Really confused about PCI Express 3.0

G

Guest

Guest
Hopefully someone can help me understand this before I pull the trigger on a motherboard for my current build.

I read about how Gigabyte and MSI have released BIOS revisions that enables PCIE 3.0 on existing B3 boards, essentially making them Gen3 boards. To me this implies that there's really not difference between a PCIE 2.0 and 3.0 slot. Sorry if I'm dumb for pointing that out, but I haven't built in awhile and I'm just trying to confirm. So the slot itself must be identical, yes?

Furthermore, my understanding is that Ivy Bridge is going to support PCIE 3.0 whereas Sandy Bridge currently does not. So, does PCIE 3.0 basically reside on the CPU? In other words, the hardware for PCIE 3.0 has been sitting on motherboards for awhile now, just waiting to be utilized by Ivy Bridge?
 
You already seem to have a pretty good handle on the differences.

yes, they are the same physical slot.

And it is the architecture on the CPU that supports the higher data rates, which is why SB do not support it, but IB does.
So even though PCI3.0 is on boards, it will not be usable until Ivy Bridge comes out.
 
G

Guest

Guest
Then why do they make the distinction on the motherboard specs? For example, right now if you search newegg and filter motherboards by socket 1155 and ATX there are only five boards that list " 2 x PCIE 3.0 slots". Those same five boards also list the number of PCIE 2.0 slots they have.

This makes me think there's more to it than just a slot and the right CPU. Why would they make that distinction of there weren't some version specific hardware on the board?
 
PCIe 3.0 FAQ - http://www.pcisig.com/news_room/faqs/pcie3.0_faq

It's going to be interesting if these early PCIe 3.0 MOBOs will actually work properly once PCIe 3.0 CPUs and GPUs are released.

I prefer a 'PCIe 3.0' Clean' environment so I don't run into so unforeseen/oddball bottleneck. I want PCIe 3.0 Clean: CPU, GPU, Chipset & MOBO.

Further, my 'biggie' is then bye-bye 'shared' bottleneck on everything: SATA, USB, add-on chipsets e.g. Marvell or whatever that ALL interfere with today's 'Swiss Army' styled MOBO. As far as the PCIe 3.0 to the GPU I expect little to NO difference on ANY GPU for quite sometime. A GTX 590 cannot saturate PCIe 2.0 x16 - it's not even close.

For 'Today's GPUs' the 4-GPU core limit would have to be lifted, example 4-WAY GTX 590's (8-cores), to hit PCIe 3.0 x8 and forget PCIe 3.0 x16. The new SB-E/LGA 2011 coming out in a matter of weeks has 32-lanes of PCIe 3.0, but there's nothing out there to make a dent.
 
G

Guest

Guest
It's going to be interesting if these early PCIe 3.0 MOBOs will actually work properly once PCIe 3.0 CPUs and GPUs are released.

Yeah that pretty much sums up my level of trust right now. I know ASRock isn't the byword it used to be which is great, but I'd feel better with a Gigabyte or ASUS board. I'm about ready to forget it and get a solid motherboard from a good name but without any PCIE 3.0 compatibility and just enjoy it.

The last time I built a gaming system was 2005 and back then there were so few motherboards that would run two lanes at 16x in SLI. Amazingly six years later that really hasn't changed. On top of that I've seen several articles and videos proving a mere 1-2% when going from dual 8x to dual 16x. Meh...

I don't know why my shorts are in a knot over PCIE 3.0 when like you said it probably won't matter and by the time it does I'll be ready to rebuild anyway.
 

Now double the bandwidth and dilute the 1%~2% into nothing.

The 'real' advantages are going to be as I mentioned the currently 'shared' remaining PCIe bandwidth to the other components: SATA, USB, add-on Chipsets, etc -- those speeds and performance are the real winners of PCIe 3.0. Soooo the PCIe 3.0 GPU only MOBO's e.g. ASRock 'Gen 3' are mostly to 'sell' their MOBOs.

ASUS is {owns} ASRock and they share, obviously, technologies and etc.