|
https://www.overclock3d.net/news/cpu_mainboard/amd_reveals_their_32-core_64-thread_naples_server_platform/1 It isn't likely due to Nvidia and IBM riding the inertia wave of leprechaun gold, but I hope this pans out and AMD makes x86 a viable HPC platform, if only because it'll embarrass the gently caress out of Intel's failed attempts at doing so
|
#
?
Jan 20, 2017 09:18
|
|
|
|
| # ? May 22, 2024 18:40 |
|
|
The pessimistic side of me wants to say: If Intel couldn't break into this market then AMD has no business even entertaining the idea. Also seems like the wrong thing is bolded in that article, I mean who needs 128 pcie lanes? . ... fake edit: Turns out the bolded stuff were hyperlinks  .
|
|
#
?
Jan 20, 2017 10:10
|
|
|
It's a sound idea, really, if you have CPU and GPU on the same die you have all the bandwidth in the world between the two, and tons of PCIe lanes to the platform (without need of external chips like the DGX-1 needed) would be great if you needed more compute units It's a lot more realistic than anything Intel has pushed 
|
|
#
?
Jan 20, 2017 10:45
|
|
|
Boiled Water posted:The pessimistic side of me wants to say: If Intel couldn't break into this market then AMD has no business even entertaining the idea. NVME data servers.
|
|
#
?
Jan 20, 2017 11:31
|
|
|
https://www.chiphell.com/thread-1697525-1-1.html More Zen rumors after translation: 1. NDA lifted end of Feb, official unveiling 1st half of March. 2. CPUs are already well stocked, but launch delayed due to fixing faults in BIOS on the mobos. 3. 8C/16T only on initial launch, 4C/8T by April 4. Zen expected to be expensive. 5. High expectations set for gaming performance.
|
|
#
?
Jan 20, 2017 11:43
|
|
|
Palladium posted:https://www.chiphell.com/thread-1697525-1-1.html Reading through they're saying a 500-600$ price is to be expected which is...what everyone sane thought? Once AMD did their demos anyway. That's still leagues better than the alternative.
|
|
#
?
Jan 20, 2017 12:22
|
|
|
wargames posted:NVME data servers. This is being discussed heavily in the Intel thread and generally perceived as being a bad idea. edit: Being a bad idea does not stop anyone from implementing said bad idea.
|
|
#
?
Jan 20, 2017 12:41
|
|
|
Boiled Water posted:This is being discussed heavily in the Intel thread and generally perceived as being a bad idea. A bad idea for home users, it's an increasingly common thing in the enterprise space because they have very different requirements and absolutely can utilize all the speed.
|
|
#
?
Jan 20, 2017 16:34
|
|
|
Boiled Water posted:This is being discussed heavily in the Intel thread and generally perceived as being a bad idea. That discussion is mostly about Intel Optane which is not what we are talking about here. For fast storage in the data center, you can get stupid fast (and expensive) NVMe/SAS PCIe cards full of storage. I am looking at a solution for our database servers.
|
|
#
?
Jan 20, 2017 16:55
|
|
|
FaustianQ posted:Reading through they're saying a 500-600$ price is to be expected which is...what everyone sane thought? Once AMD did their demos anyway. That's still leagues better than the alternative. That sounds like a really great price compared to a $1000 competitor Sounds like a really poo poo price for normal people though
|
|
#
?
Jan 20, 2017 17:03
|
|
|
1gnoirents posted:That sounds like a really great price compared to a $1000 competitor they'll be released afterwards though
|
|
#
?
Jan 20, 2017 17:10
|
|
|
I suspect it also makes good sense to come out with the absolute top-line product first: AMD has been poo poo for years and they want all the initial reviews to be as positive as possible with as great benchmarks as possible. You get headlines about how you've put out a top of the line chip, you're in a much better place to wash away that stink of failure compared to people talking about being the best on the price/performance curve, even though the latter is what actually matters.
|
|
#
?
Jan 20, 2017 17:23
|
|
|
evilweasel posted:I suspect it also makes good sense to come out with the absolute top-line product first: AMD has been poo poo for years and they want all the initial reviews to be as positive as possible with as great benchmarks as possible. You get headlines about how you've put out a top of the line chip, you're in a much better place to wash away that stink of failure compared to people talking about being the best on the price/performance curve, even though the latter is what actually matters. It's also a better business since the die harders will buy your new expensive chip rather than wait a month. Much like Nvidia release things.
|
|
#
?
Jan 20, 2017 17:26
|
|
|
Palladium posted:https://www.chiphell.com/thread-1697525-1-1.html Chiphell grain of salt I thought BIOS faults would be more of a CPU issue with Zen given what we know about it????
|
|
#
?
Jan 20, 2017 18:12
|
|
|
It's kinda crazy to see AMD seemingly doing everything right. We live in strange times. e - although I suppose not rushing to market and screwing yourself is easier when Kaby is so underwhelming for the markets Zen is targeting. K8.0 fucked around with this message at 18:19 on Jan 20, 2017 |
|
#
?
Jan 20, 2017 18:15
|
|
|
Palladium posted:https://www.chiphell.com/thread-1697525-1-1.html So BIOS is on the motherboard as usual and not on the chip? Also, define "expensive", because I think AMD is going to leave money on the table. Again.
|
|
#
?
Jan 20, 2017 18:26
|
|
|
Sidesaddle Cavalry posted:Chiphell grain of salt SwissArmyDruid posted:So BIOS is on the motherboard as usual and not on the chip?
|
|
#
?
Jan 20, 2017 18:28
|
|
|
mayodreams posted:That discussion is mostly about Intel Optane which is not what we are talking about here. For fast storage in the data center, you can get stupid fast (and expensive) NVMe/SAS PCIe cards full of storage. I am looking at a solution for our database servers. Exactly. NVMe is kinda OK for fast boot times I guess but its real use case is poo poo like database servers where you are pushing crazy IOPS. Nothing in the consumer world even comes close to saturating a SATA III connection's IOPS.
|
|
#
?
Jan 21, 2017 03:56
|
|
|
SwissArmyDruid posted:because I think AMD is going to leave money on the table. Again. Nah, I actually expect the opposite. Phenom I/BD were launched at stupid prices relative to their performance. Phenom II only wasn't embarrassed by i7 920 only by DDR2 compatibility and much cheaper mobos.
|
|
#
?
Jan 21, 2017 04:38
|
|
|
Paul MaudDib posted:Exactly. NVMe is kinda OK for fast boot times I guess but its real use case is poo poo like database servers where you are pushing crazy IOPS. Nothing in the consumer world even comes close to saturating a SATA III connection's IOPS. NVME is useful for video/3d workstations, at least.
|
|
#
?
Jan 21, 2017 04:54
|
|
|
Anime Schoolgirl posted:Those were just rumors, bios is apparently on board instead of chip with this information. Won't know for sure either way until it comes out. Palladium posted:Nah, I actually expect the opposite. Phenom I/BD were launched at stupid prices relative to their performance. Phenom II only wasn't embarrassed by i7 920 only by DDR2 compatibility and much cheaper mobos. From a purely business perspective, I still think that AMD should have gone up to $750, make Intel have to move soon, and THEN come down to $500-$600. Probably wouldn't win any favors on the touchy-feely side, though, but I care less about touchy-feely and more about "please dear god make Intel innovate again".
|
|
#
?
Jan 21, 2017 05:24
|
|
|
The low price is down to AMD needing to make a strong play to overcome the reputation lead that Intel has built over the last several years. Mindshare was something AMD always struggled with even when they had demonstrably better CPUs, and even now half the people who are excited for Zen aren't actually interested in it, they just want an affordable Intel with more cores.
|
|
#
?
Jan 21, 2017 11:21
|
|
|
Do AMD have the equivalent of Intel's microcode running on their CPU's? With microcode being proprietary I'd expect that if AMD 'open sourced' their CPU code it'd add to the reasons for people to switch.
|
|
#
?
Jan 21, 2017 11:25
|
|
|
Hows AMD chips and more specifically ryzen virtualisation solution against intels? Ive been loosely following ryzen development but have not seen much talk about it.
|
|
#
?
Jan 21, 2017 11:30
|
|
|
apropos man posted:Do AMD have the equivalent of Intel's microcode running on their CPU's? What would that actually do for anyone?
|
|
#
?
Jan 21, 2017 11:43
|
|
|
ohgodwhat posted:What would that actually do for anyone? I guess some people would prefer to know what programs are running on their machine. If they aren't expert enough to dig into the source themselves they can read reviews by people who are. It's not a clincher, but another way for AMD to attract a customer base.
|
|
#
?
Jan 21, 2017 13:09
|
|
|
ConanTheLibrarian posted:The low price is down to AMD needing to make a strong play to overcome the reputation lead that Intel has built over the last several years. Mindshare was something AMD always struggled with even when they had demonstrably better CPUs, and even now half the people who are excited for Zen aren't actually interested in it, they just want an affordable Intel with more cores. When AMD is already finding it hard to move RX470s at $160 which is head and shoulders against anything NV in the past and present in this price tag, imagine the how much more harder it would be to justify new AMD platform costs against 6 years worth of perfectly capable i5/i7s. Unless 4C/8T and lower Zen is happening at firesale prices I doubt there will be many takers at the end of the day. Palladium fucked around with this message at 13:13 on Jan 21, 2017 |
|
#
?
Jan 21, 2017 13:10
|
|
|
Snakes in a can posted:Hows AMD chips and more specifically ryzen virtualisation solution against intels? Ive been loosely following ryzen development but have not seen much talk about it. Roughly similar, and if you're a pessimist about demonstrated IPC gains then it's better.
|
|
#
?
Jan 21, 2017 15:05
|
|
|
apropos man posted:Do AMD have the equivalent of Intel's microcode running on their CPU's? Yes, it's pretty much the only way to implement the modern x86-64 instruction set with necessary compatibility for old programs, without doing much more complicated hardware designs. apropos man posted:With microcode being proprietary I'd expect that if AMD 'open sourced' their CPU code it'd add to the reasons for people to switch. People who would care about that sort of thing are probably going to demand all the other parts in the system be "open source" in a similar way before they really care to buy a system over it. It would also likely cost a lot of money/time with lawyers to go over all of AMD's current CPU microcode and ensure that it's in a state where it can be legally released into a proper open source license. There may be contracted or otherwise purchased work in there from an outside company that AMD can't unilaterally release for open source use.
|
|
#
?
Jan 21, 2017 17:02
|
|
|
Is open source microcode something that even Richard M. Stallman  would care about?As I see it, it like his microwave: if it’s not immediately obvious if the device has firmware at all, then it doesn’t matter if the answer is “yes, and the source code is proprietary”. http://www.linux-mag.com/id/255/ posted:I’m not so much concerned with what goes on in some sort of appliance where you can’t load software. If you can’t copy any software onto the thing, then in some sense it doesn’t matter whether you’re allowed to copy this particular program.
|
|
#
?
Jan 21, 2017 17:30
|
|
|
Platystemon posted:Is open source microcode something that even Richard M. Stallman For a while he used a Chinese MIPS-clone based laptop specifically because it was the only one with open source firmware, now he uses an old ThinkPad with an open source replacement BIOS. He probably would go out and upgrade to something new that had open source microcode if it could also support an open source UEFI boot environment.
|
|
#
?
Jan 21, 2017 17:39
|
|
|
apropos man posted:Do AMD have the equivalent of Intel's microcode running on their CPU's? Almost every aspect of being in the computer chip business is expensive and proprietary. Applauding AMD for open-sourcing their microcode or whatever is like applauding Hitler for converting to vegetarianism.
|
|
#
?
Jan 21, 2017 18:18
|
|
|
Hitler in my CPU?!
|
|
#
?
Jan 21, 2017 18:31
|
|
|
WAR DOGS OF SOCHI posted:Hitler in my CPU?! Yes, there is Hitler in your CPU.
|
|
#
?
Jan 21, 2017 18:36
|
|
|
fishmech posted:People who would care about that sort of thing are probably going to demand all the other parts in the system be "open source" in a similar way before they really care to buy a system over it. Lol, I demand that the aluminum smelting plant layout and exact process used to make the aluminum in my computer case be open-sourced so that I have the freedom to construct my own chemical plant to refine my own aluminum
|
|
#
?
Jan 21, 2017 18:44
|
|
|
lol if your aluminum wasn’t created in open‐source stars
|
|
#
?
Jan 21, 2017 18:48
|
|
|
fishmech posted:It would also likely cost a lot of money/time with lawyers to go over all of AMD's current CPU microcode and ensure that it's in a state where it can be legally released into a proper open source license. There may be contracted or otherwise purchased work in there from an outside company that AMD can't unilaterally release for open source use. Funny thing is, you probably need to get State Department approval to release poo poo like that, because it technically counts as a munition and is subject to ITAR and technology transfer agreements. Can't have Pakistan or the DPRK getting access to AMD microcode, they could spin up a bootleg plant someplace! Glorious 240nm Opterons!
|
|
#
?
Jan 21, 2017 19:38
|
|
|
Platystemon posted:Is open source microcode something that even Richard M. Stallman https://stallman.org/intel.html quote:Don't use Intel processors newer than Core2, because they have the "management engine" back door and no one can shut it off.
|
|
#
?
Jan 21, 2017 19:42
|
|
|
Is there anything to the management engine being dodgy? I mean not that I can get out of using Core2Duo or older Intel processors.
|
|
#
?
Jan 21, 2017 19:49
|
|
|
|
| # ? May 22, 2024 18:40 |
|
|
... Boiled Water posted:Is there anything to the management engine being dodgy? I mean not that I can get out of using Core2Duo or older Intel processors. no
|
|
#
?
Jan 21, 2017 19:50
|
|