Maxing out your PCI-E bus?

[SPL Tech]

How can one determine the data transfer speed of a particular video card? For example, PCI-E 1.0 x16 maxes out at 4GB a second. So say I threw a GTX Titan Black that was overclocked like a boss in a 1.0 slot. How would I know if the GTX Titan Black could generate enough traffic to max the slot out?

Second, it is technically possible to overclock a PCI-E bus, right? So one could get, say, 4.5GB/s out of a 1.0 slot maybe?

 

[GaldedRaw]

I think a titan would be bogged down by a PCI-E 1.0 x16 slot since it is made for a PCI-E 3.0. Though it is backwards compatible. I think there is bench mark type programs that can determine the data transfer rates. I read about it in a magazine, if only I had it on me, but I am at work and when I am back in my room the internet is so terrible it doesn't even work anymore. I cant get on the OCF at night .

I never heard of PCI-E overclocking.

 

[Silver_Pharaoh]

How can one determine the data transfer speed of a particular video card? For example, PCI-E 1.0 x16 maxes out at 4GB a second. So say I threw a GTX Titan Black that was overclocked like a boss in a 1.0 slot. How would I know if the GTX Titan Black could generate enough traffic to max the slot out?

Second, it is technically possible to overclock a PCI-E bus, right? So one could get, say, 4.5GB/s out of a 1.0 slot maybe?

On socket 1155 mobo's pci bus is tied to everything from the memory to the hard drive. Very unlikely you will be stable over 103 Mhz BCLK so any gains to the pci bandwidth would be very small.

That said, I'm not sure if a Titan will saturate the entire pci bandwidth...

 

[ATMINSIDE]

Test data

There you go, already been done.

 

[Silver_Pharaoh]

Test data

There you go, already been done.

So it appears that we are beginning to saturate Pcie 1.1 16x Bandwidth.....

 

[ATMINSIDE]

So it appears that we are beginning to saturate Pcie 1.1 16x Bandwidth.....

Yeah, as you can see there's a margin-of-error difference between 2.0x8/x16 and 3.0x8/x16.

2.0x16 has the same bandwidth as 3.0x8
People obsess over "well this doesn't have the latest PCIe". Truth is, unless you're chasing that last 1fps, its not necessary.

 

[EarthDog]

I am hoping wizzard posts an update. I have to imagine the 780ti/290x/Titan that story may change from 'margin of error' type differences... but could be wrong.

 

[ATMINSIDE]

I am hoping wizzard posts an update. I have to imagine the 780ti/290x/Titan that story may change from 'margin of error' type differences... but could be wrong.

It might, but I doubt we'll see full 2.0x16 saturation for a while still.

Edit: We're about due for another, its coming up to 2 years since his last, which is his gap from the review before the one I linked.

 

[Silver_Pharaoh]

I believe it was Earth_Dog(I maybe wrong) that stated "Bigger garden hose doesn't produce more water if it's connected to the same size spigot." (That's actually a paraphrase for those who are critical).

Truth is as long as you're running PCIe 2.0 x8 or better, you're good.

Pci 1.1 1x all the way!
For mining anyway


That's a good way to explain bandwidth to someone actually. Easy to understand!

 

[Bobnova]

Just look at GPU performance, simple enough.
The more data the card is demanding to generate frames, the more data has to go over the bus.

 

[Silver_Pharaoh]

I was playin' around with one of my new dust collectors last night. It has 3x pcie 1.1's. The CPU is a Celeron 500Mhz that only draws 27W. I'm doing some thinking, but that's for another thread.

Just FYI, if it's the old pci slots, adapters to make it pcie are $30

/OFF TOPIC

 

[EarthDog]

PCI bandwidth is not close to PCIe... 133 MB/s versus PCIe3.0 16x 15.75GB.

Don't do it. LOL!

 

[Silver_Pharaoh]

PCI bandwidth is not close to PCIe... 133 MB/s versus PCIe3.0 16x 15.75GB.

Don't do it. LOL!

Not for gaming, but mining? You could probably get away with it.
But it's for another thread

 

[EarthDog]

Maybe. I have no idea... If a 1x slot is 984 MB/s and the PCI slot is ~1/3 of that...


/back to your regularly scheduled programming.

 

[wingman99]

Here are some interesting facts.

QUOTE:While PCI-Express 1.0 pushes 250 MB/s per direction, PCI-Express 2.0 pushes 500 MB/s, and PCI-Express 3.0 doubles that to 1 GB/s. While the resulting absolute bandwidth of PCI-Express 3.0 x16, 32 GB/s, might seem overkill, the ability to push that much data per lane could come to the rescue of configurations such as 8-lanes (x8) and 4-lanes (x4). Another impressive feature of Ivy Bridge Core processors, provided they're paired with Intel Z77 Express chipset, is that the second x8 link from the CPU root complex can be split as two x4 links, making x8/x4/x4 possible, giving some motherboards 3-way SLI and CrossFireX capabilities without clogging the DMI chipset bus (that 4 GB/s pipe between the CPU and chipset), which is better left untouched by graphics cards to help with today's bandwidth-hungry SSDs.

 

[SPL Tech]

So the slot speeds do matter. Interesting. I am interested in knowing why there is ANY difference between the 2.0 X16 and 3.0 X16 since no card can max out those slots.

 

[EarthDog]

I'm sorry, but isnt that what the article is showing? That things are saturated? The percent is almost within the margin of error but we can see 4x 3.0 goes down in performance as well.

 

[Woomack]

Somehow in couple of benchmarks maximum FPS is higher on pcie 3.0/2.0 than 1.1 ... all depends what card. But minimum FPS is almost not changing so general experience in games is not really changing.
I guess that drivers are controlling additional options when pcie 2.0+ is enabled but I can't explain what or how.
For sure pcie type matters for some devices that will be limited by bandwidth or won't work at all. I mean RAID controllers and similar stuff. Once I couldn't enter RAID ROM on LSI controller when it wasn't in pcie 2.0/3.0 x8 or x16.

 

[GaldedRaw]

Its computer magic!

 

[Woomack]

Regarding pcie clock itself then at least on Intel boards it goes as high as bclk. On SB you can count on about 103MHz, on IB about 105-107, on Haswell with some luck up to 110. I mean stable as max is slightly higher.
At the same time some HDD will generate errors or shut down when you pass 105MHz, some other will work without issues up to 120MHz pcie. Generally SSD have higher tolerance for high pcie but still I see no point to overclock it.
If I find some time this weekend then I will try max on i3 4330 / ASRock Z87M OCF.