over NVIDIA and all other competitors.
Click to expand...
they still have a lot of money and time invested in the new cards so they certainly will not cost less. also with no direct competition at release why would they sell them at cheaper than the last gen cards were?
 
Status
Not open for further replies.
Back
Top