• PC Building Thread V6 - "running six RGB controller utilities at once" edition
    999 replies, posted
idk if its worth splurging on one of those right now, pcie 4.0 drives are coming soon.
Good point, but for read performance might see a 20% improvement (NAND isn't getting the same 2X speed increase as PCIe4) and for write speeds it probably won't make a difference beyond the usual generational improvements. It's also up in the air how real PCIe4 is going to roll out. So it might be "smart" to get a 980 Pro, but that'll probably come down to price and availability.
I'm not convinced the slot is what's holding back pcie drives. A 970 pro only reads at about 2/3 of the theoritical limit for pcie 3.0 x4, and that's the fastest drive on the market. Something a little more typical doesn't even swamp an x2 lane. If it was the slot holding it back I think stuff like the 900p would be using a full 16x cart.
https://flash.newegg.com/Product/9SIA53D94U7163 OEM equivalent of 960 Pro is $150, get 'em while they're hot
Guys, quick question got a NZXT S340 case, a quad fan AIO on the front doing the push pull thing, a 120 fan at the back above the gpu and a 140 at the top. Which positions would be best to put the rear and top fans for the gpu to breathe a bit more?
Push/Pull with pressure optimized fans and a 140 top and 120 rear exhaust is as good as you're gonna get with that case aside from taking the side panel off and sticking a box fan into it. You're going to want better fans and use like Speedfan or something to better adjust your fan speeds to cpu/gpu temps. You could maybe try cutting a fan into the basement or buying something like this to mount some fans next to your GPU. I know someone whos had good luck with a blower Vega card and one of these. https://www.amazon.com/GDSTIME-Graphic-Card-Graphics-Cooler/dp/B07H5KPY8P/
For now I'll go for the 4 by 2 deal I guess otherwise I'd have to change cases I guess...
Hey I put together a build, just wanted to get people's thoughts on it. https://pcpartpicker.com/list/gyXyYT any recommendations would be appreciated
With a B450 motherboard, I'd just spend the extra $10 for the improved XFR2 and better stock cooler on a 2600x. You won't want to overclock with XFR2 generally doing a better job anyways. Otherwise I'd go 2600 if it was still on sale and use that extra cash for X470. Skip the sata drives, 500GB/1TB drives can be found in a $5 bin at the computer recycler. Bare 4TB drives are $100~ and if you want better drives 8TB's can be shucked from externals for $130. https://flash.newegg.com/product/9SIA53D94U7163 https://www.amazon.com/dp/B07GCL6BR4/ https://flash.newegg.com/product/20-326-190 https://www.amazon.com/WD-Blue-4TB-Hard-Drive/dp/B013HNYV8I/ I'd also recommend looking for either 1440p or 144hz in a monitor, VA panels look pretty decent. Skip buying windows 10 entirely, keys are less than $5 on ebay and if it ever deactivates windows just buy another. https://www.ebay.com/itm/WINDOWS-10-PROFESSIONAL-PRO-KEY-32-64-BIT-GENUINE-ACTIVATION-CODE-LICENSE-KEY/352604516163 https://www.ebay.com/itm/Windows-10-Pro-Professional-Key-32-64-Bit-Activation-license-instant/233054806818 https://www.ebay.com/itm/WINDOWS-10-PROFESSIONAL-PRO-KEY-32-64-BIT-ACTIVATION-CODE-LICENSE-KEY-GENUINE/123678633545 You also definitely don't need 650w, 550w is plenty. https://pcpartpicker.com/list/ypW4cY
https://arstechnica.com/gadgets/2019/05/amd-to-launch-new-7nm-navi-gpu-rome-cpu-in-3rd-quarter/ "the company said that the first Navi parts will be priced below the $699 Radeon VII." No high-end at launch
Unless they've got a Navi design that can do HBM2 and GDDR6 (or god forbid they're doing a Navi 64 / separate die design) then that's no high-end at all for Navi. Radeon VII was your "high end".
Here's a leaked PCB that's likely for a navi gpu https://youtu.be/ckJIy0L7LHY gives us an idea of the level of performance we should expect this is most likely the highest end model considering how much power it draws
Well, it's still possible that the best Navi will be around VII/2080 performance level while being cheaper at the same time. They do have the advantage of 7nm and if they make it a pure gaming chip without all the unnecessary datacenter features, it could be a good bit smaller and more economical than the VII.
unlikely, Radeon VII is already on 7nm, and HBM is a smaller and more efficient IMC anyway. Barring significant improvements to the GCN uarch (DSBR,TBR, etc) then Navi is basically just going to be a mid-tier product, and VII will be the "flagship".
A Navi 64 with HBM2 would probably be really fast, but it would be expensive to design and make. And Nvidia could just cut the price of the 2080 ti if they feel slightly threatened (Good for consumers, bad for AMD)
That's why I said without the unnecessary datacenter features. The Vega GPU in VII has 1/2 rate FP64 (artificially nerfed in firmware), just cutting out a large portion of those fp64 units and fixing the broken Vega features could bring a significant bump in efficiency. I'm not saying they'll deliver 2080 perf for 300$/€ or whatever those crazy rumors claimed, but considering how long it's been in development, they should be able to hit that level at a somewhat more reasonable price. Remember, they're competing against Nvidia who's still using an outdated manufacturing node and making humongous dies with all those extra tensor/RT units, if they can't at least somewhat catch up to that it's gonna be a serious embarassment.
There's not really any evidence that 1:1 FP64 actually costs anything significant silicon-wise. And 2:1 FP16 is a selling point for Radeon under Rapid Packed Math. That's pretty much it for "datacenter" features. The whole "designed for the datacenter" in regards to Vega is a meme anyway. The only way anything is going to perform better than Radeon VII is if it beefs up the front-end, doubles ROPs, and/or adds DSBR/TBR, and uses HBM2 (or 384-bit G6). I don't see any of those happening, Navi is going to be mid-range. I'd love to be wrong, but RTG has a history of underperforming.
I think the rumor was 1080 performance for $300.
Well, obviously FP16 units will be a lot simpler than FP64, it doesn't cost much to implement more of them for double throughput. Yes, the 14nm Vega was meant to be a gaming card from the beginning, but I think it's pretty clear that the 7nm version was then repurposed as a compute card first. Not a meme at all. All I know is that the 7nm version is bigger than you'd expect it to be, some of it is due to extra FP64, but the extra-wide memory bus and other changes might have something to do with it too. Nobody even expected the VII, it looks like it was a quickly cobbled together product just so they can fill this long gap of no high-end releases with Navi taking too long. I don't buy the explanation that VII was always meant to be the high end product in the lineup, it's just too inefficient and expensive to produce.
Pretty sure they basically just fattened the registers, and the fatter memory bus, to get 1:1 FP64 (from the FP32 native FMAs). I mean all HBM2 cards have ECC available in the memory (and in the IMC) and that's enabled in the MI cards, same with SRIOV, but pretty much all modern AMD designs (Polaris included) have those features. The "datacenter" design is PR memery. I agree the VII was last minute, but I also think AMD knew they'd have to release something high end once they realized the Navi design wasn't revolutionary. Again, I'd love to be proven wrong, but unless we get a 60~ CU HBM2 Navi card, I don't see high-end being in it's wheelhouse.
isn't a lot of the issues with vega inherent to GCN
Thanks for the advice, gonna go with the 2600x and the B450, as well as the monitor you posted after realizing I don't actually need to buy WIN 10 since I can just upgrade my existing key lol. I haven't gotten a new computer since 2011, so this is very exciting for me. Only hard part left is to pull the trigger on all that dosh.
Let's set aside whether the design of the GPU itself is a meme, but the gaming version definitely didn't need 4 stacks of HBM, 12GB would've been more than enough. The "might reach 2080" levels is pure speculation on my side, but if it doesn't at least surpass V64, Navi is gonna be kinda useless.
Yeah I can agree the 4X HBM2 is a design for datacenter, but that's just a large memory interface that they tacked on, IMO. I think Navi is going to be V64 performance (plus or minus in some titles) for like $300 and V56 for $200
No. The only difference you would notice would be Wifi and an extra m.2 slot.
So I have a spare 760 SCC 2Gb, my wife has a 950 2Gb, they're pretty similar cards but would it be worth the hassle putting the 760 in instead of the 950?
Both of those motherboards are the same aside from an extra m.2, better vrm heatsinks, rgb and wifi and the upgraded audio on the Carbon. The seems Tomahawk has an extra usb2 header. Computer parts rarely go for their MSRP, in reality that extra $60 is just a seller which has stock of something that Newegg doesn't so they want to rip off people who simply hit checkout on pcpartpicker or follow build guides. It's only $20 more, which if you want Wifi is probably worth it if you don't have a better adapter. https://www.amazon.com/MSI-Performance-Motherboard-B450-PRO/dp/B07F85YKLJ
Some of them. Most of Vega's problems have to do with HBM2. Thank god it looks like Navi might use GDDR6.
The only issue ever with HBM has been pricing. It is simply more expensive per Gb than G5 or G6. And then packaging prices can hurt too. In reality HMB2 uses less silicon space, and less power than GDDR, and I wish AMD could get it's prices competitive with GDDR HBM could be mainstream.
There's also the problem with having to package unnecessary amounts of VRAM in order to get adequate memory bandwidth, which really hurts price.
Sorry, you need to Log In to post a reply to this thread.