|
wolrah posted:I'm a block and a half from corn fields in two different directions and I have 500mbit/sec cable service. The next town over has gigabit muni fiber. In between us, however, are a lot of unfortunate souls who have the choice of LTE based services or Frontier DSL that advertises 24/3 and might do 10/1 if you're lucky. Nebraska? I've got gigabit u/d here and my in laws 30 min away have 1mbit satellite as their only option.
|
#
?
Apr 18, 2019 00:43
|
|
|
|
| # ? May 31, 2024 07:18 |
|
|
EdEddnEddy posted:The question I have to ask, have you actually tried VR, And not just your Phone or a Rollercoaster Experience VR, but actually strapped on a Vive in a proper room scale setup and Played any of the excellent experiences that are available currently? (Or sitting in a game such as Elite Dangerous or Project Cars? Yes a flight sim and driving sim which VR was good for even before the Roomscale stuff came about.) Id buy VR just for something like Elite but its a steep price of entry for just one game. The Rift S being $400 is making me consider buying one though.
|
|
#
?
Apr 18, 2019 01:02
|
|
|
Cao Ni Ma posted:Id buy VR just for something like Elite but its a steep price of entry for just one game. The Rift S being $400 is making me consider buying one though. Hit up the VR thread in Games, but for simulators one of the cheaper WMR headsets can be a great option.
|
|
#
?
Apr 18, 2019 01:05
|
|
|
I mean, PC gaming might be dying but no more than desktop PCs in general are dying, in favor of tablets and mobiles
|
|
#
?
Apr 18, 2019 01:50
|
|
|
Comfy Fleece Sweater posted:I mean, PC gaming might be dying but no more than desktop PCs in general are dying, in favor of tablets and mobiles Tablet sales are also declining and smartphone sales are leveling off. I'm afraid the entire concept of the electronic computing is on death's door. We're going to have to go back to rooms full of classy ladies doing math soon 
|
|
#
?
Apr 18, 2019 02:03
|
|
|
Enos Cabell posted:Nebraska? I've got gigabit u/d here and my in laws 30 min away have 1mbit satellite as their only option. Ohio. I used to live in one of those kinds of areas, about 20 miles from where I am now. Verizon actually accidentally sold us DSL out there when we were officially out of range, it'd work at sub-megabit speeds most of the time but got pretty lossy at times and would totally drop out on hot summer days because the above ground lines apparently stretched out and we were so marginal it'd just stop working. We kept it because our other choices were 26.4kbit/sec dialup or satellite. Since that experience I have had ISP availability as a primary criteria when looking at places to live.
|
|
#
?
Apr 18, 2019 04:56
|
|
|
Comfy Fleece Sweater posted:I mean, PC gaming might be dying but no more than desktop PCs in general are dying, in favor of tablets and mobiles The percentage of gaming PC vs. home desktops productivity and gen usage show that gaming PC are gaining. https://www.youtube.com/watch?v=qioXYSED-MM (if you can't speak Spanish, you know what this clip is anyways)
|
|
#
?
Apr 18, 2019 06:10
|
|
|
I live in a town of 80,000 in the middle of the country and have fiber. It's also run by the city and costs $50/month. Before that I also had a fiber connection in the middle of flyover country. You can get good internet away from the coast but I know that it's not even remotely common and I've been very fortunate to have that kind of connection for almost a decade now.
|
|
#
?
Apr 18, 2019 06:57
|
|
|
Happy_Misanthrope posted:I thought Nvidia's multi-display GPU clock issue was fixed? My clocks go from 300 to 1200 mhz when I extend my display to my TV. Both are '60'hz, the only difference is my TV is likely at 59.94 vs the true 60hz of my monitor. Both are 4k. Try multi display powersaver that comes with NVinspector, it's the only tool that has reliably worked for me
|
|
#
?
Apr 18, 2019 08:36
|
|
|
I haven't seen this posted yet, behold Nvidias fix for the big RTX cards burning up inside:  Yes this is how they come from the factory now. (German) https://www.tomshw.de/2019/04/17/ka...ueck-der-woche/
|
|
#
?
Apr 18, 2019 10:47
|
|
|
Bazooka Joe to the rescue! (Bazooka Joe is bubble gum)
|
|
#
?
Apr 18, 2019 11:34
|
|
|
That is a RTX Titan in the picture. Nothing says 2500$ like five different thermal interface compunds on one cooler. The EVGA 2080ti kingpin has a much more elegant copper heatspreader, it’s briefly visible in a GN video.
|
|
#
?
Apr 18, 2019 13:57
|
|
|
eames posted:That is a RTX Titan in the picture. Nothing says 2500$ like five different thermal interface compunds on one cooler. On a $300 card, yeah, okay I get that. Kind of. On the fuckin Titan? lol gently caress off with that poo poo E: I mean for $2500 it should have poo poo I haven’t even heard of for cooling
|
|
#
?
Apr 18, 2019 14:12
|
|
|
Rexxed posted:Bazooka Joe to the rescue! Yeah no joke that looks like the underside of an elementary school desk
|
|
#
?
Apr 18, 2019 15:41
|
|
|
tehinternet posted:On a $300 card, yeah, okay I get that. Kind of. The one thing you can always be sure about companies is that they love saving money more than you do
|
|
#
?
Apr 18, 2019 15:45
|
|
|
Does anybody understand why they went with this solid cooling plate covering the entire front of the PCB rather than letting the airflow from the fans hit the components and PCB? Surely was more effective than the contraption pictured above? Even if this thermal pad workaround solves the "sudden death" issue there's little doubt in my mind that these cards will have a shorter lifetime than they could have with better cooling. I found a picture of EVGA's memory heatspreader, apparently they already used it for their watercooled Pascal cards.
|
|
#
?
Apr 18, 2019 16:32
|
|
|
sauer kraut posted:I haven't seen this posted yet, behold Nvidias fix for the big RTX cards burning up inside:   So glad I cancelled by 2080ti preorder and got a regular 2080 instead.
|
|
#
?
Apr 18, 2019 16:35
|
|
|
Stanley Pain posted:
Just never buy first party cards.
|
|
#
?
Apr 18, 2019 16:43
|
|
|
Thermal pad spam is what all of the high end cards do, and have done for a few generations if you watch the GN teardowns. That seems a bit over the top, but that cooler is over the top in a whole lot of ways (like screw count/type). Yall complain about weird stuff.
|
|
#
?
Apr 18, 2019 16:53
|
|
|
It's a very reasonable way to dissipate heat from a bunch of different surfaces that don't need great thermal assistance, but definitely need some.
|
|
#
?
Apr 18, 2019 18:04
|
|
|
So how big of a deal is pcie 4.0? Is the throughput increase noticeable everywhere? And now 5.0 is coming soon after..
|
|
#
?
Apr 19, 2019 01:33
|
|
|
Purgatory Glory posted:So how big of a deal is pcie 4.0? Is the throughput increase noticeable everywhere? And now 5.0 is coming soon after.. Only extreme edge cases (some 2080ti) saturate 16x PCI-E 3.0 lanes, and only then by a tiny amount, so for GPUs it's not really any news of interest.
|
|
#
?
Apr 19, 2019 01:34
|
|
|
VelociBacon posted:Only extreme edge cases (some 2080ti) saturate 16x PCI-E 3.0 lanes, and only then by a tiny amount, so for GPUs it's not really any news of interest. unless your cpu doesnt have many pcie lanes and wants to use them for other things, then 8x lanes pcie4 would be sufficient
|
|
#
?
Apr 19, 2019 02:02
|
|
|
What situations cause the need for PCIe lanes? I hear about it a lot but I don’t really see it mentioned outside these forums. Is it less a PC gaming thing and more some other thing?
|
|
#
?
Apr 19, 2019 02:04
|
|
|
buglord posted:What situations cause the need for PCIe lanes? I hear about it a lot but I don’t really see it mentioned outside these forums. Is it less a PC gaming thing and more some other thing? NVME uses 4, thunderbolt can use 4, gigabit ethernet uses 4. And your chipset (not cpu lanes) can only assign lanes in groups of 4, so anything that takes 1 lane is going to pull 4. Yeah it's not a huge deal but the 9900k only has 16 lanes to begin with. Chipsets provide another ~24 or so
|
|
#
?
Apr 19, 2019 02:31
|
|
|
buglord posted:What situations cause the need for PCIe lanes? I hear about it a lot but I don’t really see it mentioned outside these forums. Is it less a PC gaming thing and more some other thing? Bandwidth needs. Or if you want to think about it another way, communication needs. A little wifi card or sound card doesn't need a ton of bandwidth, cause the data its working on is relatively small and doesnt need to be passed around to a bunch of other locations in the system. On the other side of the spectrum, something like a video card (or the interconnect between two CPUs in a server) could be moving huge amounts of data between places like system RAM, compute units, etc
|
|
#
?
Apr 19, 2019 02:40
|
|
|
VelociBacon posted:Only extreme edge cases (some 2080ti) saturate 16x PCI-E 3.0 lanes, and only then by a tiny amount, so for GPUs it's not really any news of interest. 2080Tis saturate PCIe 2.0 lanes on occasion. It's going to be quite some time before anything saturates 16 3.0 lanes, and even the first crop of PCIe 4.0 storage solutions are adding about 500-750MB/sec on top of the current top-tier NVMe drives. In short, PCIe 4.0 might be nice if there's an easy way to control lane allocation to enable more/better use of add-in cards. CF/SLI/NVLink might not be worth it, but for ML applications and the like, being able to theoretically run four GPGPUs on a *consumer* board at PCIe 4.0 4x4 (which would be the equivalent to 4x PCIe 2.0 x8) would help a lot of hobbyists and students not have to spend a small fortune on a super-expensive PC.
|
|
#
?
Apr 19, 2019 03:33
|
|
|
Oh sorry. I thought the GN test was on 3.0?
|
|
#
?
Apr 19, 2019 04:11
|
|
|
VelociBacon posted:Oh sorry. I thought the GN test was on 3.0? It might have been, but I remember the results showing that at the absolute limit, that the 2080Ti *just* barely would completely saturate and slightly exceed the maximum bandwidth of a PCIe 2.0 x16 connection.
|
|
#
?
Apr 19, 2019 04:48
|
|
|
BIG HEADLINE posted:It might have been, but I remember the results showing that at the absolute limit, that the 2080Ti *just* barely would completely saturate and slightly exceed the maximum bandwidth of a PCIe 2.0 x16 connection. Only took ~12 years to saturate the PCIe 2.0 on the X38 chipset
|
|
#
?
Apr 19, 2019 05:00
|
|
|
Re PCIe 4, it isn’t only the throughout but also the faster and wider the link is, the lower the latency on the link as well. This is more important for machine learning gpu applications than graphics in general, but the link also responds quicker to turnaround data transfers.
|
|
#
?
Apr 19, 2019 15:51
|
|
|
Looking at the description of PCIe 5.0, it says 16x will run at 63 GB/s. Considering dual channel 3200 DDR4 only has 51.2 GB/s of bandwidth, it sounds like the kind of thing that won't feature in the consumer space for years even though the spec is supposed to be finalised shortly. Won't the motherboards cost a fortune too given that the traces for PCIe 4.0 are supposed to be pretty difficult to create already?
|
|
#
?
Apr 19, 2019 17:12
|
|
|
We'll have DDR5 before PCIe 5.0, and Tom's Hardware just put out caps from a DDR5 presentation yesterday: https://www.tomshardware.com/news/what-we-know-ddr5-ram,39079.html
|
|
#
?
Apr 19, 2019 17:18
|
|
|
oh loving brother, that one guy has a loving Navi video out.ConanTheLibrarian posted:Looking at the description of PCIe 5.0, it says 16x will run at 63 GB/s. Considering dual channel 3200 DDR4 only has 51.2 GB/s of bandwidth, it sounds like the kind of thing that won't feature in the consumer space for years even though the spec is supposed to be finalised shortly. Won't the motherboards cost a fortune too given that the traces for PCIe 4.0 are supposed to be pretty difficult to create already? No, because PCIe 4.0 is electrically-identical to PCIe 3.0. The issue is signalling, not physical traces. It's why there is the prospect, with Zen 2, that PCIe 4.0 could come to existing motherboards in a limited capacity, that is, to the first slot only, and then everything else would remain PCIe 3.0. To get PCIe 4.0 out to all the slots, you would need booster chips, which, obviously, won't be retrofitted to older boards, but will probably be a thing on newer boards. SwissArmyDruid fucked around with this message at 21:32 on Apr 19, 2019 |
|
#
?
Apr 19, 2019 21:27
|
|
|
What does a booster do / how does it do it Is it basically just running a second bidirectional garden hose from the well?
|
|
#
?
Apr 19, 2019 22:36
|
|
|
Statutory Ape posted:What does a booster do / how does it do it a repeater basically
|
|
#
?
Apr 19, 2019 23:07
|
|
|
PCIe repeaters fall into two categories: redrivers and retimers. Redrivers are electrical only to boost the signal and reduces the random jitter. They are just signal conditioners. Pcie equalization will be performed on the entire link, through the redriver. Retimers are aware of the pcie link equalization procedure so each link connected to a retimer goes through equalization. They convert the high speed analog to digital to extract the equalization training sets and then back to high speed analog. So they’re quite a bit more complex. They shouldn’t show up in the pcie device tree. However another method would be to use 2 port pcie switches to retime links as well, which would show up in the tree and can add some additional latency depending on how the switch works. IDT has a good explainer if you’re into that kind of thing: https://www.idt.com/document/msc/pcie-gen3-retimer-frequently-asked-questions
|
|
#
?
Apr 20, 2019 03:27
|
|
|
A few pages back ( I think it was this thread) we talked about going to nv control panel and putting fast sync on. Seems to work great for me on a gtx1060 hooked into a TCL tv Tried the same thing also on a 1060 on my laptop and the option wasn't there for the laptop screen Is it just not available or? It's a Helios 300 with the 7th gen i7/15.6/1060
|
|
#
?
Apr 20, 2019 15:47
|
|
|
Statutory Ape posted:A few pages back ( I think it was this thread) we talked about going to nv control panel and putting fast sync on. Seems to work great for me on a gtx1060 hooked into a TCL tv If your laptop uses Optimus, you're might not able to use it, because all display outputs usually go through the Intel integrated graphics. It's the same reason you can't drive gsync external display on most Optimus laptops. You need the whole chain to support the feature, or you need a laptop with an dedicated discrete card output. defaultluser fucked around with this message at 16:46 on Apr 20, 2019 |
|
#
?
Apr 20, 2019 16:35
|
|
|
|
| # ? May 31, 2024 07:18 |
|
|
HalloKitty posted:Interesting - gpu usage is higher and cpu lower with vulkan as one might expect, but there's clearly optimisation still needed stop thinking of low CPU utilization as being a good thing, think of it as "low CPU occupancy" instead. Your cores can't get enough work to keep them occupied. The reason the CPU utilization is lower on Vulkan is probably because the dev sucks at optimizing drawcall submission and NVIDIA's DX11 multithreading driver is clawing back some performance by spreading it across multiple threads. And you've got a single-thread bottleneck in Vulkan as a result, which is why the framerate is lower. Low CPU utilization and low framerate is actually a bad thing, high utilization and high framerate is a good thing (as long as utilization is not so high that you are getting stutter). Otherwise you've just got a bunch of cores (and your GPU) sitting there idle while the main thread tries to churn through fast enough to keep them busy. Single-thread bottlenecks are a practical reality in many titles even in multithreaded engines. (low CPU utilization and high framerate is great of course! but given the choice I'll take high framerate over high percent of my processor sitting idle any day.) Paul MaudDib fucked around with this message at 19:18 on Apr 22, 2019 |
|
#
?
Apr 22, 2019 03:13
|
|
