|
Combat Pretzel posted:Those assholes over at AMD need to finally talk about socket compatibility. More so with those current rumors about SP3r3/TR4+. TDP is another interesting topic, there were some slides floating around showing the 16C TR with 280W TDP as opposed to he 105W of the 3950X with the same core count. If this turns out to be true they are either using much lower quality silicon for TR (unlikely because they historically saved the best bins for it) or it is going to run much higher clockspeeds. GN‘s recent video on AMD TDP is really good, by the way.
|
#
?
Oct 18, 2019 07:25
|
|
|
|
| # ? Jun 11, 2024 11:02 |
|
|
TSMC spending a shitload to expand 7nm/5nm capacity. Sounds like they expect to be the process leader for a while. https://www.anandtech.com/show/15001/tsmc-radically-boosts-capex-to-expand-production-capacities
|
|
#
?
Oct 18, 2019 18:58
|
|
|
It's either them or Samsung, and their process is available right now and demand is high, so expansion makes sense. 7nm has a lot to offer and few players in the market.
|
|
#
?
Oct 18, 2019 19:01
|
|
|
K8.0 posted:It's either them or Samsung, and their process is available right now and demand is high, so expansion makes sense. 7nm has a lot to offer and few players in the market. and gloflo decided to not do 7nm and instead deploy lawsuits, what a poorly run company.
|
|
#
?
Oct 18, 2019 19:07
|
|
|
wargames posted:and gloflo decided to not do 7nm and instead deploy lawsuits, what a poorly run company. They haven't made a profit ever, their CEO is trying to get them to profitability - hence their investment in lawsuits instead of technology.
|
|
#
?
Oct 18, 2019 19:16
|
|
|
wargames posted:and gloflo decided to not do 7nm and instead deploy lawsuits, what a poorly run company. gloflo: "Global Floundries" Also the Saudi hedge fund thing that owns them is getting antsy and wants to see an immediate return, not dropping another couple billion to get 7nm up and running and maybe breaking even in 3-5 years. But yeah it seems like GloFo has always been exceptionally poorly run and long-term it's probably a good thing for AMD to be able to move away from those contracts and shift to TSMC. I have no idea who that leaves GF with... a lot of smaller companies on older nodes I guess? Seems like it devalues GloFo as a brand to cede process leadership, they will just be left with a dwindling supply of customers who want their lovely older nodes. Paul MaudDib fucked around with this message at 20:41 on Oct 18, 2019 |
|
#
?
Oct 18, 2019 20:38
|
|
|
Wouldn't it be a kick if AMD bought back GloFo, and sold off everything to restart R&D funding for production from 7nm and below?
|
|
#
?
Oct 18, 2019 21:18
|
|
|
SwissArmyDruid posted:Wouldn't it be a kick if AMD bought back GloFo, and sold off everything to restart R&D funding for production from 7nm and below? easier to let tsmc and samsung to take that risk of build stuff out.
|
|
#
?
Oct 18, 2019 21:22
|
|
|
AMD confirms that Zen 5 is already in developmentwargames posted:easier to let tsmc and samsung to take that risk of build stuff out. The way this perfect storm has brewed up with Intel playing catchup on a tech, process, and mindshare basis? If Mubadala were to auction off GloFo to the highest bidder, I think it would be in AMD's interest to at least put in a nominal bid. It's all highly unlikely of course, but if it were to happen, I can think of a lot of worse problems than having some production back in-house, even if that production is purely licensing Samsung process tech, like GloFo did with 14nm.
|
|
#
?
Oct 19, 2019 01:51
|
|
|
SwissArmyDruid posted:
Hell of a change from a few years ago when people were praying for AMD to get GloFo's albatross off their neck.
|
|
#
?
Oct 19, 2019 01:59
|
|
|
yeah, it's amazing how fast the script has flipped
|
|
#
?
Oct 19, 2019 02:10
|
|
|
MagusDraco posted:Hell of a change from a few years ago when people were praying for AMD to get GloFo's albatross off their neck. Let's be real, it is STILL an albatross. The WSA is still an obligation that dictates their moves going forward.
|
|
#
?
Oct 19, 2019 03:42
|
|
|
Yea but they've at least found away to alleviate it with I/O dies. If EPYC2 is a roaring success then the ISA will never be a limitation on AMD's finances. The real issue will be when AMD has moved significantly beyond Zen2, like 5-6 years down the line, why the gently caress would AMD continue to work with a foundry partner who can't supply them with a relevant process?
|
|
#
?
Oct 19, 2019 22:22
|
|
|
That second part is my point. TSMC has been running 5nm risk production since April.
|
|
#
?
Oct 20, 2019 00:15
|
|
|
What process is the x470 chipset on?
|
|
#
?
Oct 20, 2019 00:27
|
|
|
55nm.
|
|
#
?
Oct 20, 2019 00:53
|
|
|
TSMC's revenue graph from the article at the top of the page shows there's plenty of money in older processes. GoFlo's problem is they have to start investing in a new node eventually or they'll fade into embedded irrelevancy.
|
|
#
?
Oct 20, 2019 11:36
|
|
|
Supposedly GF's 7nm that got cancelled worked fine enough and was done. Some of the leaks suggested it would've been pretty similar to TSMC's 7nm overall in performance/yield/power. The rumor was GF's owners got cold feet due to the risk from dramatically increased costs and/or wanted more money now and not maaaaybe 2-5yr down the road depending on how many customers GF was able to get. While what happened was pretty disappointing realistically GF does actually have a decent shot at making some money for a couple of years at least with just its current tech since their 16nm and "12nm" processes will still be in the economic/performance sweet zone for lots of things. Then they could move to 7nm when it gets cheaper/more profitable to do so. But yeah if they wait too long their window of opportunity will go away fast if for no other reason than TSMC and Samsung will just grow their fab space to accommodate all possible 7nm-or-less process demand.
|
|
#
?
Oct 20, 2019 22:53
|
|
|
anandtech published their 15" Surface review with Ryzen and Vega https://www.anandtech.com/show/15008/the-microsoft-surface-laptop-3-15-inch-review
|
|
#
?
Oct 21, 2019 14:06
|
|
|
"More about partnerships than product," Yeah, alright, I'm down with that. I am still loving dying here for benches of the Intel-based ones for context, though. (I will, however, continue to repeat my battlecry of "HD620 CAN GET FUUUUUUUUUUUUUUCKED")
|
|
#
?
Oct 21, 2019 20:00
|
|
|
I'm thinking the Intel one will probably be overall superior, especially since it looks like it'll be coming with LPDDR4X 3700 E: this is based on Intel IGPU at or near parity based on leaks and estimates from their new architecture there and Speedstep getting a callout in the article. NewFatMike fucked around with this message at 20:42 on Oct 21, 2019 |
|
#
?
Oct 21, 2019 20:39
|
|
|
eames posted:anandtech published their 15" Surface review with Ryzen and Vega Quite a lot of low-hanging fruit for AMD to pursue: * LPDDR4X/5 * 7nm * Video acceleration in HW * WiFi MAC integration in chipset/SOC * TB3/USB4 in chipset/SOC It gets trounced in perf/watt: https://www.anandtech.com/show/15008/the-microsoft-surface-laptop-3-15-inch-review/6 Intel is also significantly ahead in the peripherals with TB3, WiFi MAC on chip, and basically can significantly reduce BOM costs for OEMs. No need to integrate a third party TB3 controller if it's on the chipset! Malcolm XML fucked around with this message at 21:25 on Oct 21, 2019 |
|
#
?
Oct 21, 2019 21:22
|
|
|
I'm going to be building a new 3700x computer in the next couple days, and just wanted to make a little checklist of stuff to do to get the most out of it. -Latest chipset drivers. -Use Ryzen DRAM calculator to set memory timings. -Ryzen Master on balanced? -ABBA BIOS? Anything else I'm missing?
|
|
#
?
Oct 21, 2019 23:44
|
|
|
Lowclock posted:I'm going to be building a new 3700x computer in the next couple days, and just wanted to make a little checklist of stuff to do to get the most out of it.
|
|
#
?
Oct 22, 2019 00:47
|
|
|
ufarn posted:Windows power options to "Ryzen Balanced" (with a higher-than-default minimum CPU usage last I checked). But they change it all the time.
|
|
#
?
Oct 22, 2019 03:47
|
|
|
Running Ryzen 5 2600X and MSI Tomahawk B450. Was messing around in the UEFI and noticed that AMD "Cool n Quiet" was set to "auto." Do I want this set to "enabled" instead? I know that this lets the CPU step down frequency when needed, but don't know how aggressive it is, or any other ramifications.
|
|
#
?
Oct 22, 2019 15:12
|
|
|
I'm on a 2600 and I haven't noticed a huge difference with turning it "on" instead of auto. It maybe runs a couple degrees cooler in idle. edit: I think I'm leving it on. My fans seems to be noticeably quieter. It used to run at 1600rpm but now it's down to 1200rpm when just using Firefox. Mu Zeta fucked around with this message at 16:16 on Oct 22, 2019 |
|
#
?
Oct 22, 2019 15:57
|
|
|
betterinsodapop posted:Running Ryzen 5 2600X and MSI Tomahawk B450. Leave it at auto. Auto will just remember to turn it off when you start overclocking. Which, since you have an X chip, you probably kinda sorta are.
|
|
#
?
Oct 22, 2019 19:12
|
|
|
Craptacular! posted:Leave it at auto. Auto will just remember to turn it off when you start overclocking. Which, since you have an X chip, you probably kinda sorta are. RE: kinda sorta overclocking: I have Precision Boost turned on, so I guess that kinda OC's the CPU? I seem to run at 4.2 when the CPU is pushed. I did NOT turn on MSI GameBoost (lol) because it sounds like it applies a pretty aggressive voltage and OC that may not be great. Coming from an i5 Sandy Bridge, it's an adjustment. The only things I've really done is enable XMP and turned on Precision Boost.
|
|
#
?
Oct 22, 2019 20:29
|
|
|
Yeah, not much to tweak. I spent weeks playing with fan curves (and cardboard baffles) in order to get the PC building satisfaction I require.
|
|
#
?
Oct 22, 2019 21:20
|
|
Boys who love boys
|
Yeah, that's kinda where I'm at. Turn on PWM fan control and gently caress around w/the curves and I guess I'm done. 
|
|
#
?
Oct 22, 2019 21:45
|
|
|
Threadripper marketing is beginning... https://www.youtube.com/watch?v=766eoe7obrQ
|
|
#
?
Oct 22, 2019 22:59
|
|
|
Combat Pretzel posted:Threadripper marketing is beginning... "drat I NEED MORE MONEY" I love youtube comments.
|
|
#
?
Oct 23, 2019 01:37
|
|
|
Finally seen pictures of that leaked TRX40 Taichi from Asrock. That's certainly a loving huge heatsink on the VRMs. A proper one with fins. Kinda overcompensating for the weaksauce ones on the X399 ones.
|
|
#
?
Oct 23, 2019 09:53
|
|
|
I'm mainly curious if AMD will bin TR3 to have superior singlecore performance compared to desktop this time around. That would shift a lot of potential buyers to HEDT and I assume that margins are higher there. The main question is if yields allow that kind of strategy shift. It would also explain the 3950X delays and the big TDP difference between 16 core ryzen 9 and leaked threadripper parts.
|
|
#
?
Oct 23, 2019 10:47
|
|
|
GN saying in their news roundup that they have independent confirmation that the new Threadripper breaks compatibility. So no 1/2-series TR in the new TRX40 boards, and no 3-series TR in x399. Same socket just wired different, ala Kaby Lake to Coffee Lake. Odd considering Rome was drop in compatible in some servers, and obvi Zen 2 on AM4 exists.
|
|
#
?
Oct 24, 2019 18:43
|
|
|
Some people were speculating maybe the new Threadripper uses a "chipset-less" design like Epyc, but first-gen Epyc gets most of the way through POST even on TR4 boards with a chipset, so it is not necessarily impossible to run a chipsetless-chip on a chipset board. It seems like they could have made it work if they wanted to.
|
|
#
?
Oct 24, 2019 18:50
|
|
|
yeah i mean both x570 and the Promontory chipsets (all the Asmedia AM4 chipsets) are just hung off of PCIe too. You could hang either off of Rome or a dang Intel CPU if you really wanted to, so i dunno.
|
|
#
?
Oct 24, 2019 19:26
|
|
|
Cygni posted:GN saying in their news roundup that they have independent confirmation that the new Threadripper breaks compatibility. So no 1/2-series TR in the new TRX40 boards, and no 3-series TR in x399. Same socket just wired different, ala Kaby Lake to Coffee Lake. So you're saying there's gonna be a bunch of used x399 boards on ebay soon.
|
|
#
?
Oct 24, 2019 22:27
|
|
|
|
| # ? Jun 11, 2024 11:02 |
|
|
Actuarial Fables posted:So you're saying there's gonna be a bunch of used x399 boards on ebay soon. I just wish there had been more X399 boards with IPMI, there is just that one board from Asrock Rack and it's still $550 and now it's on a dead platform pretty loving sweet board though, 10 SATA ports and dual 10 GbE and 5 3.0x16 ports
|
|
#
?
Oct 24, 2019 22:49
|
|