|
It certainly makes me consider Skylake-E more for my next upgrade, that's for damned sure.
|
#
?
May 26, 2015 04:09
|
|
|
|
| # ? Apr 20, 2024 00:29 |
|
|
You'll have plenty of time to save up if that roadmap is correct. Mid-ish 2017 is nearly 2.5 yr away.
|
|
#
?
May 26, 2015 14:14
|
|
|
PC LOAD LETTER posted:You'll have plenty of time to save up if that roadmap is correct. Yeah - I might settle for Broadwell-E, which is due out Q1 '16. Depends on what it offers, but 2.5y from now would mean this current system - at least 85% of the innards, will be 5 1/2 years old.
|
|
#
?
May 26, 2015 14:20
|
|
|
Meh, you'd think given DX12 and Vulkan, and their presumed multithreading awesomeness, that Intel will start shovelling out consumer hexacores, and maybe even octocores starting with Cannonlake. Either that, or introduce Skylake-E sooner.
|
|
#
?
May 26, 2015 16:56
|
|
|
So I haven't been reading the news for over 18mo now. What's the status of IHS TIM for Haswell-E and broadwell desktop?
|
|
#
?
May 26, 2015 17:50
|
|
|
Combat Pretzel posted:Meh, you'd think given DX12 and Vulkan, and their presumed multithreading awesomeness, that Intel will start shovelling out consumer hexacores, and maybe even octocores starting with Cannonlake. Either that, or introduce Skylake-E sooner. I dare say that unless Zen can make Intel sweat, they will have little compulsion to put out a general hex-core. :/
|
|
#
?
May 26, 2015 20:39
|
|
|
El Scotch posted:I dare say that unless Zen can make Intel sweat, they will have little compulsion to put out a general hex-core. :/ However outside of rendering, movie editing and other easily broken up tasks it's not useful at all. DX12 and gaming is not at all close to being multithreaded since it's super difficult to make (I am told).
|
|
#
?
May 26, 2015 20:43
|
|
|
Intel can probably get back their breathing room just by making 4 cores + hyperthreading a standard.Boiled Water posted:However outside of rendering, movie editing and other easily broken up tasks it's not useful at all. DX12 and gaming is not at all close to being multithreaded since it's super difficult to make (I am told). Pretty much everyone uses middleware/abstraction/someone else's API these days (especially on consoles, where you're probably using the system SDK AND something like UDK or Unity) and there's quite a few of those and they're cutting down on abstraction losses all the time. asm.js of all things is getting viable these days. Difficulty reaching the metal is real, but not the issue you believe it to be.
|
|
#
?
May 26, 2015 21:48
|
|
|
Am I getting a feeling from this thread that Cannonlake would be more worth an upgrade than Skylake because of the conjecture that Intel would need to push harder to complete against Zen by then? Honestly asking.
|
|
#
?
May 26, 2015 21:58
|
|
|
What's Zen? If anything, I cannot believe the mileage I'm getting out of my 2500k. A whole new rebuild isn't worth it for > 10%. The only new thing I'm going to miss is NVMe which is still in it's infancy.
|
|
#
?
May 26, 2015 22:12
|
|
|
People are really really really overestimating the CPU demand of 99% of consumer use cases
|
|
#
?
May 26, 2015 22:14
|
|
|
go3 posted:People are really really really overestimating the CPU demand of 99% of consumer use cases Which is what makes Sklyake interesting, a Skylake NUC or other <15W part probably is at performance parity with full wattage Sandy Bridge parts.
|
|
#
?
May 26, 2015 22:22
|
|
|
Tab8715 posted:What's Zen?
|
|
#
?
May 26, 2015 22:34
|
|
|
Zen is name for AMD's new chip architecture that is scheduled to be released in 2016. The only way Intel puts out anything better than the 4 core (8 thread) i7 that they've been doing would be if a comparably priced 6+ core (12+ thread) Zen chip had an actual measurable performance difference in several AAA games. Even then it would be a long shot since they'd have the 6+ core Broadwell-E line to compare against.
|
|
#
?
May 26, 2015 22:58
|
|
|
Do you think AMD will make a comeback with Zen?
|
|
#
?
May 26, 2015 23:01
|
|
|
Tab8715 posted:Do you think AMD will make a comeback with Zen? no, but it will hopefully be a first step.
|
|
#
?
May 26, 2015 23:15
|
|
|
Krailor posted:Zen is name for AMD's new chip architecture that is scheduled to be released in 2016. You sure about that? I bet that performance in AAA games is irrelevant for 95%+ of laptops and desktops sold.
|
|
#
?
May 26, 2015 23:19
|
|
|
Tab8715 posted:Do you think AMD will make a comeback with Zen? Do I hope they will? Yes. Do I think they will? No.
|
|
#
?
May 27, 2015 01:14
|
|
|
With AMDs performance I'm honestly perplexed how the company exists.
|
|
#
?
May 27, 2015 01:51
|
|
|
Tab8715 posted:With AMDs performance I'm honestly perplexed how the company exists. Low margin shitware computers for people who don't know and don't care and shop at Wal-Mart.
|
|
#
?
May 27, 2015 02:00
|
|
|
edit: wrong thread
|
|
#
?
May 27, 2015 02:16
|
|
|
Tab8715 posted:With AMDs performance I'm honestly perplexed how the company exists. There have been several years where they got by on awards from lawsuits, IIRC.
|
|
#
?
May 27, 2015 02:22
|
|
|
Tab8715 posted:If anything, I cannot believe the mileage I'm getting out of my 2500k. Yeah, I'm really happy I convinced myself to upgrade in 2011-12, and as nice as it would be to have 1000MB+/sec solid state storage, the only place I'll ever notice it is faster level load times. As for AMD, their current state does mean one thing. The time to upgrade is the second you hear they've declared bankruptcy, whether you need to or not. BIG HEADLINE fucked around with this message at 03:35 on May 27, 2015 |
|
#
?
May 27, 2015 03:31
|
|
|
BIG HEADLINE posted:Yeah, I'm really happy I convinced myself to upgrade in 2011-12, and as nice as it would be to have 1000MB+/sec solid state storage, the only place I'll ever notice it is faster level load times. The latest PCI-E SSDs take up so many lanes (and consumer CPUs provide so few lanes) that they steal bandwidth from your GPU anyway.
|
|
#
?
May 27, 2015 04:22
|
|
|
~Coxy posted:The latest PCI-E SSDs take up so many lanes (and consumer CPUs provide so few lanes) that they steal bandwidth from your GPU anyway. This is only the case with Dual-GPUs. I need to take a step back and while the performance gains aren't necessarily impressive from the desktop perspective or for those of us power users whom are still on Sandy Bridge but what does it mean of HTPCs and Laptops?
|
|
#
?
May 27, 2015 04:30
|
|
|
So if these NVMe SSDs can talk directly to the CPU without going through the PCH in a single socket layout, what happens with a multi socket layout like the X99? Forced to go through the PCH or connected directly to one of the CPUs?
|
|
#
?
May 27, 2015 04:38
|
|
|
Tab8715 posted:I need to take a step back and while the performance gains aren't necessarily impressive from the desktop perspective or for those of us power users whom are still on Sandy Bridge but what does it mean of HTPCs and Laptops? I'm planning on getting a new machine at some point, and realistically afaik everything right now is totally fine for my needs. Looking into stuff down the line is more just picking and choosing whatever new feature set theoretically seems like a good upgrade. Like uh, AVX-512? Hell if I know if anything I use will ever take advantage of it, but if they do it'll go faster! Of course there's no end to that logic, so realistically it's more of a justification to upgrade for whenever my current machine gets annoying enough to use.
|
|
#
?
May 27, 2015 06:41
|
|
|
~Coxy posted:The latest PCI-E SSDs take up so many lanes (and consumer CPUs provide so few lanes) that they steal bandwidth from your GPU anyway. Probably the only reason they're being 'generous' and giving Skylake 20 lanes. Just enough to SLI two GPUs in x8 and *one* NVMe drive at x4. That neglects the other things utilizing the available lanes, too. BIG HEADLINE fucked around with this message at 08:11 on May 27, 2015 |
|
#
?
May 27, 2015 08:09
|
|
|
BIG HEADLINE posted:Probably the only reason they're being 'generous' and giving Skylake 20 lanes. Just enough to SLI two GPUs in x8 and *one* NVMe drive at x4. That neglects the other things utilizing the available lanes, too. I don't think Skylake is getting 20 lanes from the CPU. 16 from the CPU and the PCH now has 20 3.0 lanes but they share bandwidth to the CPU.
|
|
#
?
May 27, 2015 15:51
|
|
|
Lowen SoDium posted:I don't think Skylake is getting 20 lanes from the CPU. 16 from the CPU and the PCH now has 20 3.0 lanes but they share bandwidth to the CPU. I've seen conflicting sources on this, some saying that a Haswell on a Z97 chipset supported 16 PCIe 3.0 lanes and 8 PCIe 2.0 lanes, while a Skylake on a Z170 chipset will just have 20 PCIe 3.0 lanes. Some (but not all) sources are saying that the Haswell configuration was 16 lanes managed by the CPU and 8 PCIe 2.0 lanes managed by the chipset, and the Skylake configuration is 16 lanes managed by the CPU and 20 PCIe 3.0 lanes managed by the chipset (36 lanes total). A quadrupling of bandwidth connected through the chipset seems unusual but not impossible. Hopefully it will all be cleared up soonish. Rastor fucked around with this message at 16:32 on May 27, 2015 |
|
#
?
May 27, 2015 16:16
|
|
|
Tab8715 posted:If anything, I cannot believe the mileage I'm getting out of my 2500k. A whole new rebuild isn't worth it for > 10%. The only new thing I'm going to miss is NVMe which is still in it's infancy. This reminds me of this image from an Intel slide:  So if you've got a Sandy Bridge you might expect a 33% IPC improvement when updating to Skylake. 
|
|
#
?
May 27, 2015 16:36
|
|
|
Where's Kentsfield, assholes?? It's gonna be more than 33% from a Q6600 but that's still quite disappointing for this timeframe.
|
|
#
?
May 27, 2015 17:22
|
|
|
Grim Up North posted:This reminds me of this image from an Intel slide: Aww, lil Penryn with its 2% 
|
|
#
?
May 27, 2015 17:33
|
|
|
Grim Up North posted:This reminds me of this image from an Intel slide: If the clocks are high. Sandy Bridge overclocks really well, also. That would also be at its most optimistic. It's kind of sad, but.. well, AMD isn't putting up enough of a fight.
|
|
#
?
May 27, 2015 17:40
|
|
|
japtor posted:Like uh, AVX-512? Hell if I know if anything I use will ever take advantage of it, but if they do it'll go faster!
|
|
#
?
May 27, 2015 21:31
|
|
|
Well, this is kind of an odd question. I'm probably going to have access to somebody who can replace BGA chips, does anybody know where I can find a list of compatible atom cpus for netbooks (Specifically the 1005HA eee pc, currently has an Intel Atom N270 / 1.6 GHz), and if it would actually be worth doing? I love the idea of having the little netbook that could, but am not up-to-date on atom processors at all.
|
|
#
?
May 29, 2015 22:01
|
|
|
surc posted:Well, this is kind of an odd question. Not worth doing. I imagine the bios has the microcode for the n280 chip but that's all as those were the only two Diamondville chips. Asomodai fucked around with this message at 23:40 on May 29, 2015 |
|
#
?
May 29, 2015 23:37
|
|
|
Tab8715 posted:What's Zen? Unless that 2500k is sitting at 5GHz or something you would get more than 10%, the benchmarks I saw were showing around a 8-10% increase from Haswell. MaxxBot fucked around with this message at 00:57 on May 30, 2015 |
|
#
?
May 30, 2015 00:55
|
|
|
MaxxBot posted:Unless that 2500k is sitting at 5GHz or something you would get more than 10%, the benchmarks I saw were showing around a 8-10% increase from Haswell. If you do have one of those 2500Ks or 2600ks that sits at 5GHz, then it looks like you're ahead of a 4790k still. A friend of mine has a 2600K at 5.1ghz on air comfortably, and just has no reason to upgrade.
|
|
#
?
May 30, 2015 01:15
|
|
|
|
| # ? Apr 20, 2024 00:29 |
|
|
5Ghz on air? My 2500k on water tops out at 4.5, drat
|
|
#
?
May 30, 2015 01:30
|
|
