Comments by "" (@Wingnut353) on "Gamers Nexus" channel.

  1. Yep, its a good timing for cyber monday also people can watch think it over a bit and then snag thier deals with time to spare.
    1
  2. That isn't what was said at all... games that use purely DXR won't require any changes. Games that dip into Nvidia proprietary APIs will require patches for AMD. The jury is still out on Vulkan-RT I can't imagine AMD not supporting that at all though.
    1
  3. In general it is slower with the current drivers. I'd take level1linux with a grain of salt for now, give it another month or so for things to shake out on the open source drivers as they not yet stable and like he said FPS numbers don't meant a lot on unstable drivers as they can be broken in areas that speed up rendering but cause it to render incorrectly... I think the 1080ti is probably alot slower on Linux vs windows as well whereas the Vega card probably runs similarly on Windows and Linux. Nvidia tends to integrate with Linux graphics stack poorly as it is closed source... so things like slow 2d dragging of windows (like slide show) isn't unheard of even on a 1080ti.
    1
  4. Steve droppin truth bombs...
    1
  5. To the L2 vs L3 question... L2 functions really to cache things a bit quicker that you don't want to be in L3 closer to a single core or pair of threads etc.... L3 caches things for the whole CPU complex. L3 is always the most beneficial place to increase cache as all cores can benefit larger L3 means better multi-threading performance generally... while L2 increases aren't needed as much because they mainly are there to take up slack for the L3 and the performance characteristics between L2/L3 haven't changed too much, say for instance your L3 on the new process had to be 2x slower, well then you would need to reevaluate this relationship and build a larger L2 to compensate for the slower L3... if L3 gets faster you could even make the L2 a bit smaller! Think of the Main memory like a Sea, flowing into a lake and then into a water tower and then into your glass (the CPU) Each smaller reservoir only needs to be large enough to not get drained, if you have a memory access miss prediction there isn't a lot you can do, but if the prediction is working you can keep what you will need in the cache and keep it flowing to the CPU and all of the caches working optimally. The water towers aren't a perfect analogy since data flows in both directions through the memory hierarchy but it gets the idea across I think. For instance if something doesn't fit in cache it must be spilled back into main memory and flow back upstream.
    1
  6.  @nicholasduncan7424  "Not to mention PCs can already do 4K at 144 fps. It's just expensive to do." EXACTLY... you wont' be able to do the same performance as consoles 2 years from now at the same price... you'll be some significant amount above that.... that's just the way it is. You'll have a CPU 2x as fast, but games won't be faster on it, you'll have a graphics card that is 2x as fast but it only lets you go from high (console equivalent) to ultra (with little to no visible differences). I'm not a fanboy I have several PCs and both a PS3 and PS4, this is historically how it has gone... I'm just being pragmatic.
    1
  7.  @antc1442  buildzoid is the last hairbender afaik...
    1
  8. https://youtu.be/OUtvsAmD3Ws?t=1222 Edit: Engineering sample is T Retail is Daisychain just like he says in the review here https://i.imgur.com/v6LLuTf.png that's a photo of his retail board.
    1
  9. Actually no... a cheese grater with dual Vega Duo IIs is still only $17000.... with no other options except the four GPUs on 2 cards. And that would be double the FP32 Teraflops, and 25% higher INT8 deep learing OPS as well as 100GB/s higher bandwidth per GPU, and similar or better inter GPU link bandwidth.
    1
  10. Higher conductivity dies is a major breakthrough.... and mandatory if we are to stack high frequency dies. Imagine a Threadripper in a standard AM4 package with the dies stack on top of each other similar to HBM (HBM can get away with stacking currently due to low clocks, or in the case of Aquabolt large numbers of heat conductive TSVs). Hopefully they can actually develop some complex designs with it... that'll be the kicker. Also worth noting is this https://www.nrl.navy.mil/news/releases/nrl-researchers-discover-novel-material-cooling-electronic-devices but that is cubic boron arsenide probably harder to make than bulk which is what is being reported on now.
    1
  11.  @NovemberOrWhatever  No....what I'm saying is what you are trying to do is completely impossible. If you want to compare CPUs like you want the way to do that is normalize against clock rate not actually change it....
    1
  12. It's probably just a quad rank dimm... and the BIOS is just seeing it as 4 dual rank sticks. Since consumer PC BIOS / UEFI only ever really plans on seeing single and dual rank dimms.... maybe it just doesn't even try to intialize the rest of the ranks on the second pair of dimms since the memory controller has allocated it's max number of ranks already (8). Ranks on DDR4 are basically the same as putting another set of sticks on a single card with a select line between them. In the past there have even been 8 rank DIMMs.
    1
  13. So just to reiterate... alot of people claiming improvements were on FCU + the Crimson Beta driver for FCU update... it isn't surprising that an Nvidia card isn't seeing improvements especially one that has been out so long.
    1
  14. I mean AMD is guaranteeing that I won't buy a Zen 3 CPU.... by not supporting x370 I refuse to buy a board for a single GEN of CPU. Might seem a bit crazy since I've only ran Zen 1 on this board but the fact was I expected to be able to go from Zen 1 to whatever the last CPU was on the socket period.... there is no technical reason anyone has legitimately mentioned that precludes this being possible. I am an upgrade they left on the table profit wise because CPUs are upgrades... BOARDS are not. Zen 2 just isn't that much of a differnece that I'd even bother to upgrade from a 1700x ... yes its faster but not a big enough leap to justify the upgrade cost and Zen 3 isnt' enough to justify buying a board and a CPU.
    1
  15. To be fair GN is not located that far from a huge technology area on the east cost (I'm about 3-4 hours inland from him) and probably is fairly close to the server that is hosting his session... so this is probably all ideal conditions. Alot of times my internet connection can't even stream YouTube reliably even though it is "200mbps/20mpbs"
    1
  16. Should call it the winter mod, gotta heat the house anyway right!? Fires up the dual opteron 6386 se OC engineering samples.....toasty.
    1
  17. yes mobo manufacturers that use soldered FLASH proms are cheapskates its like a difference of pennies.
    1
  18. I can't believe how you answered the Vega CU question.... which was specifically about the new 7nm Vega APUs which have 8 instead of 11 CUs on the die.... the one and only reason for that, is memory bandwidth, the 7nm Vega CUs can chew threw that bandwidth and deliver performance at higher clocks than the 12nm ones so there is less bandwidth to go around, and over 8CUs would be completely starved. So yes the answer is diminishing returns due to inadequate memory bandwith on APUS, for laptop parts they could get around this but it isn't an issue that will go away on desktop without a larger IHS or some sort of stacking or embedding passive components inside the PCB etc...to allow more room for say a stack of HBM.
    1
  19.  @burdmate  Keep saying crap like that and it will... talk to people instead of interacting on line.. BOOM problem solved.
    1
  20. electron beam lithography has been around since the early 90s with resolution easily down to 20nm on adapted electron microscopes and 10nm or less in custom designed hardware, the problem is that its extremely slow... current efforts are in trying to speed it up by various methods like using a grid of electron beams all scanning at once etc...still slower than using mask lithography though far more powerful.
    1
  21. TI SXLC2 or IBM Blue Lightening DLC3
    1
  22. Steve, Zen 3 is still on "N7+" but the node is just 7nm... note they weren't differentiating nodes at all in that sense. They dropped 7nm+ to align with TSMCs nomenclature which is 7nm for the node and N7 N7+ N6 etc basically for the gate library and particular fab setup.
    1
  23. I will never buy a Designare motherboard because of the way your pronounce it... it makes me cringe. And sadly I think you are actually right.
    1
  24. Just to reiterate... 15W 8 core.... the lower end bins will potentially be even lower wattage with smaller GPU and CPU. It is unlikely Intel's current 10nm can beat AMD in the mobile space right now and even be affordable.
    1
  25. To be fair.... Radeon VII is a disappointment but not exactly bad, its more like meh. Zen TDP is a lame thing to put on this list... AMD has the best perf/w right now in CPUS... and the TDPs are fine, they are measured differently and in honestly a more realistic way...
    1
  26. overclocks microcontroller embedded in.... you finish the sentence :D alot of micros are super easy to OC by as much as 25% too... since they are often designed with really good safety margins. Actually MCU overclocking competitions would be really fun and cheap to get into... kind of like the pinewood derby of OCing. Rules could be something like no heatsinks, no non standard copper boards 1oz copper only, any oscilator you like, any passive components you like and a specific MCU.
    1
  27. Motherboard form factors are more so a hole pattern than anything.... please take a lesson from Wendel and chill. You can call out things and inform people without going on a rant like this which frankly is not enjoyable to watch...
    1
  28. Are you able to OC that ram any even if you have to losen timings a little it may perform better if you can get it closer to the ideal 3733?
    1
  29. They say they will have a competitor to ARM... NOT develop an ARM CPU... with Nvidia buying ARM it would be extremely foolish and frankly very pointless. x86 and ARM are not that different power use wise when they are implementing the same performance levels.
    1
  30. Right to repair does nothing for people that want the right to tinker... if you can break DRM to fix your phone, you should be allowed to break DRM to improve it outright as well.
    1
  31. A black PCB direct mounted to the side of the cooler with short cables to the fans or direct header connections into the fans would be ideal. I'd rather have to take the board off to service the fans... than ever have to manage those cables. Solutions don't have to be proprietary... what we have instead is a thoughtless mess.
    1
  32. 50 bolts and a bit of metal costs nearly nothing in mass production...
    1
  33. That's why they made it after all.... that board dispenses overkill without prejudice.
    1
  34. Cat be like... let me ESD the shit out of these video cards.
    1
  35. AMD has no shame so probably... gonna try to avoid a blower for the next card I've had a Sapphire Nitro + RX480 and that was good, and I picked up a cheap Nitro Fury X recently and it is good... so not going to repeat the mistake of buying a blower card (Vega FE).
    1
  36. OK... so significantely slower than an RX550 the slowest GPU I could ever imagine buying for anything even an HTPC. RX550 seems to hit around 50-60FPS with dips, at 1080p low.
    1
  37. Wish the Asrock ITX boards had that... the metal "heatsink" blocks get pretty warm on mine with a 1700x.
    1