PC Gaming Hardware

[The Soulman]

A noobs question: is it possible to run a (bottom end) graphics card on a single PCIe lane?
If it is, how greatly is the performance degraded?

Of course it's possible to run any PCIe device (afaik) on a single lane. The performance penalty depends on what kind of PCIe (PCIe 3.0 1x is a bit faster than 1.0 3x) what graphics card, and what application you intend to use it for.

If you just intend to do stuff like watching YouTube videos at 1080p60, then I have no doubt that a modern 2.0 or 3.0 lane will be able to power a cheaper graphics card just fine for those lighter tasks. The main issue occur when you try to do stuff like 3D acceleration. Not that you would put any powerful card on that sort of connection, but most games would probably choke and die at those speeds.

Thanks, for instance Intel mobile express chipset GM45 (p9400, 8Gb ddr3) and geforce 1030 thru a mini PCIe adapter such as:
https://www.aliexpress.com/item/Mini-PCI-E-PCI-Express-Extender-Riser-Card-1x-To-16x-PCIE-Mining-Card-USB-3/32786970481.html?spm=2114.search0104.3.49.e5a83378e99q2d&ws_ab_test=searchweb0_0,searchweb201602_3_10152_10065_10151_10344_10068_5722815_10342_10343_10340_5722915_10341_10698_5722615_10697_10696_10084_10083_10618_10304_10307_10301_5722715_5711215_10059_10534_308_100031_10103_441_10624_10623_10622_5711315_5722515_10621_10620-10152_10151,searchweb201603_68,ppcSwitch_5&algo_expid=dcd9ff0a-e36e-45c4-bf31-65b70c6ba45d-7&algo_pvid=dcd9ff0a-e36e-45c4-bf31-65b70c6ba45d&priceBeautifyAB=0

Intended use is to accelerate video editing. 

[Ampera]

In theory, yes. It will work, but I'm not sure how well for that application. Skimming will likely not be as fast, but straight up rendering might be possible.

In practice I would never use something like this. If this is truly your only option (keep in mind USB 3.0 will likely be faster or par) then I don't believe it's something worth doing. That device is intended for mining rigs where all you need to do is throw a load of small sized hashes onto the VRAM and let it rip.

[paulca]

So I did a bit of reading on the CMT 'cores' of the AMD FX/Bulldozer architecture.

The trouble is that 99% of the articles explaining it are from gamers.  Who repeatedly focus back onto single thread performance due to DX11 and previous single threaded CPU bounding.

They may not be 8 full cores, but they have twice the integer execution units compared to Intel quad cores.  Intel SMT is just time switching within a core, where as AMD CMT is actual double the execution units per core.

Now the Intel gamer fan boys immediately skip past this as being double the execution hardware for multiprocessing and focus on it being effectively less hardware per core count for single threading.  As a single thread can only use one execution unit.  Which is a bit short sighted and single thread focused.

Of course in terms of gaming a single thread can only use one execution unit, which is the same in the Intel architecture.  However AMD wasted so much resource creating the architecture they got left behind in actual single execution unit performance.  That and poor implementation of the floating point unit bottle neck by having two integer units and only one floating point unit per core.

Still, in flat out multi-processing the CMT will out perform the Intel per clock cycle per core.  But gamer Intel fan boys skip past that as "Not important for gaming" and of course 90% of their benchmarks are gaming focused.

But at the end of the day what does it mean.  Well, heavy multi processing is about the only place the FX will out perform per clock speed.  It's energy hungry, larger die, higher operating temps are a factor for cooling, but again the reviews are biased by quoting problems throttling with the stock cooler (duh!). 

Gaming, no.  Compiling, heavy desktop use, multiprocessing applications, video rendering, ray tracing, etc. etc. the FX has a slight edge.

However DX12 should make most of that moot, but it's moot anyway as the same CPU architecture for the DX12 generation would be the Ryzen not the FX/Bulldozer.

If I listed the things that make me wait on the CPU performance by how much time I spend doing them, then gaming would not be top of the list.  So I don't think I made "that" much of a mistake with it.  When I bought it was heavily into compiling Linux OSes using automated builder script that took a day or more to run on a Phenom quad.  These days even large components like Firefox build in a few dozen minutes on the FX. 

Anyway, I expect the Ryzen still will beat it hands down.

[wraper]

In gaming terms probably, but aren't the FX and the Ryzen about 2 years apart?

Although FX sort of has 8 things which can be called cores, they are not really full fledged cores. It's more like 4 cores with hyper-threading on steroids. In terms of technology they are more like 7 years apart, not 2. Unless you run heavily mutithreaded task, even 4 Cores, 4 Threads Ryzen 3 1200 beats it. If you take 3.2GHz Ryzen 5 1400 which has SMT (4 Cores, 8 Threads), it beats 9590 in any scenario.

Isn't that the argument that was leveled against the original Intel dual and quad cores?  That they weren't really cores just threads?  Where as AMD actually had proper cores.  Maybe that changed with the FX I donno.

I think you are confusing it with two separate dies in Intel CPU while AMD had single die. I don't recall any confusion between Hyper-threading and actual CPU cores. FWIW both Intel hyper-threading and AMD SMT is much more than just splitting core for 2 tasks, it significantly increases performance in multithreaded applications compared with just maxing out core with single task.

[NiHaoMike]

A noobs question: is it possible to run a (bottom end) graphics card on a single PCIe lane?
If it is, how greatly is the performance degraded?

Of course it's possible to run any PCIe device (afaik) on a single lane. The performance penalty depends on what kind of PCIe (PCIe 3.0 1x is a bit faster than 1.0 3x) what graphics card, and what application you intend to use it for.

If you just intend to do stuff like watching YouTube videos at 1080p60, then I have no doubt that a modern 2.0 or 3.0 lane will be able to power a cheaper graphics card just fine for those lighter tasks. The main issue occur when you try to do stuff like 3D acceleration. Not that you would put any powerful card on that sort of connection, but most games would probably choke and die at those speeds.

If there's enough VRAM, games run just fine. That's my experience with a DIY adapter to connect a standard GPU to an Expresscard slot.

[tszaboo]

I already have an Acer Predator 21:9 1440p monitor.  However it's only 100Hz and only when running on display port.  I am running it at 60Hz on HDMI currently.

I'm not a FPS gamer, I do like smooth performance, so G-Sync or V-Sync are on, so I'm maxed out at 60FPS anyway.

As an ad-hoc bench mark Rise of the Tomb Raider pretty much runs at 60FPS until I turn some of the higher end anti-alising/over sampling or realistic hair on.

My gaming stutters are entirely caused by IO on the hard-disk, usually unrelated processes in Windows poking around at things causing stutters.  Things such as AntiVirus which if I haven't been in Windows for more than a week seems to wait till I launch a game and then kick off scans and poking around.  Thankfully with 16Gb of RAM actual game HD access is minimal beyond loading screens.

So basically, my FX 9590 might be a bit of a gaming slouch, but it's probably not worth upgrading to a Ryzen 7 just yet and with top end GPUs costing well over £300 it's probably not economical to upgrade there either yet.

That's a very good start. I didnt mention Ultrawide, because that is a trigger word for some people, and they start rambling. It is good for gaming. Just please use displayport, and try G-Sync. It is that noticeable. Those stutters what you are experiencing will be mostly gone.

[paulca]

That's a very good start. I didnt mention Ultrawide, because that is a trigger word for some people, and they start rambling. It is good for gaming. Just please use displayport, and try G-Sync. It is that noticeable. Those stutters what you are experiencing will be mostly gone.

The trouble I had with display port is that the video card and monitor fight each other.  On boot up the monitor cycles it's inputs looking for active signals, but the video card only enables the display port when it sees a client.  The video card doesn't seem to enable the display port fast enough and the monitor moves on to the next input.

I tried different things, but the only thing that worked was rebooted two or three times to the POST screen until finally they seen each other. 

Then of course I would leave the computer for 20 minutes and the monitor would go to standby and I'd have to start over.  In Linux this was fine as I could force a static monitor assignment, but in windows at times I couldn't get the desktop back at all.

So I reverted to HDMI.

[paulca]

Anyway.  I got a bonus in work so I have money burning a hole in my pocket, although I could put it towards my house savings, but I'm getting tired of saving.

The spec currently looks like:

Ryzen7 2700X, 2x8Gb 3666Mhz DDR4, Asus Crosshair VII X470 Mobo, NVidia 1070Ti, Samsung Polaris 256Gb M.2 drive, Corsair 750W PSU, Be Quiet or Crystal Design case.

Normally I wait until I can get at least double performance across the board, but this would give me 50-60% increase in most areas.

The alternative would be building a NAS for myself, aiming to run OpenNAS or FreeNAS, but it's tricky picking a cheap, small, silent, low power setup with enough oomph for running a media server but low enough power consumption when idle to leave on 24/7.

Of course if I upgrade the gaming rig, I can put a rubbish video card into the old one, turn everything down and use it as a NAS, at least temporarily, though I doubt I'll get the power consumption below 50W.

[Doctorandus_P]

Did you realize it's over 2 years old?

[Ice-Tea]

Just a spammer biding his time.