angerthosenear

More Battlefield 4 testing with Mantle vs DX11 Mantle testing in a desktop. Did everything the same as with the eGPU. Test range. Same route (ish). Comparison of hardware: \[/TD] Laptop Desktop [/TR] CPU i7-2620M (2.7->3.4GHz HT dual) i7-4770K (3.5->3.9GHz HT quad) RAM 16GB @ 1333MHz 16GB @ 2400MHz GPU R9 280X @ x1 2.0 R9 280X @ x16 3.0 Drive HGST 1TB @ 7200RPM Samsung 840 EVO 120GB [/TABLE] Laptop performance here: [URL]http://forum.techinferno.com/diy-e-gpu-projects/2109-diy-egpu-experiences-%5Bversion-2-0%5D-54.html#post82601[/URL] Mantle DX11 Desktop charts: Doesn't look like much of a difference (there isn't really), but just like with laptop eGPU performance, there was about a 10fps gain. Went from high-70s/low-80s to high-80s/low-90s. I didn't say in my laptop post since I didn't pay much attention to the fps as far as mentally logging it. But there it was high-50s/low-60s to high-60s/low-70s. I must say, that is a pretty impressive showing for a laptop eGPU implementation. I'd expect a bigger difference in more graphically intensive locations however.

Actually, the fact you installed Win8 in MBR is very important because it is the only way you can chainload from Setup 1.x -> Win8. If you do a UEFI / GPT based install, you cannot chainload. We have about the same procedure, except I have to also load the DSDT table with Setup 1.x. Are you running Win8 or Win8.1 ? I like your very well thought out computer name! ;P

I did some Mantle tests with my eGPU setup (R9 280X). See results here: http://forum.techinferno.com/diy-e-gpu-projects/2109-diy-egpu-experiences-%5Bversion-2-0%5D-54.html#post82601 Quick result: Yes, Mantle is better and has a noticeable improvement. --- As said in other post, I will test in a desktop probably tomorrow to see how the performance difference is on both setups.

R9 280X with Catalyst 14.1. I have the onscreen infos showing that it is in fact using Mantle as the renderer. I am able to switch between both DX11.1 and Mantle. --- Results (sorta kinda): I can only use log parts of what is going on. Since Mantle is a new thing, I can't log fps and whatnot with an external program. Little chart action:: DirectX 11.1 Mantle The CPU usage has dropped some while using Mantle. And yes, it is noticeable. The framerate (unloggable) is upwards of 10+ fps higher than DX11.1. So yes! Mantle is a noticeable improvement from DX11.1 for those with AMD cards. Tomorrow (perhaps), I will test the same thing, but in a desktop. I will use the same GPU, but of course the CPU would be different. We are more concerned about the change in performance rather than exact numbers.

Whelp. My top notch quality internet is out. So a minor update for now. Any good external fps loggers and whatnot? Much better than me just mentally logging (and forgetting) the fps. Testing will be done in the "Test Range" for now. From the 30 seconds I was able to play, It seemed to have better frame rate with DX11.1 rather than Mantle. Of course 30 seconds of playtime isn't worth much. Hopefully I will regain Internet, but I doubt it. So assume testing will be done tomorrow, which also means I can test DX11.1 vs Mantle both as an eGPU and on a desktop (my last part comes in tomorrow). ---- SIDE NOTE::: You CAN disable the iGPU in Win8. I had to due this due to horrendous bugs in BF4. I will do testing on this in benchmarks to see if that helps with eGPU performance.

For those running AMD cards, I can confirm that the new 14.1 Beta driver does work. I plan to get BF4 soon-ish to test Mantle vs DX performance. I will report back on this eventually ;D

What model laptop do you have? This will be the most important factor in determining what is wrong. For the laptop I used here: http://forum.techinferno.com/diy-e-gpu-projects/5622-%5Bblog%5D-crossfire-testing.html It was pretty straight forward. However I could not use the internal display, external only. Are you using the proprietary AMD fglrx drivers? --- It is possible. And I have mined with my eGPU from within Windows, haven't tried within Ubuntu however.

I got the 14.1 drivers working with my eGPU setup with R9 280X. I might be interested in testing with BF4 as well. I can revert to 13.11/13.12 to test with and without Mantle to see if there is a significant performance gain. Would be interesting how much of a gain (if any) there is while running PCIe 2.0 1x

For this experiment, I'm using my boss's Dell Latitude E6420 that he has let me borrow in the past here: http://forum.techinferno.com/diy-e-gpu-projects/2109-diy-egpu-experiences-%5Bversion-2-0%5D-52.html#post80638 i7-2620M 8GB RAM I have Ubuntu 13.10 installed on my Samsung 840 Evo 120GB drive for this experiment (the mega-fast booting was great for the countless number of reboots I had ;D ) As you saw (if you checked it out), the mPCIe slot has a slight edge on performance for whatever reason, this also might be an indication of an lower # PCIe slot (or w/e you want to call it - earlier in PCIe chain). Only the card plugged into the mPCIe slot (vs ExpressCard) was able to be the primary card. I have my friend's R9 280X and another's 7970 for testing. The 7970 is hooked up via mPCIe to @jacobsson 's PE4L-PM060A (thanks for letting me borrow your adapter!), and the R9 280X hooked up via my PE4L-EC060A. NOTE: The ports are operating at a Gen1 link (PCIe 1.0), this certainly hurts performance. But since we are concerned about the performance change / if there is one, this isn't an overly pressing matter. Hardware pics here: Software As I stated just a moment ago, I'm using Ubuntu 13.10 for testing since this is the only way (for now hopefully) that it is possible to forcibly enable Crossfire on unsupported systems with something under the minimum 4x link per card. Laptops (that are concerned about eGPUs) don't have Crossfire support and certainly don't have 8x worth of spare PCIe lanes to give 4x link per card. After nearly pulling my hair out fiddling with driver installs (Ubuntu + me + driver installs = bad day), I got the correct fglrx drivers installed and actually working. I then followed the steps @sskillz used here: http://forum.techinferno.com/diy-e-gpu-projects/5622-%5Bblog%5D-crossfire-testing.html#post80974 NOTE: you must use 'aticonfig' using 'amdconfig' might get the Crossfire option to show up in Catalyst Control Center (CCC), but it will drop Cfire upon reboot. Using 'aticonfig' will (after much fiddling) persist between reboots. I had the same issue as @sskillz with Unigine Heaven crashing when running in windowed mode. So tests are run by selecting 'Basic' (to change to those presets), the selecting 'Custom' and then 'Full Screen' option so it runs fullscreen. Same for the 'Extreme' preset. Single card (R9 280X) performance: Here are some random terminal pictures and whatnot. This will also show that some commands will show Crossfire enabled while others do not. Also that CCC shows no Cfire while it may be enabled. Honestly, no idea what is exactly going on here - much hair pulling probably. Crossfire results (what you are actually curious about ;D ): As you can see, there is a significant enough gain in the basic preset to show that the Crossfire is working. The performance gain in the extreme preset is pretty lackluster. This might be due to the PCIe 1.0 based link on the cards (2.5GB/s instead of 5GB/s a Gen2 port would provide). I cannot get the ports to run at a Gen2 port since I cannot chainload without blackscreening and many issues from going to Setup 1.x -> Ubuntu. Ubuntu seemed to have installed itself in an odd fashion (becomes a boot device separate from the drive). ---- Crossfire, it can work via eGPU (in Ubuntu for now), but as for my opinion on the matter. You might as well instead of buying two cards+adapters, buy one very nice card and a single adapter. They performance (and setup) would be much better. So, all-in-all, it is doable. Certainly not worth it. -- If you are hardware manufacturer for eGPU components, contact me if you attempt to make a SLI/Crossfire unit. I would be very interested and glad to test (if possible).

I also was only able to run the cards at PCIe 1.0. I can't get Setup 1.x to chainload to Ubuntu. Ubuntu isn't using UEFI that I know of, it just refuses to chainload. Not sure what that's about. I'm sure PCIe 2.0 would've had a much more significant boost to performance.

Can't with my laptop, could with my boss's laptop, but not enough for dual x2 link (got 3 PCI-E slots, 4th slot I think is USB based). Would be neat to see however.

LAPTOP eGPU CROSSFIRE SUCCESS!!!!!!!!!!!!!!!!!!!!!!!! Pictures and full writeup coming soon (few hours from now). Note, the performance gain,,,, well, isn't really worth all the extra money for the setup. Got a 35% increase (Unigine Heaven Basic preset) from a single card setup -> Crossfire. Might as well just get a nicer card. Almost no increase for Unigine Heaven Extreme preset. I followed the same steps as @sskillz used on his desktop setup here: http://forum.techinferno.com/diy-e-gpu-projects/5622-%5Bblog%5D-crossfire-testing.html#post80974 Not sure about Windows Crossfire external. But the setup steps would be the same if @sskillz can figure it out. I sadly have to return all the hardware, so I cannot do any future tests.

Oh, your right. The original has 8ms -> 6ms response time. So this is a bit of an improvement. 'R' for Response Time I guess.

Been lurking in this thread for a bit. Newegg.com - LG 27EA83R-D Ocean Black 27" 5ms WQHD HDMI Widescreen LED Backlight LCD Monitor IPS pivot&height adjustable 350 cd/m2 5,000,000:1 27EA83R-D for $499 until Feb. 6. Pretty solid deal for those looking for this display and have been holding off. Still no idea what the 'R' means in the model number, I don't see any changes.

I'm running Win 8.1 64bit with no issue. I have to perform a DSDT override and PCI compation for it to work, but it does. I have a Fujitsu T901, I'm not sure about other units since it can vary from unit to unit - some are very finicky.

Other Test I reinstalled the driver, but the problem is still there... Please help! I'm running driver 327.02 and rarely get driver crashes (only when messing around with mining LTC). And no DirectX crashes. Perhaps reinstall DirectX and install driver 327.02. I know I have a lot of issues with certain driver versions.

SWEET!!!!!!!!! I'll attempt to meetup with my friend so I can borrow his card so I can attempt Cfire externally. I'll have a spare drive so I'll use an installed version of Ubuntu for testing. I guess if I can get that working properly, we just gotta figure out how to force enable Cfire in Windows. There doesn't so happen to be a commandline based setup similar to aticonfig in Windows is there?

I'd rather keep my brain securely located within my skull. You may have a whitelist issue. But if you are using a custom BIOS to remove the whitelist, then the issue is something else. Not exactly sure on how to fix. What laptop do you have? There might be someone else with the same model as you that got an eGPU setup working.

Quick test to see if there is any difference between the PE4L-EC060 and the PE4L-PM060 Essentially. ExpressCard vs mPCIe I7-2620M 2x 8,192 MB G.Skill DDR3 @ 667 MHz dGPU disabled AMD Radeon R9 280X Windows 8.1 Enterprise Fan mod (full fan speed) active to reduce throttling (due to poor Fujitsu design) Running 3DMark11 EC: AMD Radeon R9 280X video card benchmark result - Intel Core i7-2620M Processor,FUJITSU FJNB231 P7018 9474/3962/3929 mPCIe: (testing now after I post and pull laptop apart) //////// And of course. It's just not plug and play happiness. Fujitsu claims there is no whitelist. But computer won't boot with it plugged in. Not sure what's going on. Attempted on a Dell Studio 1737, got BSODs after driver installs. It didn't even work with hotplugging. Oh well. Now testing on a Dell Latitude E6420 (thanks boss for letting borrow your personal laptop for testing!): i7-2620M 8GB RAM R9 280X Windows 8 EC: http://www.3dmark.com/3dm11/7842335 P6444 8121/3981/3977 EC (Re-ran with back panel off): http://www.3dmark.com/3dm11/7842928 P6452 8147/3972/3977 mPCIe (note, back panel off, not sure if there would be a difference due to thermal / airflow): http://www.3dmark.com/3dm11/7842830 P7026 9468/3967/3954 Side note: the Dell Latitude E6420 is a very nice laptop. Has two true mPCIe slots and 1 USB-based mPCIe slot. Easily upgradable. Single HDD bay however, no mSATA that I see. Nice fan. Thin and light, yet solid. *hides in bag* Not sure if an issue with the first test or something, but the mPCIe test is significantly better. But yet only matches the EC test on my laptop. So perhaps... "It depends" would be the best we can say. Conclusion: It more than likely doesn't make a difference what interface is used. Just variables when testing gave some varying results. Not really a surprise here since both are PCIe 1x connections.

SATA cables are used for data transfer already, I see no reason why they cannot be used for PCIe based signals as well. We know from SATA3 they can support 6GB/s and with PCIe Gen2 @ 1x only using 5GB/s, this is more than adequate. Not really a step back if it works while being socketed. SATA 3.2 seems like something to combat M.2 which is (can be) PCIe 4x based. So.... hm. Guess we'll see how that plays out.

Wow. Uhm... How about using a SATA cable? Shielded lanes. We know it can support 6 GB/s soooo that should be good enough. I think that is suitable for a socketed system. Can't believe we all forgot about it.

@bjorm I get a GPU usage bump whenever anything changes. So when I refresh the page or other graphical changes (page loading or loading), there is some GPU usage. If it is on internal only then that is a different story. If I monitor GPU usage / GPU memory usage. I can have over 500MB GPU mem usage while NOT gaming. Just with a bunch of stuff on my screen at once (plus 4 monitors hooked up to GPU).

This seems rather promising: AORUS Gaming Laptop X7 Brief: Very thin and light system featuring a Haswell i7 and SLI GTX 765M. 0.9in. and 6.39lbs Thermal design looks very nice with it's dual fan and 5 heatpipe design. Also has 4 RAM slots and 2x mSATA + 2.5" HDD bays for solid memory and storage options. Can output to 3 displays and allows for Nvidia Vision Surround. --- Only complaint being the fact it uses the older mSATA instead of the newer/faster PCIe base M.2. Don't know the price on it, but I'll be watching it.

I still have to wait for the the snails working at the post office / customs for the adapter to arrive. I haven't tried Ubuntu with the single card yet. Might not need the DSDT override. I'll test and report back. Hopefully I got the persistent USB thinger working. I've never been able to get it to work Couldn't get Ubuntu to work live at all.. --- What about something like this: https://www.xgamingstudio.com/forum/showthread.php?10-RELEASE-ATI-CrossFire-Xtension-A-way-to-force-crossfire-w-profiles

Internal speakers. I'll try a different VGA cable. The one I'm using at the moment does have a ferrite about 2" away from the plug on both ends. Might just be a shoddy cable regardless. It is rather old.