|
Which brings into question the reason that Intel have been leading in performance the last few years. Could they have been so cynical as to move security down the list of priorities in favour of performance, in order to stay ahead?
|
# ? May 16, 2019 07:16 |
|
|
# ? May 31, 2024 10:18 |
|
Do a comparison between Intel and Boeing. It's not a pretty picture. I name these two companies because I am familiar with them. I'm sure there are other enormous companies that are on a similar trajectory.
|
# ? May 16, 2019 07:41 |
|
Lookin' forward to the Zen2 benchmarks where the 9900k figures are measured with HT off to reflect real world use
|
# ? May 16, 2019 08:18 |
|
Intel probably about to drop the best i7 seen in years
|
# ? May 16, 2019 12:00 |
|
new segmentation strategy: i7 and i9 both have hyperthreading but only i9 can turn it off
|
# ? May 16, 2019 12:40 |
|
New Intel CPUs come with a manual jumper to disable SMT.
|
# ? May 16, 2019 12:57 |
|
i dont think current model i7 even has HT i wasnt actually making a joke about that- i bet you the next i7 we see is the best one in years
|
# ? May 16, 2019 13:15 |
|
Sure, in absolute pure single-threaded performance. While on the other hand, AMD decides to go 4-way SMT. (I'm not making that up, right? I did hear something like that?)
|
# ? May 16, 2019 13:55 |
|
SwissArmyDruid posted:Sure, in absolute pure single-threaded performance. 4 way smt might be an eypc only thing, 2 way is still ryzen but till computex gets here we are basically pissing in the wind on what they will actually do, Statutory Ape posted:i dont think current model i7 even has HT 9xxxx series i7 do not have HT https://ark.intel.com/content/www/us/en/ark/products/186604/intel-core-i7-9700k-processor-12m-cache-up-to-4-90-ghz.html
|
# ? May 16, 2019 14:09 |
|
Statutory Ape posted:Intel probably about to drop the best i7 seen in years it will be the hottest at least
|
# ? May 16, 2019 14:13 |
|
apropos man posted:Which brings into question the reason that Intel have been leading in performance the last few years. Could they have been so cynical as to move security down the list of priorities in favour of performance, in order to stay ahead? Seems much more likely that they spent decades not thinking about security at the microcode level and were blindsided by recent advances. *extreme parody of a valley girl voice form the 90s* "Uhhh, like, why would I worry about application layer security, Becky? I make hardware."
|
# ? May 16, 2019 14:29 |
|
SwissArmyDruid posted:Sure, in absolute pure single-threaded performance. I think the rumor is 4-way SMT will be the big feature for Zen 3/EPYC 3. I doubt we will get it on the mainstream/consumer platforms since > 2-way SMT has a much smaller benefit when you're not a server dealing with tons of threads in an I/O- or memory-bound situation. I wonder if they're waiting for this to drop before updating Threadripper so they can use it to create additional market segmentation between the EPYC and TR lines. Or, OTOH, I could totally see AMD do 4-way SMT across their entire product stack just as a big middle finger to Intel. Mr.Radar fucked around with this message at 15:24 on May 16, 2019 |
# ? May 16, 2019 15:20 |
|
Mr.Radar posted:I think the rumor is 4-way SMT will be the big feature for Zen 3/EPYC 3. I doubt we will get it on the mainstream/consumer platforms since > 2-way SMT has a much smaller benefit when you're not a server dealing with tons of threads in an I/O- or memory-bound situation. I wonder if they're waiting for this to drop before updating Threadripper so they can use it to create additional market segmentation between the EPYC and TR lines. Or, OTOH, I could totally see AMD do 4-way SMT across their entire product stack just as a big middle finger to Intel. As you mentioned, 4-way SMT is unlikely to help client workloads, so I don't really get the point. There's also a risk of it outright hurting games.
|
# ? May 16, 2019 15:28 |
|
LRADIKAL posted:Do a comparison between Intel and Boeing. It's not a pretty picture. I name these two companies because I am familiar with them. I'm sure there are other enormous companies that are on a similar trajectory. We all know just how well high-level corporate decision makers are held accountable for their misdeeds and failures, especially in US of A. AKA not at all
|
# ? May 16, 2019 15:57 |
|
apropos man posted:Which brings into question the reason that Intel have been leading in performance the last few years. Could they have been so cynical as to move security down the list of priorities in favour of performance, in order to stay ahead? Yes
|
# ? May 16, 2019 16:42 |
|
BangersInMyKnickers posted:it will be the hottest at least
|
# ? May 16, 2019 18:12 |
|
Munkeymon posted:Seems much more likely that they spent decades not thinking about security at the microcode level and were blindsided by recent advances. I read an article back a while back that said essentially that there's a well known and studied list of vulns which chip makers were expected to design around, but Intel didn't because of the performance cost. AMD did try, which is why it wasn't hit with as many of these issues.
|
# ? May 16, 2019 23:00 |
|
I'm getting impatient.. Lots of sales on old Ryzen right now. Is X570 going to be a major upgrade over X470? Microcenter has Ryzen 1600 for $80 right now. I'm thinking about buying that plus an X470 board, then swap the cpu with Zen2 when it comes out.
|
# ? May 17, 2019 03:35 |
|
Alpha Mayo posted:Is X570 going to be a major upgrade over X470? Biostar thankfully answered that for us. Unless you want PCIe 4 for futureproofing (which you probably shouldnt care about honestly), it's not really an upgrade at all. For AMD, it's more about taking things in house. e: I should add that it may support other Zen2 specific features that haven't been announced, if you are thinking of upgrading someday.
|
# ? May 17, 2019 04:55 |
|
Alpha Mayo posted:I'm getting impatient.. Lots of sales on old Ryzen right now. If I had a Microcenter close by I'm sure I'd have a 1600 in my server. I almost bit the other day when Newegg had a similar deal with an ASRock motherboard. That said, that deal has been going for months and I can't imagine it's going to go away once Zen 2 is announced. I personally think holding off two weeks would be worth it just to confirm that 12c/16c Zen 2 parts will work in an X470 mobo. Also, if pricing on the lower end parts is as aggressive as some have speculated, a 6 core Zen 2 CPU might not be that much more expensive than the 1600.
|
# ? May 17, 2019 05:59 |
|
I'd just like to say, today is my last day at an Intel based computer board manufacturer and LOfuckingL they're going to have a very rough go of it if they stay on the Intel roadmap. Only having 48 lanes of PCIe per proc in servers is absolutely nuking Intel in the HPC space.
|
# ? May 17, 2019 08:51 |
|
Meanwhile AMD is stuffing "bonus" PCIe lanes into their server platform.
|
# ? May 17, 2019 09:27 |
|
Mr.Radar posted:Meanwhile AMD is stuffing "bonus" PCIe lanes into their server platform. "Bonus"? Are PCIe lanes the prize that we find at the bottom of a box of crackerjacks, then? XD Look, I just hope that this improved focus on scavenging PCIe lanes on things that just sit there unused means that IOMMU groupings down the motherboard product stacks get better and more sensibly arranged overall.
|
# ? May 17, 2019 10:11 |
|
AMD is already the first company to fully embrace the lootbox model for marketing purposes, like the Phenom II X3 softlock, Ryzen 1600 8C8T, R9 290 soft-lock, RX 470/570 memory clockrate, RX 530 DDR3 variant, Vega 56 Samsung, and innumerable others. Way ahead of their time. I see people encouraging a slot-machine roll on the V56 all the goddamn time. Surely it will overclock with a VBIOS flash, after all you have at least a 1:3 chance. Let's be honest, at least NVIDIA is honest about what you're buying. Paul MaudDib fucked around with this message at 10:49 on May 17, 2019 |
# ? May 17, 2019 10:44 |
|
honest* *GTX 970 excluded
|
# ? May 17, 2019 10:49 |
|
Didnt the 970 end up being a decent value
|
# ? May 17, 2019 10:51 |
|
Please proceed, governor. What's your problem with the 970?
|
# ? May 17, 2019 10:51 |
|
Paul MaudDib posted:Let's be honest, at least NVIDIA is honest about what you're buying. Hahaha, have you ever bought a laptop with dedicated graphics before Pascal
|
# ? May 17, 2019 10:53 |
|
I used to have a Gigabyte laptop with a 860M. The Maxwell version. But the same designation (860M) was also used for a Kepler version.
|
# ? May 17, 2019 10:54 |
|
Arzachel posted:Hahaha, have you ever bought a laptop with dedicated graphics before Pascal Yes. I bought a thinkpad with a FX 880M and did a bunch of CUDA development work on it. It was also an actual good gaming processor, I could play TF2 on it in the post-phenom era when AMD was struggling with the architecture that shall not be named. So, in comparison, are we talking about AMD where they are still hawking GCN 1.0 in 2019? Actual first-gen GCN garbage that doesn't even support FreeSync? (and that's not even their mobile lineup, that's desktop, lol)
|
# ? May 17, 2019 10:59 |
|
Lambert posted:honest* The mx150/gt1030 clockrate fiasco is a better example imho
|
# ? May 17, 2019 11:07 |
|
Statutory Ape posted:The mx150/gt1030 clockrate fiasco is a better example imho Clockrate wasn't a big deal. People remember the 1030 for the DDR3 switchout. it would be a better example if AMD hadn't put out their own DDR3 versions of their DDR5 chips too. Or versions of the 470/570 with clockrates all over the board. If you've been doing GPUs for a while, janky-RAM chips have been a thing since literally forever. SE never stands for "special edition", it stands for "lovely edition". Understanding what a GT/s means is nothing too different.
|
# ? May 17, 2019 11:10 |
|
So what is this then Currently there are two variants of the MX150, a "normal" N17S-G1-A1 with 1468 - 1531 MHz and 25 Watt TDP and a power efficient variant named N17S-LG-A1 with 936 - 1037 MHz (-32%) and 10 Watt TDP (-60%) for thin and light laptops. Which admittedly is specific to mx150, but that is what I meant They didnt bother making people aware that they were in effect 2 vastly different cards with one name E: and nvidia spec sheet doesnt mention ddr3 on a gt1030 that I can find with a quick search, what do you have on that for me? Worf fucked around with this message at 11:22 on May 17, 2019 |
# ? May 17, 2019 11:20 |
|
Are they a different core count? 1060 6 GB vs 1060 3 GB is deceptive. Memory bandwidth or TDP limits, you have to understand what you're looking at. Just like buying the DDR3 model of that RX 500 series chip. Statutory Ape posted:E: and nvidia spec sheet doesnt mention ddr3 on a gt1030 that I can find with a quick search, what do you have on that for me? OK I got NVIDIA and AMD mixed up, NVIDIA had DDR4 and AMD uses DDR3. Paul MaudDib fucked around with this message at 11:26 on May 17, 2019 |
# ? May 17, 2019 11:21 |
|
Paul MaudDib posted:AMD is already the first company to fully embrace the lootbox model for marketing purposes, like the Phenom II X3 softlock, Ryzen 1600 8C8T, R9 290 soft-lock, RX 470/570 memory clockrate, RX 530 DDR3 variant, Vega 56 Samsung, and innumerable others. Way ahead of their time. I see people encouraging a slot-machine roll on the V56 all the goddamn time. Surely it will overclock with a VBIOS flash, after all you have at least a 1:3 chance. “Our yields are great, and we know we can’t sell as many units of this great processor. Let’s throw it into the garbage instead of selling it to our consumers as something better than they expected. They’d hate that.” That’s not loot boxing, Paul. Just because Intel has bad news come up doesn’t warrant an attack against the only company that is trying to push the bar forward.
|
# ? May 17, 2019 11:26 |
|
bobfather posted:Dude, every time Intel has a bit of trouble you’re always Thanosing in here to Oof, who said anything about Intel? Are you really that defensive that you have to deflect from literally GPUs over to CPUs? And what is "thanos"ing anyway? I haven't seen AMD push the market forward since Hawaii, tbh. Everything since then has been reactive. When was the last time AMD released a card that was 10% cheaper and 20% faster than NVIDIA? Uh, the 290X. AMD has been """competitive""" against NVIDIA since then, but nothing as daring as they've been doing in the CPU markets. They've never pushed forward, just managed to keep up with NVIDIA's price cuts. NVIDIA has been leading the market since then. AMD just can't afford to be competitive anymore. NVIDIA puts DDR4 on a 2018 Pascal card, AMD puts DDR3 on a 2012 GCN 1.0 card. Pretty much the same thing right? Paul MaudDib fucked around with this message at 11:39 on May 17, 2019 |
# ? May 17, 2019 11:29 |
|
Just give it a rest dude.
|
# ? May 17, 2019 11:44 |
|
Paul MaudDib posted:Oof, who said anything about Intel? Are you really that defensive that you have to deflect from literally GPUs over to CPUs? And what is "thanos"ing anyway? I’m mentioning CPUs because you mentioned the Phenom and R1600 as being loot boxes. I’m mentioning Intel because you have this beautiful habit of lying dormant in the Intel thread until another bad news cycle, after which you reliably emerge, downplay the problem, then move to the AMD thread and try to balance the scales. Last, and be truthful: What would Intel’s current product line resemble without AMD’s pressure? I mean, what would the core counts on i7’s be? Would there be an i9? How about mobile processor core counts?
|
# ? May 17, 2019 11:45 |
|
bobfather posted:I’m mentioning CPUs because you mentioned the Phenom and R1600 as being loot boxes. I actually think you may not be a regular here. This is not my only recent post in the Intel thread and it's not my only recent post in this thread either. I think you need to chill with the personal posting here. You disagree with me, I get it, gonna have to deal with it. Intel's product lineup minus Ryzen would probably have been Coffee Lake 6C in 2017. Without a process improvement there wouldn't have been a 4c8t 8700K anyway. "literally the same as last year" is not a saleable improvement. The 7700K was still better than the 6700k, and the 6700K was still better than the 4790K. What do you think Intel would have marketed in 2017? "literally nothing over last year but it's one better"? Tapeouts don't happen in oh-poo poo moments. Products take 2-3 years to bring to market absolute minimum, even when the architecture is 100% hammered out. It was probably on the Intel roadmap a long time before bobfather ever heard the word "ryzen". Paul MaudDib fucked around with this message at 11:58 on May 17, 2019 |
# ? May 17, 2019 11:51 |
|
|
# ? May 31, 2024 10:18 |
|
There would be an i9 except we would call it the i7 like we have been for years
|
# ? May 17, 2019 11:52 |