|
Ok looking at it a bit more the 10700k actually seems like a pretty good value. Load power isn't really a huge concern (although it's pretty silly) but yeah at this point I'd definitely wait to see the next Ryzens before doing anything. No reason to get any of these unless you're really desperate for something.
|
# ? May 20, 2020 22:15 |
|
|
# ? Apr 24, 2024 16:49 |
|
https://www.youtube.com/watch?v=iQVBlCfb72M&t=215s nice loving copy! *honk honk* Intel CPU and Platform Discussion: cock speed Anime Schoolgirl fucked around with this message at 03:04 on May 21, 2020 |
# ? May 21, 2020 02:58 |
|
Anime Schoolgirl posted:
^^^
|
# ? May 21, 2020 03:19 |
|
Anime Schoolgirl posted:https://www.youtube.com/watch?v=iQVBlCfb72M&t=215s nice loving copy! *honk honk*
|
# ? May 21, 2020 06:07 |
|
Budzilla posted:It's amazing that people keep saying 10nm will be out in 18 months or so after how many years of delay? It's been a disaster for Intel and the only time I will believe on a ready date for 10nm desktop is when I can go out and buy one. has it ever been reported what happened that made the 10nm transition such a cluster?
|
# ? May 21, 2020 16:06 |
|
Someone upthread indicated that it was intel’s choice of materials to use that gave them problems. Obviously, it didn’t behave as well as they had hoped.
|
# ? May 21, 2020 16:18 |
|
Everything from actual technical reasons and decisions to delicious office politics is what I've heard, the later probably enabling and exacerbating the former
|
# ? May 21, 2020 16:56 |
|
WhyteRyce posted:Everything from actual technical reasons and decisions to delicious office politics is what I've heard, the later probably enabling and exacerbating the former Management never EVER wants to eat vegetables. I assume the 10nm cluster went something like this: Engineer: Hey boss, this isn't going to work, we need to scrap this and start over, it will take a year to fix. Management: LOL no. Let's fix what we have instead of scrapping and rebuilding. 5 years later it's still not fixed. Fin.
|
# ? May 21, 2020 18:39 |
|
WhyteRyce posted:Everything from actual technical reasons and decisions to delicious office politics is what I've heard, the later probably enabling and exacerbating the former We'll probably never know for sure unless someone inside Intel leaks a post mortem analysis - Assuming management even calls for one. Wasn't Intel's CEO ratbert90 posted:Management never EVER wants to eat vegetables. I assume the 10nm cluster went something like this:
|
# ? May 21, 2020 19:27 |
|
i am v surprised there isn't a shareholder lawsuit against the 10nm bs. Intel knew it was poo poo for years
|
# ? May 21, 2020 19:31 |
|
Fabulousity posted:We'll probably never know for sure unless someone inside Intel leaks a post mortem analysis - Assuming management even calls for one. Wasn't Intel's CEO Is there a good take on how AMD managed to sneak up and turn off Intel’s hearing aid so dang badly despite being a fraction of the size and having their business split between at least two major product categories? Is it just a case of hiring the right people, cleaning the slate, and partnering with good helpers like TSMC who were already gaining experience going smaller for their work on ARM with companies like Apple, etc?
|
# ? May 21, 2020 19:39 |
|
Ok Comboomer posted:Is there a good take on how AMD managed to sneak up and turn off Intel’s hearing aid so dang badly despite being a fraction of the size and having their business split between at least two major product categories? Intel was very process driven. For years, process improvements had driven profitability, and the process group became dominant in management. The process team decided what the process would be, and everybody had to adapt their tools and software to that, no exception. With 10nm, the process group failed, and the management team was utterly unprepared to evaluate this, had no management steps to identify and investigate the seriousness of the issue, and had plenty of legacy ties that blinded them to actions they could have taken. Really, a case study in being too successful and the hubris that can develop. It’s almost sad that intel will never fess up, and this will not become a case study for biz schools to teach.
|
# ? May 21, 2020 19:55 |
|
Ok Comboomer posted:Is there a good take on how AMD managed to sneak up and turn off Intel’s hearing aid so dang badly despite being a fraction of the size and having their business split between at least two major product categories? it's mostly the fact that TSMC leapt ahead of literally everyone else in the world on 7nm, at the same moment Intel stried to bite off more than they could chew with multi-patterning (while also refusing to spring for the cost of EUV) and completely stalled out for 5 years. AMD didn't really have anything to do with that, it's just that right now you either use TSMC 7nm or your products are poo poo. Intel's nodes are actually comparatively good given the present market, GF 14nm and Samsung 10nm are terrible in comparison to Intel, Intel just isn't as good as TSMC 7nm. There literally is no other company that can do what TSMC does right now, they basically have a monopoly on 7nm class nodes for the moment, and in fact many of their competitors are abandoning the chase entirely and just staying on their previous generation of nodes. The one thing that AMD has done right is moving to chiplets a couple years ahead of everyone else. Being able to move the uncore off the die saves a ton of space and drastically raises yields, and chiplets let you bin parts of the chip separately instead of having to roll a natural 20 on every single core for it to turn out perfect. And Zen2 does that while suffering very little apparent performance hit, in contrast to their previous architectures. Also, it lets them throw a comparatively fat core on there in terms of cache and execution units, because the yields are so much better. (note that I am specifically referring to Matisse/Rome here, desktop Summit and Pinnacle were both monolithic chips, and the multi-chip-module HEDT/server variants of these sucked rear end for a variety of reasons. In fact even the monolithic products were pretty poo poo unless you had a perfectly threaded, latency-insensitive, non-AVX task. But right now AMD can build a HEDT processor out of chiplets that are basically smaller than phone SOCs and lose basically no performance in the process, which is the part that they have done right. The process itself is all TSMC and largely funded by Apple and Qualcomm and other phone tech companies, AMD is small potatoes who is massively riding on their coattails.) Intel is starting to hedge their bets by designing products on TSMC 7nm and that's where things get interesting. Their second-gen Xe GPUs will release either next year or the year after on a TSMC 7nm node. The problem is there is only so much TSMC 7nm capacity to go around - Intel's processor business alone could consume the entirety of TSMC's 7nm capacity, and it's not like that's different if the boxes say AMD on them either. It's one thing to be a scrappy upstart that takes over 60% of the DIY market, that's like single-digit percentages of the total market, TSMC doesn't have the capacity to supply the beige boxes and servers of the world. And like I said, there is basically nobody else left who is even pursuing 7nm-tier nodes besides Intel (and maybe IBM/Samsung?), so this is a real problem, TSMC has a monopoly and if Intel doesn't succeed then they will keep their monopoly. Paul MaudDib fucked around with this message at 20:13 on May 21, 2020 |
# ? May 21, 2020 19:56 |
|
Malcolm XML posted:i am v surprised there isn't a shareholder lawsuit against the 10nm bs. Intel knew it was poo poo for years Intel’s lawyers could point to the profit graph the last few years and end that sort of lawsuit real fast. The perception of DIY enthusiasts is not necessarily a reflection of the market reality. 10nm was a failed node, and Intel has been making more money than ever with more demand than ever. Both are true.
|
# ? May 21, 2020 19:58 |
|
Cygni posted:Intel’s lawyers could point to the profit graph the last few years and end that sort of lawsuit real fast. Pretty much this. Despite being on 14nm+++++++++++ because 10nm is still a clusterfuck, until Zen2 released in summer 2019, Intel still had no real competition, despite having stalled for years. Even with Zen2 crushing it, Intel has a >80% market share in personal computers, and still has a 95% market share for servers. Intel set multiple revenue records in 2019 thanks to that.
|
# ? May 21, 2020 21:05 |
|
Paul MaudDib posted:And like I said, there is basically nobody else left who is even pursuing 7nm-tier nodes besides Intel (and maybe IBM/Samsung?), so this is a real problem, TSMC has a monopoly and if Intel doesn't succeed then they will keep their monopoly.
|
# ? May 21, 2020 21:26 |
|
Paul MaudDib posted:it's mostly the fact that TSMC leapt ahead of literally everyone else in the world on 7nm, at the same moment Intel stried to bite off more than they could chew with multi-patterning (while also refusing to spring for the cost of EUV) and completely stalled out for 5 years. AMD didn't really have anything to do with that, it's just that right now you either use TSMC 7nm or your products are poo poo. Intel's nodes are actually comparatively good given the present market, GF 14nm and Samsung 10nm are terrible in comparison to Intel, Intel just isn't as good as TSMC 7nm. There literally is no other company that can do what TSMC does right now, they basically have a monopoly on 7nm class nodes for the moment, and in fact many of their competitors are abandoning the chase entirely and just staying on their previous generation of nodes. note that AMD is usually a node behind the leading edge mobile socs, which means they usually have a ton of capacity to use. those 7nm fabs aren't decommissioned once apple moves to the next node. Zen 4 is 5nm, which'll be a year after HVM of 5nm for A14 happens for a fall/winter iPhone launch nVidia also appears to have underestimated TSMC's fab advantage but they have a node's worth of architectural advantage so it didn't matter much. if intel had backported the cove cores 2 years ago amd would not be in the place it is re lawsuits: shareholder lawsuits have gone for far less than this, and stock price alone isn't enough. pissing away $billions in a broken 10nm while saying it's gonna come next tuesday is a good waste of dividends/buyback $
|
# ? May 21, 2020 21:51 |
|
Malcolm XML posted:note that AMD is usually a node behind the leading edge mobile socs, which means they usually have a ton of capacity to use. those 7nm fabs aren't decommissioned once apple moves to the next node. Zen 4 is 5nm, which'll be a year after HVM of 5nm for A14 happens for a fall/winter iPhone launch If/when/how intel loses somebody like Apple you might see some shareholder action.
|
# ? May 21, 2020 21:56 |
|
intel's MDFs are also limiting AMD in the mobile market. Hard to compete there if intel is putting cash on the hood of every processor "sold" If Zen 3/4 mobile APUs are good enough AMD might break through beyond the 2nd tier Taiwan brands to dell, lenovo et al's mainline series (not these weird other models). I suspect intel gives great deals on that + eng support for exclusivity. IIRC macbook motherboards are mostly designed by intel, but I don't know if that's actually true.
|
# ? May 21, 2020 21:57 |
|
Samsung’s having difficulty with their 7nm; NVIDIA was supposedly trying to use them for Ampere but couldn’t hit their targets and are going to pick up some of the low end Ampere parts instead. GF’s ex-IBM fabs bailed on 7nm two years ago.
|
# ? May 21, 2020 22:11 |
|
Malcolm XML posted:intel's MDFs are also limiting AMD in the mobile market. Hard to compete there if intel is putting cash on the hood of every processor "sold" This is the typical Intel playbook — inferiority of product offset by dollars. Definitely agree with the 'process hubris' explanation though, pretty spot on from looking on the outside in; I know there are some Intel goons that lurk this thread that probably have a better in.
|
# ? May 21, 2020 22:13 |
|
Malcolm XML posted:note that AMD is usually a node behind the leading edge mobile socs, which means they usually have a ton of capacity to use. those 7nm fabs aren't decommissioned once apple moves to the next node. Zen 4 is 5nm, which'll be a year after HVM of 5nm for A14 happens for a fall/winter iPhone launch regardless though, TSMC simply doesn't have the 7nm volume to serve the full x86 processor market (mobile/desktop/server). It works when AMD controls 60% of 5% of the market but Intel can't like, put Tiger Lake on 7nm and have that be a viable quantity of parts for them (though I guess some parts is better than zero). And that applies to AMD as well, there are hard limits to how much Rome and Renoir and Matisse they can push out. Even if AMD took 70% of the market today they probably couldn't fill those orders. Also there are still other companies running on those nodes too, it's not just AMD gets the whole thing. Notably NVIDIA is running Ampere on 7nm (although it's still not clear whether that includes the gaming chips as well), and Intel will probably be hitting it no later than the end of next year for DG2 production. What I'm not clear on is how much of the pipeline that various "similar" nodes share. NVIDIA is supposedly using N7 with some N7P cells as well, does that all run on the same machines? How about N7+ or N5, do they share base layer equipment or does TSMC have complete duplicates so they can run both lines in parallel at full speed?
|
# ? May 21, 2020 22:26 |
|
Paul MaudDib posted:
I think only tsmc can answer that, and they aren't going to let that kind of juicy competitive info leak out
|
# ? May 21, 2020 23:31 |
|
Malcolm XML posted:intel's MDFs are also limiting AMD in the mobile market. Hard to compete there if intel is putting cash on the hood of every processor "sold" Pretty much all intel bearing laptop motherboards are designed by intel, and for the major oems, made by intel as well. The co-marketing dollars ( eg: ultrabook branding ) can be pretty significant. It’s not just Apple, it’s dell, Lenovo, HP, all the way down. You’ll get to things like MSI or Asus branded notebooks before you see in house motherboard designs, and even then it’ll be a mix with intel designs.
|
# ? May 22, 2020 00:12 |
|
Ok Comboomer posted:If/when/how intel loses somebody like Apple you might see some shareholder action. What % of Intel's overall revenue comes from Apple
|
# ? May 22, 2020 00:12 |
|
Somewhere between "enough" and "the blow to Intel's perceived market position as a more boutique vendor than AMD, thus allowing them to continue to charge a premium over the competition despite marginal (+/- ~5%~10% gains in certain single-threaded workloads) cannot be enumerated".
|
# ? May 22, 2020 00:39 |
|
WhyteRyce posted:What % of Intel's overall revenue comes from Apple Enough that Apple gets arguably the best, most preferential treatment out of Intel’s hardware partners. Apple has custom versions of CPUs made just for them with their own model numbers. Apple has also historically gotten preferential treatment with timing/rollout and with binning, getting first dibs on chips weeks before other OEMs. Also Apple helped Intel develop and rollout Light Peak/Thunderbolt and in turn Intel helped Apple develop the mainstream revised MacBook Air so there’s a history of partnership there. I didn’t know that Intel designed the MOBOs that went into Macs. Does that include the T2 bus or the fancy MPX/PCIe connector on the Mac Pro (ie the weird proprietary/ARM poo poo)? If people knew that/that is really the case then maybe Apple doesn’t entirely deserve all of the abuse they’ve gotten over thermals over the last few years. trilobite terror fucked around with this message at 00:59 on May 22, 2020 |
# ? May 22, 2020 00:56 |
|
EoRaptor posted:Pretty much all intel bearing laptop motherboards are designed by intel, and for the major oems, made by intel as well. The co-marketing dollars ( eg: ultrabook branding ) can be pretty significant. It’s not just Apple, it’s dell, Lenovo, HP, all the way down. You’ll get to things like MSI or Asus branded notebooks before you see in house motherboard designs, and even then it’ll be a mix with intel designs. It seems like Dell does their own mobos to me; they’re definitely rolling their own UEFI firmware on top of the core source drop from whichever vendor they went with. I imagine there is an Intel CRB that all those guys crib from to make their own designs.
|
# ? May 22, 2020 01:18 |
|
Apple certainty is a check mark for halo products and mindshare. And they certainly are great at moving and keeping the premium market a thing. But now you're getting into nebulous intangible arguments. And people are getting real fast and loose with this concept of shareholder lawsuit and damages when they have no skin in the game or idea about actual financial information. Intel has a myriad of poo poo they need to fix but there is a reason why they continue to crank out money and still have huge numbers of design wins despite AMD being an extremely compelling if not better competitor right now WhyteRyce fucked around with this message at 01:37 on May 22, 2020 |
# ? May 22, 2020 01:29 |
|
Shaocaholica posted:I still have a loaded Dell Precision 670 workstation with 2x dual core pentium4 Xeons. That's 4x P4 cores in one system its amazing.
|
# ? May 22, 2020 01:33 |
|
movax posted:It seems like Dell does their own mobos to me; they’re definitely rolling their own UEFI firmware on top of the core source drop from whichever vendor they went with. I imagine there is an Intel CRB that all those guys crib from to make their own designs. Easy enough to check: if the pcb/board has square dots on it, it was designed and likely built by intel.
|
# ? May 22, 2020 02:37 |
|
EoRaptor posted:Easy enough to check: if the pcb/board has square dots on it, it was designed and likely built by intel. Are they oddly into copper thieving still in TYOOL 2020?
|
# ? May 22, 2020 06:22 |
|
Gamers Nexus looks at what impact cache and memory overclocking have on the 10600K: https://www.youtube.com/watch?v=vbHyF50m-rs With these tweaks the 10600K can match the 10900K's performance in most game tests, while still trailing significantly behind in productivity tests.
|
# ? May 22, 2020 14:30 |
I guess I can quote myself from over in the packrats thread:D. Ebdrup posted:Also, I have some potentially big news for the people who like low-power servers as much as I do.
|
|
# ? May 22, 2020 22:25 |
|
I don't have any brand loyalty either way, but that i9-10900K drawing 200+ watts "normally" and 300+ watts overclocked in that Gamers Nexus testing is some poo poo. Talk about just brute forcing your way along.
|
# ? May 22, 2020 22:31 |
|
Seems to me Intel CPUs now have the same problem as AMD GPUs, they're basically factory overclocked to within an inch of their lives, without saying so. It's not sustainable but they can just hope it keeps them competitive until a future architecture is better.
FuturePastNow fucked around with this message at 06:55 on May 23, 2020 |
# ? May 23, 2020 01:22 |
|
D. Ebdrup posted:I guess I can quote myself from over in the packrats thread: Since you seem to be tracking this stuff pretty closely, do you have any idea what the hell is going on with Tremont being so late, or why Denverton seems so rare when those C2xxx atoms were in everything (until the catastrophic hardware failure). I'm asking because it looked like Atom was heading to a bigger, wider place where it's not as far behind Core, but the volume mix of Atoms seems to have fallen.
|
# ? May 23, 2020 01:40 |
|
Twerk from Home posted:Since you seem to be tracking this stuff pretty closely, do you have any idea what the hell is going on with Tremont being so late, or why Denverton seems so rare when those C2xxx atoms were in everything (until the catastrophic hardware failure). Intel has been having capacity problems because they planned to be on 10nm years ago and 14nm is having to make up the difference. On top of generally higher sales volumes they are also shipping desktop and mobile dies that are 60% bigger than they were producing before. Intel’s solution has been to focus on the high-margin stuff at the expense of things like atom. That's also why they let AMD walk all over them in the i3 and pentium (and even i5) segments - those are high-volume low-margin parts and they frankly don't want to make them if it comes at the expense of selling an i9 or xeon. Paul MaudDib fucked around with this message at 01:53 on May 23, 2020 |
# ? May 23, 2020 01:46 |
Yep, ↑ is basically it.
|
|
# ? May 23, 2020 12:08 |
|
|
# ? Apr 24, 2024 16:49 |
|
Malcolm XML posted:I suspect intel gives great deals on that + eng support for exclusivity. IIRC macbook motherboards are mostly designed by intel, but I don't know if that's actually true. Missed this discussion when it was fresh, but I highly doubt it's true. IIRC there was something like that going on all the way back in 2004-2005, when Apple switched to Intel and needed assistance because their engineering team had been designing PowerPC Macs for a long time and wasn't familiar with x86 at all. However, much past that, Apple probably designed their own. Steve Jobs never liked farming poo poo out. These days, Apple has diverged from standard PC architecture quite a bit. Consider, for example, the Apple T1 and T2 chips. Lots of modern Mac designs (such as every MacBook Pro with a Touchbar) have one. These are Apple-designed chips loosely derived from Apple's iPhone chips. They do kitchen sink work - they drive the touchbar's display, scan the touchbar's touch sensor, implement all the security junk around the fingerprint sensor, and even act as the MacBook's SSD controller. (Apple bought some Israeli SSD controller startup several years back, so now they're doing almost all their SSD design and firmware in-house, they only rely on external vendors for the NAND flash.) So, yeah, pretty confident Intel had nothing to do with any of that. If Intel was designing motherboards for Apple, it would be very generic-PC.
|
# ? May 23, 2020 13:16 |