|
eames posted:4C CFL-S doesn't exist, the quadcores are all rebadged KBL-S dies on the old process. wiki chip shows it as being on 14++, as do the reviews https://www.techpowerup.com/reviews/Intel/Core_i3_8350K/ https://en.wikichip.org/wiki/intel/core_i3/i3-8350k
|
# ? Oct 16, 2017 23:02 |
|
|
# ? Jun 12, 2024 13:55 |
|
People on overclock.net found evidence in datasheets and the pinout seems identical to KBL-S, unlike the other SKUs. Even if it is CFL-S, silicon lottery reports that it clocks lower than the 7600K so if it really is 14nm++ then it's a bottom tier bin. silicon lottery posted:The 8350Ks unfortunately aren't clocking as high as the 8600K/8700K. In fact, they're clocking a little bit lower than Kaby Lake 7600Ks. http://www.overclock.net/t/1638821/coffee-lake-binning/70#post_26389321 stats should be on their site later on today eames fucked around with this message at 23:15 on Oct 16, 2017 |
# ? Oct 16, 2017 23:10 |
|
14++ actually uses bigger gates in places than 14+, not smaller. I don't know if that has any impact on clocks. 8350K launched $50 cheaper than the 7600K, though. I really don't see any reason for Intel to lie or rebrand it though, they launched the Kaby Lake-R refresh laptop chips like a week before Coffee Lake and could have easily just called the i3's KBL-R and nobody would have cared. I'm sure someone will delid one eventually (i looked around a little and couldn't find anyone who has done it, ha) and maybe stick it under a scope to see. I personally wouldn't buy one with the 8400 on this earth.
|
# ? Oct 16, 2017 23:20 |
|
Cygni posted:I really don't see any reason for Intel to lie or rebrand it though Z270/Z370
|
# ? Oct 16, 2017 23:23 |
|
I've been using my i7 920 with some noctua cooler since 2008 with the same Arctic silver 3 paste. Idle temps have been around 40C, gaming 55C, full load 65-67C as long as I remember. So at least AS3 hasn't needed a repaste yet. I probably won't look at how the paste looks as long as it keeps working. I plan to try AS3 with my new 8700k if I ever get one. I haven't bought a new tube since early 2000's and as long as it works, why change...
|
# ? Oct 16, 2017 23:39 |
|
VulgarandStupid posted:Digital Foundry finally released their i5-8400 video, showing how well it does. It's 1.5X faster than a Sandy Bridge in most benchmarks and out performs Ryzen in most games, but probably not Cinebench. Also 2666Mhz RAM is quite nice on it. So once you can get a cheaper H370/H310/B350 board, Ryzen will really need to drop their prices or will fall out of favor for most. This guy's talking about people still on Sandy/Ivy can get a 3770K with 2400 RAM. I have 1600 RAM, because everything above that was officially labelled as an overclock and that word causes me to think of system crashes and broken warranties. How big an improvement should it be, or should I just keep waiting for Ryzen refresh instead? I really like the 1700 as a nice balance between sacrificed gaming FPS for better encoder throughput when streaming, but feel I could get better bang for buck if I can hold out five more months. I'm actually weirdly okay with my system. In spite of the fact that people have told me I should use hardware encoding for streaming (though from watching tech vloggers with top end hardware Nvenc adds it's own FPS hit) I seem to be having no troubles. It's like things are really not that bad for me but I'm keeping an eye out because every review is like "YOUR SYSTEM IS ALMOST DEAD AND WORTHLESS"
|
# ? Oct 17, 2017 01:41 |
|
If youre happy with it, don't worry about it man. I personally think that this is a good purchasing moment for both AMD and Intel, as I don't expect huge things out of the Ryzen refresh or Ice Lake. So cool stuff is a ways away. But if youre happy, don't even worry about it. Fancy graphics card or fancy pants screen or big SSD would probably be a better daily use upgrade.
|
# ? Oct 17, 2017 03:38 |
|
RAM is not worth upgrading. Increasing your turbo boost (overclocking) certainly is. My opinion on RAM comes from my experience going from 1600 to 2133 with my 2500k. If my processor is not OC'd up to 4.5ghz where it should be it seems like the computer is broken. I tried the faster RAM speed, my motherboard only got it stable at 1866 and honestly I can't tell the difference. It is SO lost in the noise of SSD's, graphics card upgrades and the aforementioned overclock. I play games like Rocket League, World of Tanks, DOOM, X-COM 2 and they all run great at over 100fps at 1440p. (except x-com 2, duh).
|
# ? Oct 17, 2017 03:43 |
|
Jago posted:RAM is not worth upgrading. Increasing your turbo boost (overclocking) certainly is. So you went up one ram speed increment and you can't tell the difference so it's not worthwhile. Those graphs and stuff that Digital Foundry aren't any more scientific, are they?
|
# ? Oct 17, 2017 06:11 |
|
Craptacular! posted:This guy's talking about people still on Sandy/Ivy can get a 3770K with 2400 RAM. I have 1600 RAM, because everything above that was officially labelled as an overclock and that word causes me to think of system crashes and broken warranties. How big an improvement should it be, or should I just keep waiting for Ryzen refresh instead? I really like the 1700 as a nice balance between sacrificed gaming FPS for better encoder throughput when streaming, but feel I could get better bang for buck if I can hold out five more months. The problem is, if you're talking about getting a new processor (the 3770K, which is still $200+) and RAM, you may as well just get a new Coffee Lake, Z370 and DDR4 RAM to begin with, or a Ryzen system.
|
# ? Oct 17, 2017 06:18 |
|
Overclockable and already-overclocked quad cores without hyperthreading have ~12-18 months of comfortable use left. Hyperthreaded OCed quads probably have 24-36 months of headroom. 67/7700K owners shouldn't be sweating much.
BIG HEADLINE fucked around with this message at 06:55 on Oct 17, 2017 |
# ? Oct 17, 2017 06:23 |
|
VulgarandStupid posted:The problem is, if you're talking about getting a new processor (the 3770K, which is still $200+) and RAM, you may as well just get a new Coffee Lake, Z370 and DDR4 RAM to begin with, or a Ryzen system. I already own a 3770K, and have since January 2013. And with 8GB DDR3-1600, I haven’t really had much issue beyond not being able to use anything above Medium textures in Quake Champions because their memory utilization is so awful. I mulled buying another 8GB, decided not to, but thought I’d ask if faster RAM would do much of anything at this level. Craptacular! fucked around with this message at 08:01 on Oct 17, 2017 |
# ? Oct 17, 2017 07:59 |
BIG HEADLINE posted:Overclockable and already-overclocked quad cores without hyperthreading have ~12-18 months of comfortable use left. Hyperthreaded OCed quads probably have 24-36 months of headroom. 67/7700K owners shouldn't be sweating much.
|
|
# ? Oct 17, 2017 07:59 |
|
Craptacular! posted:I already own a 3770K, and have since January 2013. And with 8GB DDR3-1600, I haven’t really had much issue beyond not being able to use anything above Medium textures in Quake Champions because their memory utilization is so awful. Here's some good videos for you. They compare the 3770K against itself OCed and with or without fast RAM. https://www.youtube.com/watch?v=2ZxZiksWtRQ https://www.youtube.com/watch?v=tbGT-u4i3EY https://www.youtube.com/watch?v=4sx1kLGVAF0 https://www.youtube.com/watch?v=VDo-j00vUtw I guess you should use this to weigh whether its worth getting 16GB of faster RAM vs possibly upgrading the whole shebang sooner. It definitely makes a difference in some games, but don't just go one increment faster like the guy above and expect it to be noticeable. DDR3 is fairly expensive right now, but not as expensive as DDR4, so maybe it is a good idea. When I checked the other day, fast DDR3 RAM was barely more expensive than 1600. VulgarandStupid fucked around with this message at 10:23 on Oct 17, 2017 |
# ? Oct 17, 2017 10:20 |
|
I attempted to go two up, but my six year old board wasn't having it. It's not going to be a big noticable jump even if you get results like the digital foundry article. Maybe certain specific titles, and it's a lot of money for single digit performance increases. I think I made clear that I was offering anecdotal evidence as well.
|
# ? Oct 17, 2017 12:03 |
|
I recently upgraded a rig from 2500k/8GB 1333 to a 2600k/16GB 2133. The 2600k and 16GB of ram cost me $180 total. I figure I can still sell the i5 and ram for maybe $100 or so. Should be a decent enough upgrade to let the system last a bit longer. Should be plenty for 1080p/75 for a few more years.
|
# ? Oct 17, 2017 13:36 |
|
BIG HEADLINE posted:Overclockable and already-overclocked quad cores without hyperthreading have ~12-18 months of comfortable use left. Hyperthreaded OCed quads probably have 24-36 months of headroom. 67/7700K owners shouldn't be sweating much.
|
# ? Oct 17, 2017 14:55 |
|
about 3 years
|
# ? Oct 17, 2017 17:21 |
|
Based on what?
|
# ? Oct 17, 2017 19:42 |
|
Sigh, got an I7-8700K on backorder with Newegg, and now they've got bundles with that + certain motherboards in stock and shipping, while my order sits in "15-20 day" limbo. I guess it's a good sign that maybe it'll actually ship before 20 days as opposed to the "end of November" some places are reporting, but it would have been nice had they shipped the backorders before putting them aside for bundles. If I didn't already have my motherboard, I'd just cancel and order the bundle.
|
# ? Oct 17, 2017 19:53 |
|
Paul MaudDib posted:Hilariously enough I'm sure the SLI/CF situation is going to get even worse in the DX12/Vulkan era since now you will need to deliberately code for it from day 1 rather than just having NVIDIA hack together a profile for you after the fact. SLI/CF are very obviously in a death spiral of low-adoption/low-support right now. Doesn't it also mean that once some of the engine builders have coded for it you'll get excellent SLI support from all the games using that engine without the game devs or nvidia having to do anything?
|
# ? Oct 17, 2017 19:57 |
|
TVs Ian posted:Sigh, got an I7-8700K on backorder with Newegg, and now they've got bundles with that + certain motherboards in stock and shipping If they were they've all sold out since you posted, as far as I can tell.
|
# ? Oct 17, 2017 20:00 |
|
scrubs season six posted:If they were they've all sold out since you posted, as far as I can tell. Yeah, they were up for a few minutes, but went pretty quick.
|
# ? Oct 17, 2017 20:03 |
|
The 8700 I ordered on 10/6 from Amazon still shows "we will let you know when we have an estimated delivery date" :/
|
# ? Oct 17, 2017 20:10 |
|
I was cleaning up my PCPP setups and remembered my old draft for a streaming/recording PC that just encodes video. What are some of the most viable CPU+mobo picks for something like that - surely there has to be something more sensible than just going with the latest Intel consumer CPUs?
|
# ? Oct 17, 2017 20:25 |
|
ufarn posted:I was cleaning up my PCPP setups and remembered my old draft for a streaming/recording PC that just encodes video. I mean you could get a cheap xeon server like a Dell T20 or Lenovo T140 but those $200-$400 servers are just quad core haswell xeons. Make good 4 bay Plex Servers/NASes though.
|
# ? Oct 17, 2017 20:28 |
|
ufarn posted:I was cleaning up my PCPP setups and remembered my old draft for a streaming/recording PC that just encodes video. Intel consumer CPUs are the best performance per dollar out of anything in their stack. If you're really after bang for the buck, look at Pentium G4560s, untouchable value.
|
# ? Oct 17, 2017 20:30 |
|
BIG HEADLINE posted:Overclockable and already-overclocked quad cores without hyperthreading have ~12-18 months of comfortable use left. Hyperthreaded OCed quads probably have 24-36 months of headroom. 67/7700K owners shouldn't be sweating much. Hyper threading adds MAYBE a 5% performance improvement under very specific workloads that typically aren't games. You're really overselling its value.
|
# ? Oct 17, 2017 20:38 |
|
I'm a HT decrier, I run my 5820K with HT disabled because it hurts performance in stuff I do, but I think the idea is once stuff starts really using 6 cores the 4C8T variants of processors will be able to hold on a little longer than the straight 4C4T versions. Or maybe once stuff wants to use 6 cores 4 extra virtual ones won't do any good anyways.
|
# ? Oct 17, 2017 20:44 |
|
TVs Ian posted:Sigh, got an I7-8700K on backorder with Newegg, and now they've got bundles with that + certain motherboards in stock and shipping, while my order sits in "15-20 day" limbo. My Newegg 8700k backorder placed about 45 minutes after they went live on the morning of 10/5 just tipped over to verification, shipping in 24 hours, status. I suspect they got a shipment large enough to settle their backorders and put the remainder up as combo deals.
|
# ? Oct 17, 2017 20:44 |
|
craig588 posted:I'm a HT decrier, I run my 5820K with HT disabled because it hurts performance in stuff I do, but I think the idea is once stuff starts really using 6 cores the 4C8T variants of processors will be able to hold on a little longer than the straight 4C4T versions. HT doesn't add cycles, it just allows you to jam instructions through the gaps in the execution pipeline. In the best case, you're pushing instructions through there that have no latency requirements (for some reason) so they won't negatively impact anything but in real world scenarios you're just incurring additional latency for that little bit of extra output. This is great for large computational loads that run in batches like you get in business environments where throughput trumps transactional latency, but its generally poo poo for games and overcommit past your physical cores and in to the logical HT ones is a recipe for poo poo performance and latency/concurrency problems.
|
# ? Oct 17, 2017 20:51 |
|
BangersInMyKnickers posted:Hyper threading adds MAYBE a 5% performance improvement under very specific workloads that typically aren't games. You're really overselling its value. Curious what you're basing this on? Digital Foundry's tests for i5 vs i7, for example, show 30% for games that multithread well. Which is consistent with the gains for many other (non-gaming) parallel workloads.
|
# ? Oct 17, 2017 20:56 |
|
Ordered the 8700k + auros gaming 7 combo and its already in packaging. A bunch of the newegg backorders shipped also but some got cancelled lol.
|
# ? Oct 17, 2017 21:01 |
|
crazypenguin posted:Curious what you're basing this on? Digital Foundry's tests for i5 vs i7, for example, show 30% for games that multithread well. Which is consistent with the gains for many other (non-gaming) parallel workloads. We've done extensive testing of it with business workloads and the rub basically is that if you are seeing large HT gains, it because your worker threads are full of holes because your code is poorly optimized. Once the load is back up because something is busy, those holes in the pipeline disappear and now that second thread you were trying to jam through a single core is forced to queue along side the primary thread and the latency goes to poo poo. Yeah, your max FPS or whatever during periods were you aren't CPU bottlenecked is going to improve but the low FPS periods where it is CPU bottlenecked is going to be worse than if HT wasn't being utilized because of the additional latency/queuing/contention. Avg numbers doesn't mean much when your floor is down and standard deviation is up in both directions. HT is a blunt, dumb tool and optimization inside the existing worker threads on 1 thread per core is better bang for the buck because of the increased consistency.
|
# ? Oct 17, 2017 21:03 |
|
Are these high-speed trading or scientific numerical simulations? Pauses waiting on RAM are going to always be present for games. Trees and graphs are too natural of a data structure for pretty much everything non-graphics (and even a lot of graphics, scene graphs and the like). Those will always involve a lot of pointer indirection, even if it can be reduced, and let's face it... Game devs aren't always the bestest most efficient coders imaginable. (e: I suppose I should say this is because the business only demands "good enough" and rarely offers the time to do an actual good job.) crazypenguin fucked around with this message at 21:21 on Oct 17, 2017 |
# ? Oct 17, 2017 21:13 |
|
scrubs season six posted:Doesn't it also mean that once some of the engine builders have coded for it you'll get excellent SLI support from all the games using that engine without the game devs or nvidia having to do anything? That's the dream. However it doesn't seem to be quite working out how everyone hoped. Prior to DX12 there were only 2 companies that really cared about good multi-card performance; nVidia and AMD. It was fairly common for one of them (usually NV) to embed a dev or two at the studios making AAA games whose only job was to provide help to the game devs on optimizing their engine for that brand of card and develop driver optimizations (including SLI profiles) for their respective GPU company. This is how you get driver updates for specific games released right before they come out. This worked great for both parties; the game devs got a free developer who was focused on performance optimization and the GPU companies got games that were optimized for their GPUs which results in better benchmark scores which then drove sales. It was certainly in the GPU companies interest to get SLI/Crossfire scaling well because that means you'll get more people buying more video cards especially at the high end where margins are better. This was especially true for nVidia since SLI is a custom multi-card implementation they could create market segmentation by only allowing certain cards to be used in SLI. The move in DX12/Vulkan to lower level graphics coding has introduced the double-edged-sword of mostly platform agnostic graphics optimizations. This allows developers to finely tune their engines and ultimately allows for better performance. However I think this will also result in fewer instances GPU companies embedding devs at studios since it's no longer going to be easy for them to optimize a game/driver for a particular GPU. Any help they provide is just as likely to benefit the competition. This means that now more and more game devs are going to be left on their own to optimize their games. Game devs make money selling you games so there's very little monetary advantage in them trying to add code for multi-gpu support; it would only benefit a tiny fraction of their user base and cost development time that could be better spent creating a new game/DLC they could sell you.
|
# ? Oct 17, 2017 21:14 |
|
NJ Deac posted:My Newegg 8700k backorder placed about 45 minutes after they went live on the morning of 10/5 just tipped over to verification, shipping in 24 hours, status. I suspect they got a shipment large enough to settle their backorders and put the remainder up as combo deals. Not mine yet, but it was about 4 hours in. Hopefully that's what they did and it'll still process today.
|
# ? Oct 17, 2017 21:23 |
|
TVs Ian posted:Not mine yet, but it was about 4 hours in. Hopefully that's what they did and it'll still process today. Some peoples preorders got cancelled
|
# ? Oct 17, 2017 21:34 |
|
BangersInMyKnickers posted:We've done extensive testing of it with business workloads and the rub basically is that if you are seeing large HT gains, it because your worker threads are full of holes because your code is poorly optimized. And we know how important code optimization is for game developers.
|
# ? Oct 17, 2017 21:39 |
|
|
# ? Jun 12, 2024 13:55 |
|
NJ Deac posted:My Newegg 8700k backorder placed about 45 minutes after they went live on the morning of 10/5 just tipped over to verification, shipping in 24 hours, status. I suspect they got a shipment large enough to settle their backorders and put the remainder up as combo deals. Yup, I ordered mine 5 minutes after backorders went up and just got my confirmation email. Now to debate on canceling my Amazon preorder right now or wait until I get the shipping number from NewEgg.
|
# ? Oct 17, 2017 21:52 |