Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Status
Not open for further replies.
How much memory does a good rendering box require?

When you think about it, DDR3 DIMMS are now readily available in 2GB sticks.

If they offered a board with 12 DIMM slots (2 per channel x 3 channels x 2 processors) fully populated, it would have 24GB. That seems insane to me. What desktop system needs that much RAM?

To put this in context, I just parted out a high-end Vista gaming rig that was hard-pressed to use the full 4GB of RAM I had in that machine.

If I were to buy a MP today, I would probably only configure it with 6GB (1GB sticks x 3 channels x 2 processors).

I guess the bottom line is, would MP users find it limiting to have only 6 DIMM slots with a capacity of 12GB of RAM?

"640k of RAM is enough for anyone"

That quote is always so safe and effective....but seriously, expandability never hurt anyone. It improves the machines useful lifespan, if nothing else.
 
While I know nothing about the design and manufacture of Apple's board, I can guarantee that it is *NOT* based on an existing Intel board. (Because I know all about the Intel boards.)

The Mac Pro uses the 5400X chipset. Intel has two boards that use the 5400-series chipset.

The S5400SF server board, which has 16(!) DIMM slots flat on the board, and a single PCI-e x16 slot, since it's meant to be used in a 1U rack-mount chassis. (Indeed, it is sold exclusively as the "SR1560SF" server system, not as a lone board.)

The D5400XS is the ultra-high-end dual-socket desktop board, which has only four memory slots, flat on the board; but a raft of expansion slots, including four PCI-e x16 slots.

Neither one looks *ANYTHING* like the Mac Pro's motherboard. And neither one uses a memory riser.

Oh, and 4 GB sticks of DDR-3 are available. So six sticks would be 24 GB. Eighteen sticks (three per channel, per processor,) would be 72 GB. I commented that it would be funny to have that much RAM in a system with three of the new 32 GB ultra-extreme Intel SSDs in a RAID-5. You wouldn't even have enough space to suspend-to-disk, much less hold an OS or data.
I'm thinking in terms of systems engineering, not appearance. A few modifications due to custom specs can make a significant difference. Risers, repositioning the CPU sockets for additional PCIe socket length, and the location of the ODD_SATA ports would. Then there's the color selection of the solder mask and silk screen legend. ;) Add in a few other changes, such as firmware, and possibly a different component here and there (i.e. NIC chip, etc), and you have a custom board that physically resembles nothing else made by the same company, but is still very similar as far as a schematic diagram (not layout). Modified reference design. :D

I guess it depends on how you look at it; schematic, or final product. ;) :p
 
"640k of RAM is enough for anyone"

That quote is always so safe and effective....but seriously, expandability never hurt anyone. It improves the machines useful lifespan, if nothing else.

I agree that more expansion is better, but a product manager has to make trade-offs... 6 DIMM slots, 12 DIMM slots, 18 DIMM slots... all have trade-offs associated with them.

So what's the most RAM anyone here would consider running in their MP?

6GB for me.
 
That's pretty funny. Just how big do you think the molecules in air are exactly? Hole size isn't the issue.....it's the ratio of open area to closed area.

Well, the open amount looks to me to be inadequate to provide enough flow. You knew that's what I meant, smartass, not the size of air molecules. My position is it doesn't "look" to be enough. Are you able to prove it does? My position doesn't need to be proved or disproved.It looks too constricting. You won't even say if it is or not.
 
Well, the open amount looks to me to be inadequate to provide enough flow. You knew that's what I meant, smartass, not the size of air molecules. My position is it doesn't "look" to be enough. Are you able to prove it does? My position doesn't need to be proved or disproved.It looks too constricting. You won't even say if it is or not.

To be quite honest, I didn't know what you meant. You said the "holes are too small to pull enough air thru" and that is all I had to go on.

Looking at the image, it is impossible to tell if the ratio of open area to closed area is less, the same, or greater. We would need to see a closeup of the pattern to know for sure.

S-
 
To be quite honest, I didn't know what you meant. You said the "holes are too small to pull enough air thru" and that is all I had to go on.

Looking at the image, it is impossible to tell if the ratio of open area to closed area is less, the same, or greater. We would need to see a closeup of the pattern to know for sure.

S-

I actually disagree. Air, as any fluid, has viscosity and is also subject to friction. This means that the smaller the holes get, the greater the overall area of holes vs solid to achieve the same capacity of flow. Air particle size is virtually irrelevant... besides, which part of 'air' are you referring too? Oxygen, Nitrogen, CO2... etc etc?

Oh, and it is pretty easy to see from that image that the overall area of holes is significantly less than the current design, not to mention the fact they are smaller...
 
That's a crapload of SATA connectors.
What is interesting is I can't find either board on ASUSs website at all. I was wanting to see if the colors meant anything. I would assume they have to do with the on board RAID, but am not positive.
 
Just wondering, is there any reason why a 285 or 295 graphics card is off the table?

I know that Apple has historically used lower end cards (except the pro option). Apple products aren't game machines, they are for digital lifestyle, or work. But CSS4 (and similar apps) now use CUDA. CUDA (or openCL - the big feature in SL) is a pretty huge deal, and a good graphics card will give a lot of bang for your buck.

Given how good GPUs are at handling graphics, and given how much better a 295 is than the 260, would a single socket i7 with a 295 be able to run CSS as fast or faster than a dual socket Gainstown MacPro with a 260?
 
CUDA / openCL is not just for gaming. Photoshop, CAD programs, it's going to be everywhere.

The 285 has lower TPD to the 260 (182W vs 183W), and lower idle power use, doesn't it? The 295 might be a bit hot though, at 289W.
 
CUDA / openCL is not just for gaming. Photoshop, CAD programs, it's going to be everywhere.

The 285 has lower TPD to the 260 (182W vs 183W), and lower idle power use, doesn't it? The 295 might be a bit hot though, at 289W.

So far, Apple has never used high end consumer cards. You may be able to get the GT200 core through Apple maybe offering a quardo card, but that will be about it. It is mostly due to Apple's insistence on including video drivers as apart of the OS and not as a separate downloadable. While this makes for better stability it also means features a card can do go unused for a long time. Plus since Apple only updates video drivers every once and a while new cards are made unusable (where as on a PC at least you get some sort of basic compatibility mode). The ones they do write are for specific cards, which seems silly because if Apple were using OpenGL you should be able to use any card.
 
Just wondering, is there any reason why a 285 or 295 graphics card is off the table?

I know that Apple has historically used lower end cards (except the pro option). Apple products aren't game machines, they are for digital lifestyle, or work. But CSS4 (and similar apps) now use CUDA. CUDA (or openCL - the big feature in SL) is a pretty huge deal, and a good graphics card will give a lot of bang for your buck.

Given how good GPUs are at handling graphics, and given how much better a 295 is than the 260, would a single socket i7 with a 295 be able to run CSS as fast or faster than a dual socket Gainstown MacPro with a 260?

I believe it is a case of Apple not needing to offer a large range of GPUs to sell systems to the customers they are targetting. There are downsides to offering more options for Apple, though they may seem irrelevant to us as consumers wanting choice. The 295 is basically ruled out as they would need SLI drivers (I believe) which don't seem like something Apple are willing to deal with.

Maybe we will see things change with the change in role of the GPU, but I wouldn't expect anything this year.
 
So far, Apple has never used high end consumer cards. You may be able to get the GT200 core through Apple maybe offering a quardo card, but that will be about it. It is mostly due to Apple's insistence on including video drivers as apart of the OS and not as a separate downloadable. While this makes for better stability it also means features a card can do go unused for a long time. Plus since Apple only updates video drivers every once and a while new cards are made unusable (where as on a PC at least you get some sort of basic compatibility mode). The ones they do write are for specific cards, which seems silly because if Apple were using OpenGL you should be able to use any card.

I guess it depends what you consider high end. The 7800GT (G5), 1900XT and 8800GT were in the high end bracket of cards as defined by the manufacturers at the time of release. They just weren't the highest end in those brackets, which makes sense as such cards have a huge premium for little performance gain.
 
I believe it is a case of Apple not needing to offer a large range of GPUs to sell systems to the customers they are targetting. There are downsides to offering more options for Apple, though they may seem irrelevant to us as consumers wanting choice. The 295 is basically ruled out as they would need SLI drivers (I believe) which don't seem like something Apple are willing to deal with.

Maybe we will see things change with the change in role of the GPU, but I wouldn't expect anything this year.

So high-end consumer graphics cards are a "nascent" market?
 
I agree that more expansion is better, but a product manager has to make trade-offs... 6 DIMM slots, 12 DIMM slots, 18 DIMM slots... all have trade-offs associated with them.

So what's the most RAM anyone here would consider running in their MP?

6GB for me.

32GB. HDR Photoshop, 3D animation and multi track audio with lots of synths.
 
So high-end consumer graphics cards are a "nascent" market?

Just some areas. I would say CUDA/OpenCL and SLI/Crossfire (for non-games at least) are. Those areas are ones that could see Apple change their current ideology towards the GPU.
 
SLI/Crossfire would be very useful with OpenCL.

No it wouldn't. In fact it would be totally pointless. SLI/Crossfire allow the cards to work together to render scenes faster. OpenCL allows the use of the cards for general purpose computing: it can already use multiple cards rather like having multiple CPUs: it doesn't need SLI or Crossfire support.
 
I guess it depends what you consider high end. The 7800GT (G5), 1900XT and 8800GT were in the high end bracket of cards as defined by the manufacturers at the time of release. They just weren't the highest end in those brackets, which makes sense as such cards have a huge premium for little performance gain.

True, Apples biggest problem is they write drivers for specific cards versus using the family of cards.
 
No it wouldn't. In fact it would be totally pointless. SLI/Crossfire allow the cards to work together to render scenes faster. OpenCL allows the use of the cards for general purpose computing: it can already use multiple cards rather like having multiple CPUs: it doesn't need SLI or Crossfire support.

OpenCL like OpenGL would only be able to handle multiple cards (in SLI or otherwise) if the OS is aware of the cards, just because Snow Leopard has OpenCL doesn't mean that the GTX 295 is all of a sudden going to start working. It doesn't work now and OpenGL has been able to support SLI since it first came back (read: a while).
 
OpenCL like OpenGL would only be able to handle multiple cards (in SLI or otherwise) if the OS is aware of the cards, just because Snow Leopard has OpenCL doesn't mean that the GTX 295 is all of a sudden going to start working. It doesn't work now and OpenGL has been able to support SLI since it first came back (read: a while).

At an OS level the OSX is aware of the cards. OpenCL is designed to scale across the cards by default; OpenGL is not: the drivers "trick" OpenGL (and DirectX) into thinking that they pair (or more) of cards are actually one card. OpenCL doesn't need this.
 
At an OS level the OSX is aware of the cards. OpenCL is designed to scale across the cards by default; OpenGL is not: the drivers "trick" OpenGL (and DirectX) into thinking that they pair (or more) of cards are actually one card. OpenCL doesn't need this.

At a hardware level the system knows there is a card in it. At an OS level it know there is a card in it. It doesn't know what the card is thus the OS either doesn't boot or it kernel panics. Or is Apple finially getting with the time and providing a generic driver that any GPU would run with a la Windows?
 
Status
Not open for further replies.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.