Topic: good card, crappy board = what performance? (precise specs included) (Read 813 times)

My motherboard is one of those weird ones that has two x16 PCIe slots, but one operates at x4 speed if it is in crossfire mode, yet both cards still perform the same hashrates.

You can even buy PCIe 1x splitters that turn a single x1 slot into two or four x1 slots.  Just notch the connector on those split-off x1 slots, toss some x16 cards into them and they will perform just the same as if they were in the x16 slots.

ohhhhhh I didn't even think about total bandwidth from the results of the hash calculations compared to pixel related data that makes perfect sense!

*digital hug*

Sweet, I'm totally ordering it.

Not likely going to be an issue for bitcoin mining. The amount of bandwidth used is very small compared to a game so even an old crappy PCIe x1 slot would do fine. I believe that's why people can use extender/splitters to connect like 16 cards to one mainboard. A game also needs a lot more "smarts" hence the CPU is a bottleneck there.

oh and one single review on newegg of this card actually helped.  It stated he's using a single card and it's an AMD Athlon II X4 system (dunno if it's AM2+ or AM3) and after overclocking it, he got it to 260MH/s, which is significantly lower than the person who tested it on an i7 system.

That's not a bad idea but since these are screwey one-off OEM parts (welcome to e-machines!) there probably won't be many drivers available for the rest of it.  Not that I'd need sound but I need networking.  But since I just got done reloadeding Vista onto it, I could just terminate explorer.exe while the miner is running.  Either way, the screen saver and Aero are getting turned off, and about 100 other things that aren't relevant to a mining system.

The main concern I had about a non-GPU bottleneck between the system seen in the benchmark and the system I have is the PCI-E controller.  I had heard a while ago that intel or AMD or someone was planning to put the PCI-E controller on the CPU as well as the CPU cores and the GPU.  Did they actually get around to doing that in the sandy bridge based i series procs?  Cuz I have a feeling that would make a biiiiiiiiiiiig difference.

The bottleneck is going to be in the graphics card drivers and the GPU chip itself, mining is taking like 1% cpu on my i7 right now.

It'd probably be worth your while to install ubuntu on it and NOT load a GUI, that way all the GPU will be dedicated to mining.

I had this Nvidia GTS450 in my old socket 939 based board and the framerate in most games wasn't real good.  Now I built this i5-2400 socket 1155 based system to replace it and moved the card over to it and tada, 60FPS in every game with even higher settings than in my old system.  So apparently the board affects the GPU performance to some extent.

Now, who knows how badly the original x16 slot (not 2.0 or 2.1) and DDR1 and an IDE hard drive and a relatively limited processor by today's standards affected game performance though.

So that's great but I want to get a Radeon HD 5830 for obvious reasons.  It can't go in this i5 computer though cuz I'm gaming on it all the time and it makes my room really hot. The only spare system I have that would be reasonably fast is a piece of crap emachines I got for $110 at the pawn shop with a 17" LCD monitor as a mandatory part of the package.  I actually just needed the monitor for a customer so I bought the whole package and sold the monitor for $70. So what's in this (now) $40 computer you ask?  An Intel E210882 board (that's printed on the board but CPU-Z says it's model D945GCL) and an Intel socket 775 3.0GHz Pentium D with 4MB L2 cache and 800MHz fsb, i945 chipset, 4 GB of DDR2-6400, and a copy of Vista 32 bit  So it's not bad but it's bad enough I don't want to sell it to a customer (I sell and repair comps btw).

So that radeon card under perfect conditions in a killer i7 machine is supposed to get 280-300MH/s.  How low do you guys think that lovely Vista emachines system described above would drop that rate if I put the HD 5830 in it?  I assume it's somewhere between like 5% and holy crap so hopefully someone's tried something similar and has some ballpark numbers