• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

2 Different HD4870X2's....

32bit os will map 4gb total, that includes all memory from GFX, system and other perif cards.
Also the word is that each of the 4870x2 cores has 256bit bus to each cores memory using a new bridge to bring them together.
Also as others said the bus width is not a problem.
 
gddr5 with 256bit bus is fine as it offers the same/better bandwidth that gddr3 with a 512bit bus. See bandwidth for the 4870 vs the GTX280.

Hence why ATI chose the gddr5 route and Nvidia stuck with the 512bit bus route. Two different ways of acheiving the same end result.

Gddr5 memory more expensive than gddr3 but 512bit pcb more expensive than 256bit pcbs.

Thats incorrect. 256-Bit bus is a joke, and even GDDR5 can hardly save it.

GTX 280 memory bandwidth = 141 GB/s

4870 memory bandwidth = 115 GB/s


ATI should have gave the 4870 a 384-Bit bus like the 8800GTX had.

As for the 4850 with it's 256-Bit bus and GDDR3 it only has 63 GB/s. With that very low bandwidth and only 512MB, the card is definitely not a good idea for high res monitors.
My GX2 had similar memory specs, but with a little more bandwidth (70 GB/s), and high res gaming on that was not a nice experience. My 280 is such a big improvement.
 
Last edited:
makes little difference when ati are more efficient with that bandwidth, unless you prescribe to the american everything-has-to-be-bigger way of thinking.

Anyway, to write it off as a joke shows a severe lack of understanding of the technology involved on your part. do not forget that the memory speed on the gtx280 is 1.1ghz. on the ati 4870 its 900mhz. what happens if the memory on a 4870 is running at 1.1ghz? 1.1ghz x 4 (quad pumped gddr5) = 4.4ghz effective on a 256bit bus? 141gb/sec of bandwidth.......do you see the connection? go ahead do the math. what about 1200mhz ram speed? 153gb/sec which funnily enough, is the same figure you arrive at when using 1200mhz gddr3 on a 512bit bus. remember at the end of the day, gddr5 transfers twice the info per clock that gddr3 does. twice per clock, half the bus width.....SAME bandwidth.


like greebo said, its not the width of the bus thats important, its the end result. even then, theres far more to it than raw bandwidth.
 
Last edited:
The extra 75W power supplied to them..
Hi Greebo, I was talking to some younglings at Xtreme and asked them a few questions regarding HD4850 vs HD4870.

One of the things I was interested to know was did the HD4870 have a big overclocking advantage over its little brother due to the fact it had two six-pin PCI-E feeds

HD4850: 75w PCI-E Slot + 75w PCI-E 6-pin = 150w

HD4870: 75w PCI-E Slot + 75w PCI-E 6-pin + 75w PCI-E 6-pin = 225w


The answer I kinda got was that the HD4850 could take as much power as it wanted and wasn't limited to 150w, so if thats true it means the HD4850 can be overclocked as hard as its bigger brother if you are able to perform the necessary vMods.

My understanding of these subjects are limited so I hope I managed to get my point across.
 
Aw did i hurt you feelings james? ;(

ATI cant do magic, they may be more efficient with bandwidth, but when you have 63 GB/s on the 4850 you can only do so much.
You can never have too much bandwidth, and the 4870/4850 could do with a 384-Bit bus.

But your'll likely run into the 512MB limit on high resolutions anyway before you hit any bandwidth problems.
 
Aw did i hurt you feelings james? ;(

ATI cant do magic, they may be more efficient with bandwidth, but when you have 63 GB/s on the 4850 you can only do so much.
You can never have too much bandwidth, and the 4870/4850 could do with a 384-Bit bus.

But your'll likely run into the 512MB limit on high resolutions anyway before you hit any bandwidth problems.

read my post. you are totally incorrect. gddr5 on a 256bit bus will provide exactly the same bandwidth as gddr3 on a 512bit bus.


there is a reason ati chose a 256bit bus, and thats cost. the cost of the memory will drop, the cost of the pcb wont. amd made an intelligent decision going with gddr5/256 and while nvidia could build a monster gddr5 card with a 512bit bus, remember they have to save money and fast. dont be too suprised if they switch to 256bit when they go gddr5. exactly the same bandwidth dont forget.

also, before you cite gtx and gx2 benchmarks as proof, remember you are only talking about nvidia hardware. have you actually looked at any high res ati benchmarks?
 
Last edited:
Aw did i hurt you feelings james? ;(

ATI cant do magic, they may be more efficient with bandwidth, but when you have 63 GB/s on the 4850 you can only do so much.
You can never have too much bandwidth, and the 4870/4850 could do with a 384-Bit bus.

But your'll likely run into the 512MB limit on high resolutions anyway before you hit any bandwidth problems.

You are so so wrong. So what if the 4850 only has 63 gb/sec bandwidth and not suitable for large res. It's a budget card. Comparing it to a gx2 and gtx280 and saying it's bandwidth is not enough is pointless. You are trying to compare it to cards costing 4x as much?

And your gx2 was performing poorly at high res not because of its 70gb/s bandwidth but because of mirco stuttering due it been SLI effectively on one card. Bandwidth doesn't come into it.

If it wasn't for the stuttering the gx2 beats the GTX280 at very high res in a lot of games even though the GTX280 has 141gb/s of bandwidth.

And do you think the GTX280 is actually using all of it's bandwidth?

And yes youn can have too much bandwidth.

As said gddr5 will be able to go to 8000Mhz effective speed and about 400gb/s bandwidth hence ATI moving over to gddr5. Nvidia with 512bit bus and gddr3 is stuck with a maximum bandwidth of 141gb/sec unless they get the memory on the GTX running at 3000Mhz which they will never do. Hence Nvidia moving to gddr5 in a few months.

Stop trying to defend your GTX280 purchase by making things up.
 
I was talking about pure bandwidth numbers at stocks speeds. And the 280 has more bandwidth then.

I agree going GDDR5 was a good decision by ATI. And i think the main reason NV did not do this is because the 280 core seems like a highly teaked and improved G80, and that core has no support for GDDR5.

And unrelated, does the 4870X2 2GB have both cores share memory from the 2GB? So there is actually 2GB usable VRAM? Or will there be 1GB usable VRAM?
...I'm guessing it's the later though as 2GB usable VRAM would make no sense at all at this point in time.
 
I was talking about pure bandwidth numbers at stocks speeds. And the 280 has more bandwidth then.

like i said - to write a 256bit bus off as a joke shows a severe lack of understanding of the technology involved on your part when the only cause of that drop in bandwidth is a lower memory speed, not the bus.
 
As for the 4850 with it's 256-Bit bus and GDDR3 it only has 63 GB/s. With that very low bandwidth and only 512MB, the card is definitely not a good idea for high res monitors
Hi Mr B,

I've been gaming on a high res monitor longer than most so I know what an uphill struggle its been to achieve playable framerates at 1920x1200, I started HD life powered by a Radeon 9800Pro and have ended up with a HD4850 and I can honestly say its a great card for high res, especially when using lots of AA.

It's not the best card out by any means but to me it certainly packs a massive punch for the price and I would have no trouble recommending it to gamers with 24" monitors! :)
 
And your gx2 was performing poorly at high res not because of its 70gb/s bandwidth but because of mirco stuttering due it been SLI effectively on one card. Bandwidth doesn't come into it.

It had nothing to do with SLI. I would run in to it without SLI enabled. And i know it was not bandwidth, it was almost completely to do with the 512MB usable VRAM not being enough.
And i will prove this once i get XP installed so VRAM usage monitoring software will work (none works on Vista yet) and display how much VRAM is being used while playing at high res. It will go well over 512MB usage at high res with AA.

If it wasn't for the stuttering the gx2 beats the GTX280 at very high res in a lot of games even though the GTX280 has 141gb/s of bandwidth.

Greebo, i've owned the cards, i've done a million tests. I've got a 2560x1600 monitor. Stop acting like you know what your on about here.
Ignoring the stutturing when you hit the 512MB limit, things are totally unplayable. A Slideshow, as it starts swapping with system memory. Theres no comparison between GX2 and 280 when this happens, the 280 will run the game WAY better. I always like to game at native 2560x1600 res, and any 512MB card just dont cut it, some games like GRID wont even run with no AA or AF because of still not enough VRAM. With 1GB VRAM i can now play that game with 16x AF + 4x AA at native res.


Stop trying to defend your GTX280 purchase by making things up.
:rolleyes:
 
Last edited:
i will [demonstrate] this once i get XP installed so VRAM usage monitoring software will work (none works on Vista yet) and display how much VRAM is being used while playing at high res. It will go well over 512MB usage at high res with AA
that sounds like a great idea and I'm sure people will find your results interesting!

Just have to say that perhaps ATI texture compression is helping to keep things running smoothly? I don't know too much about these things but if the compression is better on the Radeons then having only 512MB frame buffer is not such a big deal as you seem to believe?
 
Last edited:
@ 1920x1200 and below i don't think either the bus or the 512mb is a problem on the 48xx series as benchmarks show and those of us with the cards know.

30" monitors and i'd have to agree with mr.b about 512mb cards.
 
@ 1920x1200 and below i don't think either the bus or the 512mb is a problem on the 48xx series as benchmarks show and those of us with the cards know.

30" monitors and i'd have to agree with mr.b about 512mb cards.

Nvidia cards yes, ATI are better due to better texture compression but yes eventually you will hit that memory limit.

But Mr B can't have it all ways.

First its memory bandwidth which makes the 4850 not as good as the GTX280, then it's only 512Mb of ram, then the 4850 is not better than a GX2, blah, blah, blah.

Who ever said the 4850 was as good as a GTX280?

The 4850 is competing with a 8800GT, not a GTX280.

The 4870 is against the 9800GTX, again only a 512mb card.

Hence it doesn't need more than 512Mb and it doesn't need more memory bandwidth.

It does what it is meant to do, beat a 8800GT and is a great card for up to 1680 x 1050.

If the 4 series had 384bit bus and 1gb gddr5 like you wanted them too Mr B, then they would not cost £117 and £164.

There is a market for every type of card from budget to high end. It would be a sad world if both ATI and Nvidia only realesed their top high end £400 cards like you want them to.

I fail to see why you are so hung up about ATI's cards? Its bordering on pure hatred the way you describe them sometimes? :confused:
 
Last edited:
looks to me like the bandwidth does not help as much as you are wanting us to believe. In both of these games, the 4870 has almost 90% the performance of the GTX 280, the 4850 with a measley 60something gb/sec bandwidth has about 70% the performance. Not to mention having two 4850s will slaughter a GTX 280 for a fraction of the cost....




cod4-2560.gif
etqw-2560.gif
 
Last edited:
Back
Top Bottom