hmm, I have "old" Ge 4 Ti 4200 by XFX and "new" Ge 4 Ti 4200 by XFX seemed to have different RAM placement...
I see that these (4400 and 4600) have similiar design with easy RAM access. With all 4600 loooking the same, 4400 the same and 4400-4600 very similiar, making a complete-VGA block for many ppl wouldn't be a bad idea.
but... RAM from my 4200 is more rectangular, on both sides and almost on edges of card - hard to cool with water. Some older cards also had that nasty RAM positioning. Anyway that block doesn't have to be for everyone.
Generally there are 3 ways to make a block for whole VGA, when we cool it from one side only.
1. monolithic block fo whole card, water flows over RAM & GPU, there is proper flow distribution inside, multilevel bottom needs tricky drilling (a lot of precision needed as RAM & GPU have diffferent heigh, problem may appear with users who do some lap job with their GPUs)
good things:
- performance should be nice on cards for which block was made (including lapping level)
- looking nice
- easy to mount
bad things:
- hard to make
- lots of precision needed
- heavy
- without lapping Ge4 Ti have far worse cooling and block cannot be for both ones with and without lapping
2. Set of connected blocks, which may be connected in series or paralelly or some kind of hybrid of first two.
"block" consists of 2 or more blocks. One is for GPU, and there might be one for whole mem (multilevel drilling needed too, to avoid touching some stuff between mem), or 2 for 2 chips each or 4 - each for each mem chip.
good things:
- if designed for this purpose, RAM block(s) might work with almost any GPU block
- will work nice with both lapped and not lapped GPUs
- if there are slight differences between cards, this design should be more portable
bad things:
- many blocks = not so easy to mount
- with more than one block for mem, mounting may be more complcated
- need to deal with intake and outakes in many blocks (problem with "compatiblity" with GPU blocks, which may be solved in the way to make it work only with some or only with yours
)
3. One really big block on GPU, to which there will be connected copper/heatpipe sticks from mem.
There are no blocks for mem, but only copper plates on them, all connected with huge block on GPU which would take heat from all parts.
good things:
- possibility to use big, complicated GPU block
- will work with both lapped and unlaped GPUs
- relatively easy to manufacture
- easy to mount
- with small design changes (like additional RAM "heat sticks" mountings), it would be possible to make block for both ATi & nVidia cards
- if heat from RAM was transferred to kind of thick plate placed on GPU, on which proper block was mounted, it would be possible to have all-vga TEC cooling, but this way heatpipes shouldn't be used but copper sticks due to other range of temperatures, there might be probably used heat pipes but operating under TEC conditions (it's my previous all-VGA block design which I haven't made as I'm no longer interested in TEC cooling, maybe in future but not now)
- innovative design
bad things:
- RAM won't be cooled as nice as in solutions 1 and 2 but still very well, with even possibility of neat TEC cooling
- need to use some thermal insulation for whole card o avoid condensation when using TEC...
- innovative design
In solution 3. "innovative design" is in both god and bad sides. It's this way because this may be a good solution and in "bad things" because ppl haven't tried it and may be not perfect.