|
The Seamicro hardware works and works well, it just had pretty loving horrid failure rates and once AMD got a good long look at that they were nope.avi
|
#
?
May 7, 2015 14:47
|
|
|
|
| # ? Apr 25, 2024 18:48 |
|
|
Never heard of that, do you have a link with a story about it or just personal experience? I'd assume it'd be due to implementation (manufacturing) issues rather than something wrong with the design itself per se. There was some thread over at RWT's forums a while back talking about major cost cutting going on but didn't get into specifics that I can remember and can't find the post now. I wonder what AMD would use Seamicro's tech for though. HT3.1 seems already pretty decent and scales out to 16 socket systems if necessary which is more than enough when you consider they'll probably have 16-32 CPU per socket (so 256-512 CPU's total in a 16 socket system, which admittedly would probably be a unusual beastie, 4 sockets would be the far more common max so 64-128 CPU's) when Zen launches. They already have the controllers and everything designed and tested for HT3.1 with lots of business partners too. I'd assume Seamicro's real value was for its patents.
|
|
#
?
May 8, 2015 02:28
|
|
|
PC LOAD LETTER posted:Never heard of that, do you have a link with a story about it or just personal experience? I'd assume it'd be due to implementation (manufacturing) issues rather than something wrong with the design itself per se. Competitiveness with DMI 3.0? That's supposed to hit 8 GT/s, and is slated for Skylake.
|
|
#
?
May 8, 2015 02:34
|
|
|
Was Seamicro's stuff that fast though? Yea total bus bandwidth was high but per socket/CPU? I thought they were more about connecting lots of slower CPU's as seamlessly as possible with as low power usage as possible. HT3.1 would definitely be a little long in the tooth by late 2016/early 2017 for a Zen launch but it'd probably still be good enough to do the job for a 4 socket system. Especially if they're putting a decent number of PCIe lanes on die too for each socket. \/\/\/\/\/\/\/\/edit: thanks. PC LOAD LETTER fucked around with this message at 15:42 on May 8, 2015 |
|
#
?
May 8, 2015 03:01
|
|
|
PC LOAD LETTER posted:Never heard of that, do you have a link with a story about it or just personal experience? I'd assume it'd be due to implementation (manufacturing) issues rather than something wrong with the design itself per se. Comments from Verizon guys who had to deal with it.
|
|
#
?
May 8, 2015 14:31
|
|
|
If you haven't already, you should listen to Scott Wasson (techreport.com) and David Kanter (realworldtech.com) talk AMD. The front end is predominantly CPU, of what I have heard so far (I'm listening to it right now as we speak) it then branches into HBM, and then GPUs. I was very happy to hear these two pick eachother's brains. EDIT: Herp derp, link: https://www.youtube.com/watch?v=9RA6t3Mx0ms SwissArmyDruid fucked around with this message at 03:22 on May 9, 2015 |
|
#
?
May 9, 2015 01:06
|
|
|
SwissArmyDruid posted:If you haven't already, you should listen to Scott Wasson (techreport.com) and David Kanter (realworldtech.com) talk AMD. The front end is predominantly CPU, of what I have heard so far (I'm listening to it right now as we speak) it then branches into HBM, and then GPUs. I was very happy to hear these two pick eachother's brains. Got a link?
|
|
#
?
May 9, 2015 02:10
|
|
|
VostokProgram posted:Got a link? http://techreport.com/review/28246/the-tr-podcast-bonus-video-amd-zen-fiji-and-more
|
|
#
?
May 9, 2015 02:17
|
|
|
Herp derp, I copied the link into my clipboard and everything, just forgot to hit ctrl + v.
|
|
#
?
May 9, 2015 03:23
|
|
|
http://www.tomshardware.com/reviews/intel-core-i7-5775c-i5-5675c-broadwell,4169-6.html In case someone was considering an A10 for a media/light gaming box, Intel just airdropped the Broadwell i5-5675C into AMD's last remaining base because why not.
|
|
#
?
Jun 2, 2015 12:40
|
|
|
It's about twice the price though isn't it?
|
|
#
?
Jun 2, 2015 13:43
|
|
|
Dead Goon posted:It's about twice the price though isn't it? The 7850K has a MSRP of $175 and the 5675C has a reported MSRP of $250. In TomsHardware's tests, the 5675C got 93% higher FPS in GTAV. Seems like a worthy upgrade, even ignoring the massive CPU performance delta between the two. The 7850K has been on somewhat of a firesale lately however, it's only $125 on Amazon.
|
|
#
?
Jun 2, 2015 13:51
|
|
|
For a 93% performance increase, I'd pay the premium. AMD needs a miracle.
|
|
#
?
Jun 2, 2015 14:12
|
|
|
BurritoJustice posted:The 7850K has a MSRP of $175 and the 5675C has a reported MSRP of $250. In TomsHardware's tests, the 5675C got 93% higher FPS in GTAV. Seems like a worthy upgrade, even ignoring the massive CPU performance delta between the two. Oh, not too much difference in price then and obviously the performance blows AMD out of the water.
|
|
#
?
Jun 2, 2015 14:47
|
|
|
Angry Fish posted:For a 93% performance increase, I'd pay the premium. AMD needs a miracle. AMD should just do a Kickstarter for Zen at this point. I'd toss in $100. They just need $1 billion or so, right?
|
|
#
?
Jun 2, 2015 15:52
|
|
|
sincx posted:AMD should just do a Kickstarter for Zen at this point. I really think we won't have an independent x86 competitor for Intel in the next decade, and that is a shame.
|
|
#
?
Jun 2, 2015 16:54
|
|
|
AND they did a thing that clarified that if AMD gets acquired, the x86 licensing agreement gets poofed.
|
|
#
?
Jun 2, 2015 16:57
|
|
|
SwissArmyDruid posted:AND they did a thing that clarified that if AMD gets acquired, the x86 licensing agreement gets poofed. Doesn't Intel want to avoid anti-trust suits?!
|
|
#
?
Jun 2, 2015 17:04
|
|
|
Angry Fish posted:Doesn't Intel want to avoid anti-trust suits?! I'm not sure Intel cares. Everyone else is just searching for a way to make ARM better than x86.
|
|
#
?
Jun 2, 2015 18:09
|
|
|
Doesn't intel need the amd64 license?
|
|
#
?
Jun 2, 2015 18:52
|
|
|
FaustianQ posted:I'm not sure Intel cares. Everyone else is just searching for a way to make ARM better than x86. "Let's find a way to make this small subcompact Daewoo from 1999 that I bought for $800 on craigslist back in 2003 tow and haul the same loads as my uncle's $60k F-450 turbo-diesel." -- The Industry? I never understood the ARM thing. ARM is cheap to license, but the whole architecture was built in the 90s and early 'aughts as a really inexpensive embedded chip. I know the x86 is also long in the tooth, but at least it was intended for personal computers.
|
|
#
?
Jun 2, 2015 19:17
|
|
|
I'm not sure where you get that idea at all. Please look up the history of Acorn computing and ARM. The appeal of ARM is that it's an easy-to-understand, low power RISC ISA that scales out to a zillion cores really easily. A lot of the server world is running on stuff that's very memory-intensive but has low CPU requirements, and they're trying to pack density in without overscheduling interrupts on CPUs, which ARM does really well at. When RISC died 10-ish years ago because the existing vendors didn't see the writing on the wall, Alpha, SPARC, and POWER (less so MIPS) were incredibly expensive without performance to warrant it (in many common business use cases). x86 won because it was cheap enough and performed well enough. Now AMD and Intel are busy fighting over whether Intel can take away the rest of AMD's market share with $200 CPUs while the rest of world is saying "hey, if they can shove one of these CPUs into a Chromebook and sell the whole thing for $200, how much would a CPU+memory+board that runs most of my stuff at an acceptable speed cost? $30?" And they don't have to worry about licensing a ton of poo poo from Intel or AMD with cross-patent bullshit to do it. Whether or not ARM actually outperforms x86 at similar TDP is the big question everyone's after, and Intel's sort of proving here and there that massively power-gated x86 can come close to beating ARM at the embedded game, but performance isn't the question you should be looking at. It's price/performance.
|
|
#
?
Jun 2, 2015 19:44
|
|
|
adorai posted:Doesn't intel need the amd64 license? They already have it, and AMD or a successor can't "take it back" as it were.
|
|
#
?
Jun 2, 2015 21:13
|
|
|
Nintendo Kid posted:They already have it, and AMD or a successor can't "take it back" as it were.
|
|
#
?
Jun 2, 2015 21:45
|
|
|
evol262 posted:I'm not sure where you get that idea at all. Please look up the history of Acorn computing and ARM. In theory, couldn't POWER replace x86-64? But yea, I worded it wrong and you explained it better, ARM is the future because AMD is being awful and Intel are dicks and the hope is that ARM eventually gives 90% of the performance for 50% of the cost and 10% of the hassle.
|
|
#
?
Jun 2, 2015 21:51
|
|
|
adorai posted:So you are telling me that the contracts are written such that Intel can revoke the x86 license from someone who buys AMD, but someone who buys AMD cannot then revoke the x86_64 license? That was some short sighted poo poo right there. It is probably written so that if Intel is bought they lose the x86_64 license and AMD keep the x86 licensing.
|
|
#
?
Jun 2, 2015 22:12
|
|
|
Varkk posted:It is probably written so that if Intel is bought they lose the x86_64 license and AMD keep the x86 licensing. Hang, what about irrelevant but still existing VIA? They have x86-64 chips, how does it work for them?
|
|
#
?
Jun 2, 2015 22:18
|
|
|
Varkk posted:It is probably written so that if Intel is bought they lose the x86_64 license and AMD keep the x86 licensing. That's the gist of it from what I can understand. Kitguru has some statements and corrections by AMD employees, and the cross-license agreement can be found here. Even beyond that, who'd want to compete with Intel in the x86-64 space?
|
|
#
?
Jun 2, 2015 22:51
|
|
|
Samsung and Nvidia? Maybe Qualcomm? As above, VIA is still floating about but I have no idea how they get to use x86-64 so it could be moot.
|
|
#
?
Jun 2, 2015 23:13
|
|
|
FaustianQ posted:VIA is still floating about but I have no idea how they get to use x86-64 Nobody has noticed yet.
|
|
#
?
Jun 2, 2015 23:23
|
|
|
FaustianQ posted:Samsung and Nvidia? Maybe Qualcomm? As above, VIA is still floating about but I have no idea how they get to use x86-64 so it could be moot. VIA had patents which they used as leverage to obtain an x86 license. Under the terms of the agreement they can't make chips compatible with Intel sockets.
|
|
#
?
Jun 2, 2015 23:45
|
|
|
Rastor posted:VIA had patents which they used as leverage to obtain an x86 license. Under the terms of the agreement they can't make chips compatible with Intel sockets. So could a company gobble up VIA and actually do something with x86 on their own socket or will Intel cry foul? Do they also have an AMD64 license, and same rules apply? Because if so the reason I could hazard it hasn't been done is because no one really cares or thinks it a wasted effort. Maybe the death of AMD could force Intel to make x86-64 public like ARM? I have no idea how litigation could make that work but it'd be interesting as hell to suddenly see if ARM could survive that.
|
|
#
?
Jun 2, 2015 23:51
|
|
|
ARM isn't really public either. It's a huge mess, and the ISA isn't free (but at least it's only $1m or something), but aarch64/armv8-a and armhfp/v7 are finally sane enough for widespread general-purpose OS support. Same for ppc64le, if IBM can not be IBM and price it reasonably. The hardware is incredible. So is the price.
|
|
#
?
Jun 3, 2015 00:10
|
|
|
evol262 posted:ARM isn't really public either. It's a huge mess, and the ISA isn't free (but at least it's only $1m or something), but aarch64/armv8-a and armhfp/v7 are finally sane enough for widespread general-purpose OS support. It's IBM, of loving course not. I mean, from what I understand POWER could supplant x86-64 and maybe ARM but I guess gently caress that noise?
|
|
#
?
Jun 3, 2015 00:16
|
|
|
FaustianQ posted:Samsung and Nvidia? Maybe Qualcomm? As above, VIA is still floating about but I have no idea how they get to use x86-64 so it could be moot. How much money does desktop processors make? From what I gathered, there's bigger money in the mobile, server, and HPC markets. Even AMD said they'd concentrate on the server/HPC market once their chips start performing. Samsung and Qualcomm are already making bids for the low-power/mobile segment with their ARM SoCs, and NVIDIA already has their Tesla (co-processors but still) and Tegra lines for HPC and mobile markets. It seems like it'd be easier and less expensive to just chip away at Intel's market shares in the notebook/mobile market given recent trends than to directly compete with them. VIA entered into a cross-license agreement with Intel that's basically over at this point. VIA's claim to x86(-64) came from Cyrix who reverse engineered Intel's i486 and, through a series of legal disputes, got a cross-license agreement with Intel until they bought out.
|
|
#
?
Jun 3, 2015 00:21
|
|
|
evol262 posted:ARM isn't really public either. It's a huge mess, and the ISA isn't free (but at least it's only $1m or something), but aarch64/armv8-a and armhfp/v7 are finally sane enough for widespread general-purpose OS support. I really wish our upgrade schedule at work had lined up with Power 8 availability, but instead I'm still on old Power7+  You're right about the price though, without storage included I think our current machine ran about 400k $ or so (Power 770). On the other hand, it's not like Power chips are at all useful in the traditional sense for normal desktops or gaming type loads. Current chips are all about huge massive thread loads with slightly worse than haswell single threaded performance.
|
|
#
?
Jun 3, 2015 00:35
|
|
|
P8 is a lot cheaper and about 60% faster in single-threaded, but it's still $2k minimum for a dev kit, I think
|
|
#
?
Jun 3, 2015 01:23
|
|
|
I don't know if the WCCFtech link was posted already but Carizzo looks to be a 10-15% IPC improvement compared to previous iteration steamroller. Also better power draw and thermals, but this is definitely too little to late. If Carrizo had been Bulldozer than maybe AMD wouldn't be in hotwater now.
|
|
#
?
Jun 3, 2015 02:35
|
|
|
FaustianQ posted:I don't know if the WCCFtech link was posted already but Carizzo looks to be a 10-15% IPC improvement compared to previous iteration steamroller. Also better power draw and thermals, but this is definitely too little to late. If Carrizo had been Bulldozer than maybe AMD wouldn't be in hotwater now. So at a 15% IPC improvement, what Intel processor family has the same performance at the same clocks?
|
|
#
?
Jun 3, 2015 03:31
|
|
|
|
| # ? Apr 25, 2024 18:48 |
|
|
Nintendo Kid posted:So at a 15% IPC improvement, what Intel processor family has the same performance at the same clocks? IIRC, there is currently a 40% IPC gap between Haswell and AMD CPUs, 15% of that would be 25%, against an SB CPU it'd be more 15% behind. Soo...uh, maybe low end i5 SB?
|
|
#
?
Jun 3, 2015 03:37
|
|