|
Oh man. They made those with no advertising at all. And supported by one motherboard. lol
|
#
?
Aug 10, 2016 21:35
|
|
|
|
| # ? May 14, 2024 09:17 |
|
|
Intel acquired Nervana: http://techreport.com/news/30497/intel-moves-to-achieve-deep-learning-nervana-with-acquisition Not interesting, unless you're into AMD, because then this Nervana chip looks familiar:  FuturePastNow posted:Oh man. They made those with no advertising at all. And supported by one motherboard. lol Again, good idea in theory, but the tech just wasn't there. DDR3 bottlenecks ahoy.
|
|
#
?
Aug 11, 2016 01:27
|
|
|
People are hilariously calling the death of AMD based on those AoTS benchmarks, no wonder they got pulled. Someones getting fired, and AMD needs to do some damage control IMHO. Not because the results were bad, but because that's all that will be in the public mind.
|
|
#
?
Aug 11, 2016 01:55
|
|
|
Sinestro posted:Well, just being vaguely competitive in gaming is pretty impressive for AMD, sadly. They're 23% behind the i7-4790 with a 11% clock speed advantage right now. Measured in the very unscientific environment of AotS at least, the Zen ES' performance per clock is 1.76 points/MHz compared to the average Vishera performance per clock of 1.08 points/MHz. That's 63% more performance per clock, which is... kind of loving incredible. Dollars to donuts they're hoping for the actual silicon to break 4ghz though, because they sorely need it if they want mindshare.
|
|
#
?
Aug 11, 2016 02:14
|
|
|
Anime Schoolgirl posted:That's a couple percentage points higher than their estimate of 40% over Excavator, which is 15% over Vishera. Oh, it's not objectively good, it's just looking like there's some shred of evidence that those claims might be true by any possible metric.
|
|
#
?
Aug 11, 2016 02:41
|
|
|
I mean, there's still the chance that they can get some more IPC before the end of the year, right? ....right? ..........right?!
|
|
#
?
Aug 11, 2016 02:46
|
|
|
SwissArmyDruid posted:I mean, there's still the chance that they can get some more IPC before the end of the year, right? The 1D and 2D Zen serial numbers showing huge differences is also an odd thing. Thermal throttling?
|
|
#
?
Aug 11, 2016 02:56
|
|
|
I doubt it's thermals if the 24C and 32C are hitting 2.9Ghz @ 150W and 180W TDP as stated by the original leaker. It has to be limited by what seems to ail Polaris, issues in the process greatly limiting clockspeed without ramping voltage up crazy. AMD reps have to be screaming till GloFos ears bleed at this point to fix this poo poo as they look for flaws in their libraries. 3.2/3.5/4.0Ghz for Zen would put it in an amazing place, basically sitting next to Broadwell and again, if they can keep it at 400$ it'd be a drat good buy because it's closest desktop competitor is an i7-5960X which goes for a 1000$. On the other hand, based on what AMD has done in the past, I'm leery of this going A0 (because Vishera) and them charging comparable prices to Intel; because gently caress it, why not be retarded. 400$ for something that can trade punches with i7s really attracts buyers, builds mindshare and still has plenty of margin for a 143mm˛ die. I don't know about that 2D though, it's incredibly slow even compared to AMDs current offerings, do they do desktop validation for parts destined for mobile? Like, think Zen 2.0/2.4Ghz, 35W 8C/16T for mobile? Testing with disabled functions/cores?
|
|
#
?
Aug 11, 2016 04:30
|
|
|
Anime Schoolgirl posted:No, but if the actual chips themselves can't break 3.2ghz, we have a problem. GIGAHERTZ WARS P4 vs. Athlon XP all over again 
|
|
#
?
Aug 11, 2016 18:36
|
|
|
Prescription Combs posted:GIGAHERTZ WARS That's what I was thinking, that they could go back to the equivalent Intel rating system somehow.
|
|
#
?
Aug 11, 2016 19:06
|
|
|
Prescription Combs posted:GIGAHERTZ WARS Yeah, except for the fact that AMD is behind in both clock speed and IPC
|
|
#
?
Aug 11, 2016 19:12
|
|
|
Paul MaudDib posted:Yeah, except for the fact that AMD is behind in both clock speed and IPC
|
|
#
?
Aug 11, 2016 19:18
|
|
|
I don't know if it's bupkis or not, but a lot of people are claiming AotS's CPU framerates is a calculated metric based on the idea of where the framerates would sit if it wasn't GPU bottlenecked, and it's simply loving up with the Zen chip. I honestly don't know enough to say otherwise. I'm just excited for the possibility that recommending an Intel CPU will not always be the correct answer, especially APUs. AMD better get on the NUClike ball game.
|
|
#
?
Aug 11, 2016 19:30
|
|
|
there's also the part where you have off the wall poo poo like locked i5 haswells beating a 5960x I think it's not a very good benchmark at all
|
|
#
?
Aug 11, 2016 20:57
|
|
|
FaustianQ posted:I don't know if it's bupkis or not, but a lot of people are claiming AotS's CPU framerates is a calculated metric based on the idea of where the framerates would sit if it wasn't GPU bottlenecked, and it's simply loving up with the Zen chip. I honestly don't know enough to say otherwise. Hell yes I would love the poo poo out of an AMD version of a NUC. Proper dual channel RAM with an FX-9830P at 45W TDP would probably be pretty sweet.
|
|
#
?
Aug 12, 2016 00:10
|
|
|
They've definitely worked on the idea. I've seen some prototype PCs they built that were something like a quarter of an inch thick, around 8 inches long and four inches wide. Just a weird little blade thing. Historically, AMD has been more about showing what could be done and hoping an OEM does something with it rather than offer a relatively complete concept like Intel (whose free marketing dollars also don't hurt).
|
|
#
?
Aug 12, 2016 00:19
|
|
|
So it looks like the 2.8ghz/3.2ghz engineering sample of Xen beats the pants off the 8 core 4.0/4.2ghz 8350 by a margin of ~40%, and surpasses the 3.4/3.8ghz 4670k. And yet techspot spins it negatively. http://www.techspot.com/news/65945-leaked-benchmarks-amd-zen-falls-short-intel-haswell.html The earliest Xen silicon falling just 7 fps behind a more highly clocked Haswell i-7 4790 is surely a good sign? I mean, we didn't see promising results like this from early Bulldozer silicon. And as we saw in the Phenom -> Phenom II transition, and from Bulldozer to Excavator, AMD has proven crafty at squeezing out additional performance from even the biggest stinker of a uarch, so who knows what we might see them pull off with this new Jim Keller-made hotness combined with the industry partners they have now?
|
|
#
?
Aug 15, 2016 17:23
|
|
|
Bloody Antlers posted:So it looks like the 2.8ghz/3.2ghz engineering sample of Xen beats the pants off the 8 core 4.0/4.2ghz 8350 by a margin of ~40%, and surpasses the 3.4/3.8ghz 4670k. And yet techspot spins it negatively. Ashes of the Singularity doesn't seem to benefit from a bunch of cores, so it's a very impressive result. If they spin it as a 4790 beating Zen, they could spin it as Skylake 4 core beating Broadwell-E 10 core. http://m.hardocp.com/article/2016/06/24/dx11_vs_dx12_intel_6700k_6950x_framerate_scaling/5
|
|
#
?
Aug 15, 2016 18:46
|
|
|
I think I read somewhere that Ashes benefits more from clockspeed than anything else, so I'm cautiously optimistic that Zen might be pretty good. If it can be similar across the board to the results shown with Ashes, my wallet will be ready to support AMD for the first time in 10 years.
|
|
#
?
Aug 15, 2016 19:05
|
|
|
From what I read, that AoS benchmark only uses four cores. So assuming the production chips are clocked much higher than 2.8GHz (weren't the Engineering Sample Bulldozer chips also around 2.8GHz?), Zen should be pretty good. Not exactly groundbreaking, but competitive.
|
|
#
?
Aug 15, 2016 20:39
|
|
|
that appears to not be true. This test from a French site seems to show it continuing to scale past 4  http://www.jeuxvideo.com/dossier/498894/nvidia-gtx-1070-notre-test-de-la-seconde-carte-pascal-du-fabricant/498969.htm Article was posted June 9th, who knows how much that has changed since with patches and drivers since. SwissArmyDruid fucked around with this message at 20:56 on Aug 15, 2016 |
|
#
?
Aug 15, 2016 20:54
|
|
|
This is really exciting. Going of a 2.8Ghz base ES chip, those numbers look strong and if they can pull a 4Ghz base out of that thing by some method, that thing should be within striking distance of current i7 chips if not faster. If they can not be stupid for once in the last 10 years, and put it at a fantastically good price point, that would be the swift kick in the nuts to Intel I have been hoping for, for years. Lets see a $400-500 AMD chip go head to head with a $1000+ Broadwell-E with probably near feature parity in a sense too since the -E's usually are lacking a few of the newest features of their Consumer chips by a generation or so. If Zen can have some more native stuff closer to the current Z170 series then that would make it an even more impressive competitor that Intel didn't fully see coming.. However this is AMD we are talking about.... soooo. $1000 Zen's will probably be a thing they try or something thinking they have the next Athlon 4000 Black Edition or something again.
|
|
#
?
Aug 15, 2016 22:07
|
|
|
They're probably going to have to crank the voltage to get the speed and stability they need.... again. 
SwissArmyDruid fucked around with this message at 00:44 on Aug 16, 2016 |
|
#
?
Aug 15, 2016 22:45
|
|
|
If Zen is any real competition, I give it 3-6 months before we get a mainstream socket 6-core from Intel.
|
|
#
?
Aug 15, 2016 23:28
|
|
|
FuturePastNow posted:If Zen is any real competition, I give it 3-6 months before we get a mainstream socket 6-core from Intel. This is called Skylake-X. Also 4.0Ghz is an achievable boost clock inside 95W but only if we're talking half the cores/threads. Unless the program is heavily threaded, a 4 or 6 core Zen will be superior due to higher clockspeed ceiling. Although it's not impossible AMD wouldn't offer HEDT Zen at 140W if the clock scaling is worth it.
|
|
#
?
Aug 16, 2016 01:33
|
|
|
FaustianQ posted:This is called Skylake-X. Skylake-X looks like the HEDT platform and not the mainstream platform?
|
|
#
?
Aug 16, 2016 02:40
|
|
|
The idea is that when AMD finally gets its poo poo back together, maybe Intel will stop holding back 6/8-core parts with appreciable clocks and 24 PCIe lanes from mainstream parts, and consequently, pricing.
|
|
#
?
Aug 16, 2016 02:46
|
|
|
SwissArmyDruid posted:The idea is that when AMD finally gets its poo poo back together, maybe Intel will stop holding back 6/8-core parts with appreciable clocks and 24 PCIe lanes from mainstream parts, and consequently, pricing. I'm having trouble imagining a "mainstream" use for either of those things. A core 2 duo is still good enough for 80% of people that (still) own PCs. Edit: The high end parts are probably going to be sold as server chips. SYSV Fanfic fucked around with this message at 03:06 on Aug 16, 2016 |
|
#
?
Aug 16, 2016 03:03
|
|
|
SYSV Fanfic posted:I'm having trouble imagining a "mainstream" use for either of those things. A core 2 duo is still good enough for 80% of people that (still) own PCs. Gaming + recording or streaming video using x264 for ideal video quality and compression is a giant CPU pig, and something that's becoming popular in nerd circles. You can use shadowplay, quicksync, or AMD's hardware encode, but the video quality at any given bitrate is worse. This task scales across cores well, and the popular 4 core i5/i7s aren't quite enough, even Skylake.
|
|
#
?
Aug 16, 2016 03:21
|
|
|
Prescription Combs posted:Skylake-X looks like the HEDT platform and not the mainstream platform? I might be confusing it with something else then, Kabylake-S? I know Intel has announced a "mainstream" 6 core part. SYSV Fanfic posted:I'm having trouble imagining a "mainstream" use for either of those things. A core 2 duo is still good enough for 80% of people that (still) own PCs. Ehhhh...not really? Otherwise the Athlon X4 845 would be the king of budget processors for mundane office work. I really do wish Intel would have a nice budget 4C processor though as a Xeon E/X5450 does have a nice thread/performance ratio, and it's really the reason I do suggest the Athlon X4 845 over the Celerons/Pentiums. I dunno, maybe I just have something against dual cores and my personal experience on their limitations. I'd actually like to see Intel more clearly restructure thier line up: Celeron as dual core, Pentium as dual core with hyperthreading, i3's as quads, i5s as hexacores, i7s as quad and hexacores with hyperthreading.
|
|
#
?
Aug 16, 2016 03:25
|
|
|
FaustianQ posted:I'd actually like to see Intel more clearly restructure thier line up: Celeron as dual core, Pentium as dual core with hyperthreading, i3's as quads, i5s as hexacores, i7s as quad and hexacores with hyperthreading. The problem I see about this is that cost per transistor has stopped decreasing, and Intel's non-atom architecture cores have only been getting bigger. Intel would be giving up some of their margin to be delivering more transistor chips at lower price points.
|
|
#
?
Aug 16, 2016 03:43
|
|
|
SYSV Fanfic posted:I'm having trouble imagining a "mainstream" use for either of those things. A core 2 duo is still good enough for 80% of people that (still) own PCs. A Core 2 Duo, really? They absolutely aren't, unless you think the average person also overclocks and has a very good video card. I'm typing this on a Core 2 Duo I keep around because I'm weird like that and it'll have issues just playing vines or webms, let alone browsing a picture heavy page on facebook. There's people around here who game and think they're still fine with their Core 2 Quads but they have very nice GPUs that take a lot of the brunt for games and even modern browsers (thanks to video decode and 3d acceleration they have these days) on top of their overclocking and double the cores. A core 2 duo is at minimum an 8 year old part and even those last ones were just upclocked 8.5 year old parts.
|
|
#
?
Aug 16, 2016 04:02
|
|
|
fishmech posted:A Core 2 Duo, really? They absolutely aren't, unless you think the average person also overclocks and has a very good video card. I'm typing this on a Core 2 Duo I keep around because I'm weird like that and it'll have issues just playing vines or webms, let alone browsing a picture heavy page on facebook. My parents have a core 2 duo with a gt 710 added to it ($25 on sale) for video decode/browser acceleration. If there is a free as in beer benchmark for general use I can run on it, let me know. Most of the people I know that own PCs do not game, and are quite happy with their celeron/pentium branded bay trails/braswell CPUs that have a weaker per core performance than a core 2 duo. There is a reason Intel is talking up its pivot to b2b services. SYSV Fanfic fucked around with this message at 04:39 on Aug 16, 2016 |
|
#
?
Aug 16, 2016 04:36
|
|
|
My mom works-from-home (citrix client) on a bulldozer APU with zero complaints (once I put an ssd in there)
|
|
|
#
?
Aug 16, 2016 04:55
|
|
|
My e6750 seems to handle FreeNAS fine. My T100 (running some variant of the good Atoms) can do fine, as long as I'm not opening more than one intensive page and not much in the background. For ease of use, though, you definitely "need" something better than those old chips.
|
|
#
?
Aug 16, 2016 06:54
|
|
|
I suddenly want to know how well an Athlon 64 X2 3800+ would hold up with an SSD, midrange GPU and modern AIO. Was it a chipset limitation or a thermal limitation on old Athlons? Like the idea would be to compare several systems so drop a P4E, Conroe, Yorkfield, Phenom, Opteron, etc into that chart for comparison with as many systems similarities as possible, then compare price and performance to more modern stuff. More for curiosities sake.
|
|
#
?
Aug 16, 2016 08:04
|
|
|
SYSV Fanfic posted:I'm having trouble imagining a "mainstream" use for either of those things. A core 2 duo is still good enough for 80% of people that (still) own PCs. Regardless of whether or not there is a mainstream use, it is anti-consumer to hold back good poo poo like that, whether by segmentation, binning, or by price walling.
|
|
#
?
Aug 16, 2016 14:06
|
|
|
It's worth noting a lot of things like video playback won't work well on CPUs that weak without hardware video decode acceleration, which Bay Trail and Braswell have and an E6600 or something sure as hell doesn't. Heck, the 14nm atoms have pretty drat robust GPUs on most of the SKUs that take the load off the CPU whenever possible.
|
|
#
?
Aug 16, 2016 15:36
|
|
|
SYSV Fanfic posted:My parents have a core 2 duo with a gt 710 added to it ($25 on sale) for video decode/browser acceleration. If there is a free as in beer benchmark for general use I can run on it, let me know. Most of the people I know that own PCs do not game, and are quite happy with their celeron/pentium branded bay trails/braswell CPUs that have a weaker per core performance than a core 2 duo. You put a graphics card from this year in, of course that's going to help things, and even though it's based on an "old" core design it's still one 4 years newer than the CPU and system. That's hardly a standard setup for a Core 2 Duo, which is more likely by far to have either ancient, terrible, Intel integrated graphics or low end cards from that time. You're just reinforcing how unacceptable a Core 2 Duo system is - and I'm not sure why you think gaming is relevant, the modern internet and browsers are quite intensive now that everything's gone to poorly optimized HTML5 that demands acceleration support. And no, those modern systems are nowhere near as slow as a Core 2 Duo based system.
|
|
#
?
Aug 16, 2016 15:47
|
|
|
|
| # ? May 14, 2024 09:17 |
|
|
a n3050 is only very slightly behind a core 2 duo e6600, which is really impressive once you realize that the n3050 uses less than a tenth of the power
|
|
#
?
Aug 16, 2016 17:32
|
|