Originally Posted by article Behind the clock speeds, the cut-down GF114 GPU remains the same. This means that it has seven streaming multiprocessors (SMs), providing a total of 336 stream processors, rather than the eight SMs of a full-fat GTX 560 Ti 1GB. The loss of an SM also drops the tessellation unit count to seven (from eight), and the ROP count down to 28 (from 32).
Once again this is wrong. The ROP count on the GTX560 is 32 just like the GTX460. The ROPs and memory controllers are seperate from the Shader Modules and are grouped 8 at a time (4 memory controllers with 8 ROPs each; for a total of 32). The GTX560 (460) have all memory controllers operational and thus it has 32 ROPs.
The below link shows a high level overview of the GF114(104):
Originally Posted by Cei This is what the 560 should have been at launch, a significantly cheaper card with good performance.
The GTX560 was available at this price from launch. I remember seeing a few below £150 on overclockers right on launch day. BIT only reviewed a very expensive model that had no reason to exist (thanks to the GTX560TI and HD6950).
Originally Posted by xaser04 The GTX560 was available at this price from launch. I remember seeing a few below £150 on overclockers right on launch day. BIT only reviewed a very expensive model that had no reason to exist (thanks to the GTX560TI and HD6950).
Ah, fair point. I hadn't looked too closely as I've got a 560Ti ;)
is it me, or does it just show that none of these new cards are really worth it, unless you can tell the difference between DX10 and DX11?
The 4870 provided a perfectly playable framerate in most of the tests, and up to large resolutions. why would we want to upgrade during this generation? My 4890 will be running until it gets very long in the tooth.
Originally Posted by xaser04 The ROP count on the GTX560 is 32 just like the GTX460.
No other review I have read has shown anything but 32 ROPs for this card, where are BIT getting a reduction of 4 from?!
D'oh, this is, of course, correct. Sorry for any confusion. Copy changed.
In regards to the 4870 and other DX10 GFX card numbers, yes, they do use the DX9 renderer, and are included for legacy comparisons. If people think this is misleading somehow, I'll remove them, and all the non-DX-11 compliant cards from the results, but I always thought they were a nice legacy inclusion.
While we were aware of cheaper GTX 560 (Non-Ti) at launch, these weren't the cards partners were interested in sending in. Happily prices have settled now, with the GTX 560 carving out a space in the market, especially now the GTX 560 Ti is much more expensive than it has been.
Originally Posted by Baz The 4870 numbers are, ofc, using the DX9 renderer, and are included for legacy comparisons. If you think this is misleading somehow, I'll remove them, and all the non-DX-11 compliant cards from the results, but I always thought they were a nice legacy inclusion.
NoNoNoNo!!! Don't remove them!
I, along with many I suspect, still have a 4870 card.
We always whinge that there isn't a slightly older card in the mix for comparison, so I love the fact it is in there. Even those from the *other side* can see the numbers and get a good feel for how their <insert countless G92 cards here> compares.
If people are having an issue, perhaps a bracketed (DX9) after the 4870 would do?
Originally Posted by Baz Sorry legacy fans, HD 4870 and GTX 260 (216) results removed. The results will still be in older reviews though, so you can check back to them.
TBH the HD5770 can be used as a good proxy for GTX260/HD4870 performance so I don't see a problem with removing the latter cards from the graphs to avoid confusion. I would however like to see them included in games that arn't DX11.
Originally Posted by Baz Sorry legacy fans, HD 4870 and GTX 260 (216) results removed. The results will still be in older reviews though, so you can check back to them.
Why the hell you do that??? Cei and Adoctum, the only two to comment on this point so far where both infavour of keeping them. So am I. Atleast let's have some debate.
Originally Posted by Baz Sorry legacy fans, HD 4870 and GTX 260 (216) results removed. The results will still be in older reviews though, so you can check back to them.
Poor decision, seriously. There's definite value in being able to compare older cards, even if they're running DX9/10 - I'd imagine that most people wait 2-3 generations, if not longer, to upgrade GPUs, so they need to see what their old card stacks up like. When I was purchasing my 560Ti, I was trying to compare with an 8800GT...
Stick a label in brackets afterwards - ie: ATi HD4870 (DX9)
I've asked Harry to re-insert the test results of the old cards into the graphs, so they'll be back in shortly. We'll now mark them with a DX9 or DX10 appendage where appropriate to flag up its the same game, but not the same test settings as the more modern DX11 cards.
I don't see the point in including the 4870, I love my card to death, but admit it's getting on a bit now and it's performance is around the 5770 minus DX11 features (yes it can tessellate, but thanks to AMD slipping it in early, no developer bothers to do the AMD specific calls to get it working).
Including it in benchmarks is only fair if it is stacked with the same settings!
Thanks Lizard, that's the sensible solution. Sorry, but you don't just completely remove stuff that's still very relevant. Alot of people are still running these older cards because they're holding up, and when reading reviews it makes it easier to see where you stand and how upgrades are looking.
Originally Posted by Glix I don't see the point in including the 4870, I love my card to death, but admit it's getting on a bit now and it's performance is around the 5770 minus DX11 features (yes it can tessellate, but thanks to AMD slipping it in early, no developer bothers to do the AMD specific calls to get it working).
Including it in benchmarks is only fair if it is stacked with the same settings!
but useful for users upgrading from 4850 cards see what they are getting for an upgrade
Originally Posted by Glix I don't see the point in including the 4870, I love my card to death, but admit it's getting on a bit now and it's performance is around the 5770 minus DX11 features (yes it can tessellate, but thanks to AMD slipping it in early, no developer bothers to do the AMD specific calls to get it working).
Including it in benchmarks is only fair if it is stacked with the same settings!
but useful for users upgrading from 4850 cards see what they are getting for an upgrade
As a 4850 owner who is upgrading later in the summer i say a bit fat yes to this. Comparing what current cards can do compared to eachother is handy, but its nice to know if im going to be betting 50% better performance in x game than i do know or 200% better. Anand has its GPU bench database to compare, but the 4850 was only tested on the old setup and so isnt included in the latest database, so you need to cross reference 2 databases to work out roughly how it'll improve your gaming.
Comments 1 to 23 of 23
ReplyOnce again this is wrong. The ROP count on the GTX560 is 32 just like the GTX460. The ROPs and memory controllers are seperate from the Shader Modules and are grouped 8 at a time (4 memory controllers with 8 ROPs each; for a total of 32). The GTX560 (460) have all memory controllers operational and thus it has 32 ROPs.
The below link shows a high level overview of the GF114(104):
http://www.anandtech.com/show/4344/nvidias-geforce-gtx-560-top-to-bottom-overclock
No other review I have read has shown anything but 32 ROPs for this card, where are BIT getting a reduction of 4 from?!
The GTX560 was available at this price from launch. I remember seeing a few below £150 on overclockers right on launch day. BIT only reviewed a very expensive model that had no reason to exist (thanks to the GTX560TI and HD6950).
Ah, fair point. I hadn't looked too closely as I've got a 560Ti ;)
The 4870 provided a perfectly playable framerate in most of the tests, and up to large resolutions. why would we want to upgrade during this generation? My 4890 will be running until it gets very long in the tooth.
D'oh, this is, of course, correct. Sorry for any confusion. Copy changed.
In regards to the 4870 and other DX10 GFX card numbers, yes, they do use the DX9 renderer, and are included for legacy comparisons. If people think this is misleading somehow, I'll remove them, and all the non-DX-11 compliant cards from the results, but I always thought they were a nice legacy inclusion.
While we were aware of cheaper GTX 560 (Non-Ti) at launch, these weren't the cards partners were interested in sending in. Happily prices have settled now, with the GTX 560 carving out a space in the market, especially now the GTX 560 Ti is much more expensive than it has been.
NoNoNoNo!!! Don't remove them!
I, along with many I suspect, still have a 4870 card.
We always whinge that there isn't a slightly older card in the mix for comparison, so I love the fact it is in there. Even those from the *other side* can see the numbers and get a good feel for how their <insert countless G92 cards here> compares.
If people are having an issue, perhaps a bracketed (DX9) after the 4870 would do?
The last review with them is http://www.bit-tech.net/hardware/graphics/2011/06/10/sapphire-hd-5850-xtreme-1gb-review/
Thanks Baz for correcting the ROP point.
TBH the HD5770 can be used as a good proxy for GTX260/HD4870 performance so I don't see a problem with removing the latter cards from the graphs to avoid confusion. I would however like to see them included in games that arn't DX11.
Why the hell you do that??? Cei and Adoctum, the only two to comment on this point so far where both infavour of keeping them. So am I. Atleast let's have some debate.
Poor decision, seriously. There's definite value in being able to compare older cards, even if they're running DX9/10 - I'd imagine that most people wait 2-3 generations, if not longer, to upgrade GPUs, so they need to see what their old card stacks up like. When I was purchasing my 560Ti, I was trying to compare with an 8800GT...
Stick a label in brackets afterwards - ie: ATi HD4870 (DX9)
Including it in benchmarks is only fair if it is stacked with the same settings!
but useful for users upgrading from 4850 cards see what they are getting for an upgrade
As a 4850 owner who is upgrading later in the summer i say a bit fat yes to this. Comparing what current cards can do compared to eachother is handy, but its nice to know if im going to be betting 50% better performance in x game than i do know or 200% better. Anand has its GPU bench database to compare, but the 4850 was only tested on the old setup and so isnt included in the latest database, so you need to cross reference 2 databases to work out roughly how it'll improve your gaming.
June is coming to an end, are we skipping the "Hardware Buyers Guide" this month?
-
« Previous
-
1
-
Next »
Discuss in the forums