Motherboard tier list: H87 chipset

H87 motherboards also do not have full crossfire support. The PCI lanes run at x16/x4 speeds, the x4 PCI lane will cripple the second graphics card's performance.

isn't this exactly what the recent gpu myth article seemed to refute? that the 4x pipeline is more than enough and does not seem to hinder ANY card since usage was VERY low.

Which article?

this one http://www.tomshardware.com/reviews/graphics-performance-myths-debunked,3739.html

it looks into measuring the bandwidth used during gaming over the pcie lanes and shows some surprisingly minimal usage

Sorry for the late reply, but yes there is a noticeable difference between x4 and x8 or x16.

@MeteorsRaining, the same issue with PCI lanes running at X4 speeds can be found on lower end Z87 and Z97 motherboards, so you might want to go back and edit your lists.

Sounds good.

I think this list is a great idea btw, just trying to help fine tune it.

i read you link and it does not speak about x4 as i was hoping to see. and they also attach words like "meaningful" "notable" "faster" for 1.5 fps differences. i fail to see how 1 fps in any situation is "meaningful" i'm really trying to understand this and would love to see some similar numbers from actual x4 tests and such to back the "will cripple a card" statement. i am trying to understand this and read as much as i can find but have yet to see any real world data to back the pcie beliefs. your puget article uses 2 titans at 4k resolution and shows no practical fps difference to suggest that the pci lane is messing with performance by bottlenecking the data flow even at x8 gen 2. a need for more bandwidth would surely show when cutting it by 1/4 going from x16 gen 3 to x8 gen 2 with a drop in fps or some obvious performance issue. hence, the titan is using less than 100% of an x8 gen 2 slot. if it is using more than 50% of the x8 gen 2 then another cut in bandwidth to x4 would show in dropped fps. it would have to be utilizing close to 100% of the x8 slot to all of a sudden go from 100% card ability to not usable and crippled? this lets me believe some small bottlenecking at x4 with a titan is possible (probable even) but is your mobo article going to suggest that someone wanting to sli 2 titans do so on an h87 platform to suffer such performance drops and not a newer 40 lane cpu that will give each titan the cpu's full attention (which it clearly does not need)? someone with an h87 using sli will be using much older and slower cards which the data shows me won't suffer any from the x4 limitation since 2 titans at 4k will barely feel it (again no actual proof but the common belief we all have).

this is from the 4th page of the myth article "Our final graph shows, for the first time, directly (versus indirectly via FPS measurements), how extraordinarily small PCIe bandwidth requirements are for a modern graphics card (assuming the readings are meaningful and not "unreliable" as Nvidia says). PCI Express 2.0 gave us 8GB/s of bi-directional throughput from 16-lane links. Both cards fail to hit 10% of that, requiring a sustained figure under 0.8GB/s. That is a mere 5% of what you have available to a single card on a Haswell-based platform."

this is 2 cards together (750 ti and similar to what someone might try to do with an h87 mobo). some quick math tells me that x4 gen 2 is half of x8 gen 2 bandwidth which is half of x16 gen 2. so 10% usage at x16 = 20% usage at x8 = 40% usage at x4 on gen 2 pcie slot. i'd like to see some real world numbers to go with this but i have never seen any actual proof that "the x4 PCI lane will cripple the second graphics card's performance" 40% usage of the available bandwidth is hardly crippled.

again if there is some real world data to refute what i am saying, please show it to me. this is a very common issue/question and it would be nice if everyone was on the same page.

what you say is true about cutting the % usage but the article shows that the 3.0 x16 slot is not 400% more than needed but 2000% more than needed (5% total usage observed with the 750 ti) ramp it up some for the high end card and assume maybe 10% usage and the slot is still giving 1000% more than needed. so your still at 125% of total need on gen 2 x4. again it does seem logical that the 4x would impact a super card by only offering (a guess based on data avaiable) maybe 90% of bandwidth need but anything less seems able to run all out unimpeded.

i will read the new link when i get home today. i understand real world can/will differ from theory and i hope to see the real world data to go with the lab data.

what i got from the puget article (real world data) is that the titan uses less than 100% of a x8 2.0 slot which means it uses less than 25% of a x16 3.0. how much less we don't know but data seems to show a lot less than 100%. so the 3.0 x16 slot is at least 400% more bandwidth than needed for a titan and most likely more than this since i doubt the titan is using 100% of the x8 2.0 slot.

can someone please run these tests again lowering to x4 so we can know for sure?? i'm getting a headache trying to differentiate fact from what everyone "just knows"

I see most the conversation is about gpu's on the H87 chipset but I was looking for information on capabilities of booting OS from PCIe x4 ssd on the H87. I have the ASUS H87M-PRO and in the bios there's the CSM option with PCIe boot device option in the sub menu but so far all I get at boot up is "The current BIOS setting do not fully support the boot device" . I've already tried every possible setting in there.What am I missing ?