|
BeastOfExmoor posted:You mean 8 cores? 1600 (and i5-8400) already have 6 cores. I was thinkin more like 6 cores all the way to the entry level for the current prices. Like a 6/12 in the sub-$150 price range would be really cool/disruptive. Im just pulling stuff outta my bhole though. Another retailer leaked today (this time in Japan) that Ryzen 2xxx series aka Ryzen+ is coming in March, so the date seems pretty likely. https://www.tweaktown.com/news/60389/amds-next-gen-ryzen-2-processors-reportedly-launch-march/index.html
|
#
?
Jan 5, 2018 20:47
|
|
|
|
| # ? Jun 5, 2024 14:19 |
|
|
Khorne posted:GPUs are a security nightmare, but due to the nature of GPU computing I don't really see what you described happening. Multi-user GPU environments (SR-IOV) are such a niche setup it's not a big deal. Even if you are virtualizing, normally you do passthrough of the whole device, not let multiple users run at the same time. And you can't even run SR-IOV without a Quadro or a Radeon Pro or a full-on datacenter card. Similarly, even a single user will rarely run many concurrent programs on a GPU at once - you run maybe one GPGPU app or game, plus your desktop renderer, tops. So there's really very little value to breaking it anyway - the only interesting place to go is breaking the drivers and getting kernel mode. There are almost certainly exploits here, but they're just way down on the list in terms of number of users who might be exploitable and the impact you can have. Your time's better spent finding an exploit that works on more than 0.00001% of computers. Paul MaudDib fucked around with this message at 21:09 on Jan 5, 2018 |
|
#
?
Jan 5, 2018 20:58
|
|
|
Cygni posted:I was thinkin more like 6 cores all the way to the entry level for the current prices. Like a 6/12 in the sub-$150 price range would be really cool/disruptive. Im just pulling stuff outta my bhole though. Here's 6/12 for $190, would getting it down to $150 be that big a deal? http://www.microcenter.com/product/478826/ryzen_5_1600_32ghz_6_core_am4_boxed_processor_with_wraith_spire_cooler
|
|
#
?
Jan 5, 2018 21:07
|
|
|
Cygni posted:I was thinkin more like 6 cores all the way to the entry level for the current prices. Like a 6/12 in the sub-$150 price range would be really cool/disruptive. Im just pulling stuff outta my bhole though. More cores than four cores doesn't make sense all the way down the stack, though. I do agree that these super-thermally-constrained dual-cores than Intel keeps shoving into ultrabooks need to loving die as a concept, but as long as Apple exists, they will choke the gently caress out of those things with no active cooling from now until doomsday. As has been demonstrated a number of times (https://www.youtube.com/watch?v=9DyUitTVWlw) those things are throttling so hard that just plugging up the ports on a Macbook Air and setting it into a tray of water completely cured the throttling that the Air does any time it spins up a workload that can't be resolved in under ten seconds. Back to the original point, while dual-core anything in the notebook space really needs to stop, bumping everything to six cores doesn't make sense in may cases still. Even ARM still doesn't do more than four cores of each type in big.LITTLE chips.
|
|
#
?
Jan 5, 2018 21:12
|
|
|
Twerk from Home posted:Here's 6/12 for $190, would getting it down to $150 be that big a deal? http://www.microcenter.com/product/478826/ryzen_5_1600_32ghz_6_core_am4_boxed_processor_with_wraith_spire_cooler Most people don't have access to the lovely prices of Microcenter. But I was thinkin like the $109 original launch price of the R3 1200. High thread count stuff that low in the stack would, provided it was fast enough, likely force Intel to respond. In the long run, that would mean more people getting higher thread CPUs faster, which would hopefully encourage more games and applications to take advantage of that glut of power. But again, just pullin stuff out my bhole. SwissArmyDruid posted:More cores than four cores doesn't make sense all the way down the stack, though. Oh i wasnt talkin' about mobile. Considering AMD still has barely launched Ryzen mobile today, I have no idea what their plans are there.
|
|
#
?
Jan 5, 2018 21:13
|
|
|
Cygni posted:Most people don't have access to the lovely prices of Microcenter. But I was thinkin like the $109 original launch price of the R3 1200. High thread count stuff that low in the stack would, provided it was fast enough, likely force Intel to respond. In the long run, that would mean more people getting higher thread CPUs faster, which would hopefully encourage more games and applications to take advantage of that glut of power. I'm skeptical that price differences among sub-$200 CPUs really matter, especially in the desktop space. The rest of the computer, especially RAM, is really drat expensive too. I'm all for more power to more people, but the CPU doesn't really dictate the price of the whole machine.
|
|
#
?
Jan 5, 2018 21:24
|
|
|
Paul MaudDib posted:Multi-user GPU environments (SR-IOV) are such a niche setup it's not a big deal. Even if you are virtualizing, normally you do passthrough of the whole device, not let multiple users run at the same time. And you can't even run SR-IOV without a Quadro or a Radeon Pro or a full-on datacenter card. I contract for a place that has a few hundred Tesla cards. The situations their customers get themselves into entirely by accident would never happen with CPUs. Khorne fucked around with this message at 21:33 on Jan 5, 2018 |
|
#
?
Jan 5, 2018 21:29
|
|
|
Khorne posted:I was just saying GPUs are a sysadmin nightmare. Not even in SR-IOV. In any shared computing environment. Say you have no virtualization at all, exclusive_process mode, and 4 tesla cards in a chassis with a scheduler only allowing 1 user per gpu. Users still need to have a level of trust in each other even though you [likely] can't access anything in memory. I'm not even sure some of the abuses count as an exploit. Yeah I would expect there is a significant amount of leakage/vulnerability at the runtime level, ideally you want each of the users in their own virtualized environment with one user per runtime, not one GPU per user on a shared runtime. Paul MaudDib fucked around with this message at 22:05 on Jan 5, 2018 |
|
#
?
Jan 5, 2018 21:41
|
|
|
Cygni posted:Essentially, yeah. It's on a slightly improved process, and they will be launching new chipsets (rebadged? i cant think of any new features to add). But they have explicitly said its the same architecture, so maybe a few hundred more mhz? A few more hundred Mhz would be my target as well. 16 threads @ 3.8-4.0 ghz for cheap is GOOD but not really super compelling considering how long overclocked sandy bridge has managed to hold out. Those 16 threads @ 4.4 or 4.5 though and things start looking much better. Yeah it's only about 10% on paper but considering the IPC improvements for the last 5 years has been just creeeeeeping along, that 10% clock translates into at least a year of extra relevency. And speaking of futureproofing, isn't 7nm rolling around in 2019? These "improved Ryzen" parts would only have been out for no more than 15-18 months before being "obsoleted" if the promises of Zen2 are to be believed. Smaller dies = 32c/64t Threadrippers @ 5.0 ghz? HOLY SHEEEEEEEIITT TLDR: I'm just a big fat babby and don't want to get a new CPU for 6-7 years after seating it.
|
|
#
?
Jan 5, 2018 23:10
|
|
|
Seamonster posted:And speaking of futureproofing, isn't 7nm rolling around in 2019? Given the 10nm delays and investment required to manufacture it, I wouldn't hold my breath for 7nm until 2119.
|
|
#
?
Jan 5, 2018 23:17
|
|
|
Twerk from Home posted:I'm skeptical that price differences among sub-$200 CPUs really matter, especially in the desktop space. The rest of the computer, especially RAM, is really drat expensive too. I'm all for more power to more people, but the CPU doesn't really dictate the price of the whole machine. Shaving off 50 bucks CPU might let you go from a 1050ti to 1060 which is a pretty big deal if you're on a tight budget.
|
|
#
?
Jan 5, 2018 23:38
|
|
|
Perhaps not as outright awful as RAM but SSD prices have increased per GB as well. And innovation in terms of drive size has utterly stagnated too. Where the hell are our 4TB drives for ~$800?
|
|
#
?
Jan 6, 2018 00:51
|
|
|
Spinning rust, that's where.
|
|
#
?
Jan 6, 2018 01:11
|
|
|
spasticColon posted:When is Ryzen+ expected to drop? Thanks to miners - most likely not soon... , as well as the positions of video cards, of mining support prices for new hardware, and, perhaps, nvidia and amd also have to do with the high demand for bitkoint - such a vicious circle. P.S. - wait until obsolete 14nm
|
|
#
?
Jan 6, 2018 01:24
|
|
|
unstucker posted:Thanks to miners - most likely not soon... , ...?
|
|
#
?
Jan 6, 2018 01:30
|
|
|
Seamonster posted:Perhaps not as outright awful as RAM but SSD prices have increased per GB as well. And innovation in terms of drive size has utterly stagnated too. Where the hell are our 4TB drives for ~$800? High capacity SKUs exist but they are for the people that will pay for high margin products. In a supply constrained environment why waste it in race to the bottom client products
|
|
#
?
Jan 6, 2018 01:30
|
|
|
Seamonster posted:TLDR: I'm just a big fat babby and don't want to get a new CPU for 6-7 years after seating it. Considering how often this comes with mobo and memory changes, I don't blame you. I have a HTPC running on an i3 4360 and if I were going to change it I would rather just buy a more powerful Haswell and stick it in the slot than deal with the generational changes everywhere else.
|
|
#
?
Jan 6, 2018 01:36
|
|
|
Supposedly AMD's gonna give the AM4 Socket quite a longevity compared to Intel, so that would mitigate it a bit.
|
|
#
?
Jan 6, 2018 01:39
|
|
|
Cygni posted:Essentially, yeah. It's on a slightly improved process, and they will be launching new chipsets (rebadged? i cant think of any new features to add) Cygni posted:But they have explicitly said its the same architecture, so maybe a few hundred more mhz? Cygni posted:If yields are good, maybe we get 6 cores all the way down the stack or more aggressive pricing, both of which would be cool for gamers targeting the R5 1600/ i5 8400 bang-for-the-buck zone. For anyone with a current AM4 chip or a Skylake or newer system your probably best off saving money for a better GPU or monitor until Ryzen2/+/whatever they're calling it now comes out in super late 2018/early 2019. That sounds like it could be a big IPC + clock speed jump there. Unless maybe you need more threads I guess but most won't need more than what a 8C/16T or even "just" a 6C/12T Ryzen/Intel chip can offer.
|
|
#
?
Jan 6, 2018 02:15
|
|
|
Seamonster posted:Perhaps not as outright awful as RAM but SSD prices have increased per GB as well. And innovation in terms of drive size has utterly stagnated too. Where the hell are our 4TB drives for ~$800? Innovation hasn't stagnated it's just the supply shortage keeping prices high which makes higher capacities make less sense, things should be better by 2019 sometime when more production capacity is coming online.
|
|
#
?
Jan 6, 2018 02:38
|
|
|
Seamonster posted:Perhaps not as outright awful as RAM but SSD prices have increased per GB as well. And innovation in terms of drive size has utterly stagnated too. Where the hell are our 4TB drives for ~$800? Where are you seeing SSD prices go up? It sure looks like prices have either held or gone down anywhere I look. And as far as innovation in pure size, Samsung's got 16TB drives that fit in the 2.5 inch SAS form factor and Seagate has 64 TB demonstrators in 3.5 inch form factor (though you can't buy those Seagates on the market). Considering as you can still only go up to 12 terabytes in a single spinning disk drive, that's pretty impressive.
|
|
#
?
Jan 6, 2018 04:23
|
|
|
fishmech posted:Where are you seeing SSD prices go up? The sale prices this year were just about last year's normal prices. SSDs have gone in price by about $30-$50 a piece depending on capacity. I picked up 3 1tb 850 evos for about $260 1.5 years ago. They've hit $300+ normal price this year.
|
|
#
?
Jan 6, 2018 05:42
|
|
|
oops
|
|
#
?
Jan 6, 2018 07:07
|
|
|
I suppose but fix is already out and can’t you just outright disable PSP anyway?
|
|
#
?
Jan 6, 2018 07:19
|
|
|
Haha boy I still can't believe either AMD or Intel though it'd be a good ideal to implement these sorts of "secure and trusted" shitshows into their systems or chips. Its virtually guaranteed they'll either miss something in testing or years down the road the latest greatest security methods are found to have some sort of exploitable flaw and then they've got a mess on their hands. If they'd at least give the end users some way to disable it permanently (ie. a jumper on the mobo) then I wouldn't care but neither seem to want to do things the right way here. 8-bit Miniboss posted:I suppose but fix is already out and can’t you just outright disable PSP anyway? PC LOAD LETTER fucked around with this message at 07:24 on Jan 6, 2018 |
|
#
?
Jan 6, 2018 07:20
|
|
|
8-bit Miniboss posted:I suppose but fix is already out and can’t you just outright disable PSP anyway? Also, looking at the actual exploit description linked, "remote" involves a crafted certificate going to the TPM so I'm not sure if disabling the TPM wouldn't get it anyway? My Ryzen mobo's firmware doesn't allow disabling the PSP, but I did disable the TPM just in case.
|
|
#
?
Jan 6, 2018 07:27
|
|
|
Paul MaudDib posted:Multi-user GPU environments (SR-IOV) are such a niche setup it's not a big deal. Even if you are virtualizing, normally you do passthrough of the whole device, not let multiple users run at the same time. And you can't even run SR-IOV without a Quadro or a Radeon Pro or a full-on datacenter card. LOL tell that vdi deployments at big banks and law firms Also consider poo poo like webgl which allows for Gpu access but cannot be sandboxed well from the rest of the system. And even every browser is GPU accelerated so...
|
|
#
?
Jan 6, 2018 14:18
|
|
|
fishmech posted:Where are you seeing SSD prices go up? It sure looks like prices have either held or gone down anywhere I look. And as far as innovation in pure size, Samsung's got 16TB drives that fit in the 2.5 inch SAS form factor and Seagate has 64 TB demonstrators in 3.5 inch form factor (though you can't buy those Seagates on the market). Considering as you can still only go up to 12 terabytes in a single spinning disk drive, that's pretty impressive. NAND gas been in undersupply since 2016. Chip prices have gone up.
|
|
#
?
Jan 6, 2018 14:20
|
|
|
Malcolm XML posted:LOL tell that vdi deployments at big banks and law firms can you please tell me how this post should best imply my undying hatred for AMD? I'm really not sure how best to proceed from such an ambiguous post. (vdi deployments at banks probably aren't using SR-IOV for the most part, VDI does not imply multi-user on the same runtime let alone the same device)
|
|
#
?
Jan 6, 2018 14:44
|
|
|
Paul can hate on AMD all he wants. He puts a lot of effort into his posts and I appreciate that.
|
|
#
?
Jan 6, 2018 17:33
|
|
|
Actually I guess banks might be Citrix'ing everything and that might be GPU-accelerated... dunno, our Citrix sure isn't. There are compliance regulations in banking that are definitely abnormal compared to everyone else. Like MITM'ing all employee connections as a matter of policy, hence the weakening of TLS 1.3. I actually don't know if that uses SR-IOV or if it runs at the Citrix level as a single application at the runtime level, but yeah either way it's certainly possible that leaks too. GPU drivers are really bad and as a general rule of thumb you shouldn't expect anything out of them that you haven't experimentally verified, and even then implementation details could easily change with the next uarch (where "implementation details" are virtually any bare-metal optimizations you might want to do). That goes double for anything security-related. (especially now that everybody is going to be running old drivers that NVIDIA approved for datacenter use...) Paul MaudDib fucked around with this message at 19:06 on Jan 6, 2018 |
|
#
?
Jan 6, 2018 19:00
|
|
|
redeyes posted:Paul can hate on AMD all he wants. He puts a lot of effort into his posts and I appreciate that. Thing is I actually don't hate AMD... Ryzen's benefits are self-evident, you are getting twice the cores at a given pricelevel. It's just the lack of official ECC support/validation and the segfault problem that hold me back at the moment. When there are official Ryzen Pro boards with ECC support and the segfault bug is fixed (i.e. second-gen) I am probably going to buy one. If there was a mATX Threadripper board I probably would already have one. It's just that the things I do are either gaming (Intel wins) or server stuff where segfault is unacceptable and official validation is a major major plus, and I don't mind paying $100 extra per system or w/e to make it happen for either of those roles. Also, tbh most of the AMD mobos are cheap garbage in comparison to Intel's products. A B350 board for $50 is not really a selling point for me. That heavily colors my opinion on Ryzen... and the patch really doesn't change the basic dynamics of either the gaming market (Ryzen loses) or the server market (first-gen Ryzen is a non-starter). But the day AMD ships a 8-12 SATA mobo in mITX or mATX with ECC support someone needs to let me know, again especially if it's threadripper. When that happens, AMD is going to have to try and stop me from shoving money into their hands, because the basic product really is solid... it's just that it flops in most of the details. Vega is pretty much inexcusable though, nobody but miners is really interested at the prices that AMD wants to make them at. Unfortunately, that is probably going to screw over the gaming market next generation too... Paul MaudDib fucked around with this message at 19:42 on Jan 6, 2018 |
|
#
?
Jan 6, 2018 19:35
|
|
|
Paul MaudDib posted:If there was a mATX Threadripper board I probably would already have one.
|
|
#
?
Jan 6, 2018 19:44
|
|
|
Llamadeus posted:Just announced! Naturally, it's Asrock: https://www.asrock.com/news/index.asp?iD=3888 Sweet jesus, Asrock did it again 
|
|
#
?
Jan 6, 2018 19:49
|
|
|
as always ASRock uses all the necessary magic power to make it all happen They know they're the good kind of crazy.
|
|
#
?
Jan 6, 2018 20:01
|
|
|
What is this? A 32 layer PCB?
|
|
#
?
Jan 6, 2018 20:10
|
|
|
Llamadeus posted:Just announced! Naturally, it's Asrock: https://www.asrock.com/news/index.asp?iD=3888 4 DIMM slots, 8 SATA slots, 3 m.2 slots and 1 u.2, 3 front USB headers, 3 PCI-e 16x slots, dual Intel Gigabit, dual band WiFi, at least 8 power phases.
|
|
#
?
Jan 6, 2018 20:10
|
|
|
Now do ITX.
|
|
#
?
Jan 6, 2018 20:11
|
|
|
Cygni posted:Now do ITX. Put 4 so-DIMM slots on the back along with the 3 m.2 slots.
|
|
#
?
Jan 6, 2018 20:15
|
|
|
|
| # ? Jun 5, 2024 14:19 |
|
|
I'm already eyeballing the board design and I think I see a Mini-ITX TR board in there.
|
|
#
?
Jan 6, 2018 20:16
|
|
