Upgrading Mac Pro 2008: 8GB RAM or 2x Vertex 3 60GB SSD's?

[Goldfinger]

Apparantly your MP does need that extremely expensive DDR2 RAM, unfortunately. I upgraded my 2010 Mac Pro a few months ago to 16GB (DDR3 ECC) and it was less than around €80 IIRC (for 8GB).

I really wonder why DDR2 is that much more expesive.

And yes that 660MB/sec figure is indeed correct. I was confused before.

 

[nanofrog]

SATA-II has a nominal 300MB/s throughput per channel. SATA use 10/8 encoding so it's 300MB/s not 375MB/s. As with all other busses, there will be overhead and losses from this nominal figure that bring real-world throughput down.

Max Available Bandwidth for 3.0Gb/s SATA = 375MB/s (think in terms of units; capacity per second in this case).

The 8b/10b encoding scheme is nothing but a ratio in terms of how it affects throughput (8/10 = .8, so 375MB/s * .8 = 300MB/s).

Granted, you could consider this picking nits, but the units do not denote any encoding scheme.

As such questions tend to want real world results, I tend to go straight to that answer, which in this case, is ~ 275MB/s per port (latency accounts for the remaining performance loss).

In the case of the ICH limit, Intel set that one as a means keeping the DMI interconnect from being saturated by any particular controller (for example, SATA throughput could starve out other things, such as Ethernet). And instead of making a more complicated band management scheme (which would require more transistors), they chose to simply set limits per controller passing data over the DMI channel. In the end, it comes down to money; lower transistor count = lower production cost by increasing the part count per wafer (higher yields). It also lowers the R&D spent on chipset/ICH development at the time, further reducing costs (now part of this may be passed onto the end consumer, but it also allows for Intel to increase their margins as well).

 

[handheldgames]

One more thing... (had to say it)

Go over to NewEgg and read the reviews from users of the Vertex 3 SSD. Then read the reviews of the Samung 470 Series. A hella eye opener. May change your opinion on the drive your considering.

 

[speacock]

Max Available Bandwidth for 3.0Gb/s SATA = 375MB/s (think in terms of units; capacity per second in this case).

The 8b/10b encoding scheme is nothing but a ratio in terms of how it affects throughput (8/10 = .8, so 375MB/s * .8 = 300MB/s).

Granted, you could consider this picking nits, but the units do not denote any encoding scheme.

That is nit-picking

OK, there's 375MB/s available but as only 8 out of every 10 bits is used to carry user data, the effective maximum bandwidth of SATA-II is 300MB/s

We're in wild agreement that the net effect is that the real-world max is something slightly less than 300MB/s though.

----------

Apparantly your MP does need that extremely expensive DDR2 RAM, unfortunately. I upgraded my 2010 Mac Pro a few months ago to 16GB (DDR3 ECC) and it was less than around €80 IIRC (for 8GB).

I really wonder why DDR2 is that much more expesive.

And yes that 660MB/sec figure is indeed correct. I was confused before.

Someone mentioned it before - supply and demand - there's not much of it being made any more.

 

[sparkie7]

One more thing... (had to say it)

Go over to NewEgg and read the reviews from users of the Vertex 3 SSD. Then read the reviews of the Samung 470 Series. A hella eye opener. May change your opinion on the drive your considering.

Samsung 470 is only SATAII and SLOW compared to the Vertex 3. Most failures on Vertex 3 seem to be on PC.

I have had a Vertex 250 for 2 years, no issues (after they shipped me a replacement LOL) since!

 

[nanofrog]

Samsung 470 is only SATAII and SLOW compared to the Vertex 3. Most failures on Vertex 3 seem to be on PC.

Other makes using the same controller are also having issues (SandForce 2281). As per seeing most of these failures under PC's, that makes sense, as they have a much larger portion of the marketshare.

But do your self a favor, and don't look at this as a PC vs. Mac situation, as it isn't (enough information out there on SF-2281 based disks to show that it's the controller, not the system). Some have mitigated things with firmware adjustments, but they may not have it all sorted.

BTW, OWC's disks based on the SF-2281 seem to be in better shape than the Vertex 3 last I saw (they had released a new firmware revision that mitigated the issue, if it didn't completely solve it).

There's a thread in here somewhere on this as well, so it might be worth a search (links you might want to look over).

 

[sparkie7]

ok thanks nanofrog