|
I currently have a delidded 8700K, and my brother has an 8400. We are both using mITX boards/cases. At this point, it looks like the 10600K is basically an 8700K, the 10700K is a 9900K and the 10900K is a new 10-thread beast that's going to thermal throttle itself into oblivion if it even hears the word ITX. I don't think there are huge gains from upgrading from an 8700K, but my brother has been complaining about frame drops in some games... so maybe get a 9700K for him at some point?
|
# ? May 24, 2020 18:14 |
|
|
# ? Apr 18, 2024 04:30 |
|
You might want to see if it's his graphics card bottlenecking his system first. I'm running a 4790k and a 1070 and in theory I need to replace the graphics card to get better FPS in the games I play. I'm planning on grabbing a 10600k* only because I'll need to replace my motherboard/ram anyway so I may as well get the new socket now, and one of the new 3-series Nvidia cards later. *which can apparently be overclocked to match the performance of a stock 10900k...
|
# ? May 24, 2020 22:14 |
Nutsak posted:You might want to see if it's his graphics card bottlenecking his system first. I'm running a 4790k and a 1070 and in theory I need to replace the graphics card to get better FPS in the games I play.
|
|
# ? May 25, 2020 16:59 |
|
I can't imagine upgrading after 2 CPU generations. I guess when I was in high school or college my hardware moved faster. Right now I'm still on 4th gen
|
# ? May 26, 2020 18:13 |
|
still using a Gulftown
|
# ? May 27, 2020 17:26 |
|
Shaocaholica posted:I can't imagine upgrading after 2 CPU generations. I guess when I was in high school or college my hardware moved faster. Right now I'm still on 4th gen Is it time to upgrade my Tbird?
|
# ? May 27, 2020 17:33 |
|
FuturePastNow posted:still using a Gulftown My second tier server is a HP Z400 running a X5650 (upgraded from a W3565), although the DIMMs have started making GBS threads the bed en mass now, had two failures in three months and I think a third one just failed. At least it’s DDR3 so I still have a chance of finding some new-ish DIMMs. It actually outlived its replacement, the DIY microserver I built poo poo the bed due to the USB-to-SATA adapter I was booting from, and the Z400 was taking up the slack. have a X99 WS/IPMI and a 1660v3 to replace it but I’ve talked myself into a Fractal 7 XL to replace it instead of the random case I was going to use, but not quite into spending the money yet, as this month has been Once I get that fixed the Z400 will become a machine for hosting my SCSI card to drive my ancient 4x5 scanner and maybe for the firewire if I ever get a nice Nikon film scanner for 120/35mm, so... it will live on. Paul MaudDib fucked around with this message at 18:18 on May 27, 2020 |
# ? May 27, 2020 18:07 |
|
I apologize if this has been answered but would a 6700k be a reasonable upgrade from a 6600k or am I looking at buying a new mobo/processor
|
# ? May 27, 2020 18:48 |
|
Cage Kicker posted:I apologize if this has been answered but would a 6700k be a reasonable upgrade from a 6600k or am I looking at buying a new mobo/processor at this point you're pretty much throwing good money after bad, just pony up for a 3600 and a new mobo.
|
# ? May 27, 2020 18:58 |
|
Cage Kicker posted:I apologize if this has been answered but would a 6700k be a reasonable upgrade from a 6600k or am I looking at buying a new mobo/processor If you can get it for <$50, sure, why not? If you mean is it worth the $300 that retailers seem to want for it ($250 on eBay, $175ish used), then no, it is not. As Paul said, just get a 3600 + mobo for about the same price.
|
# ? May 27, 2020 19:03 |
Paul MaudDib posted:My second tier server is a HP Z400 running a X5650 (upgraded from a W3565), although the DIMMs have started making GBS threads the bed en mass now, had two failures in three months and I think a third one just failed. At least it’s DDR3 so I still have a chance of finding some new-ish DIMMs.
|
|
# ? May 27, 2020 20:24 |
|
ratbert90 posted:Is it time to upgrade my Tbird? This is the Intel thread homie
|
# ? May 27, 2020 20:43 |
|
I still think the Intel and AMD threads should be collapsed into a single x86 thread.
|
# ? May 27, 2020 20:47 |
|
Farmer Crack-rear end posted:I still think the Intel and AMD threads should be collapsed into a single x86 thread. then start poo poo against an ARM thread
|
# ? May 27, 2020 20:49 |
|
Agreed to both.
|
# ? May 27, 2020 20:51 |
|
Figures that's what the intel thread would say about now
|
# ? May 27, 2020 20:53 |
|
ARM in the datacenter is the 2010s equivalent of Linux on the desktop
|
# ? May 27, 2020 20:58 |
|
Shaocaholica posted:This is the Intel thread homie Sorry! Is it time to upgrade my Prescott? I think I have a failing RDRAM module.
|
# ? May 27, 2020 21:13 |
|
ratbert90 posted:Sorry! Is it time to upgrade my Prescott? I think I have a failing RDRAM module. My dual core dual socket 90nm Netburst Xeons are fine for daily driving
|
# ? May 27, 2020 21:18 |
|
It's linux in the data center and arm on the desktop!
|
# ? May 27, 2020 21:24 |
|
D. Ebdrup posted:I've been looking at X5690s on ebay for my server to replace the E5620s - they're insanely cheap and aside from the IPC improvements between Westmere-EP and Sandy Bridge it's basically the same CPU as what I have in my current workstation/gaming machine, as far as I can tell. Just a tad bit noisier. like usual, the top Westmere SKUs have a very steep price increase for a pretty small increase in performance. Last I checked the X5670 was $20 and the X5690 was still like $90. I think that's not money well-spent there, the X5670 absolutely makes sense for a cheap performance boost on existing hardware but if you're going to drop an additional $140 (for two CPUs) for basically 300 MHz more clock, that's a signal that you actually do care about performance/efficiency/etc and I think you're better off trying to move to something newer. it's ultimately your money though and if you want to cram on a bunch of cheap RAM and you don't want to shell out for Sandy Bridge-E or Haswell-E then maybe it makes sense. Food for thought though, a 1660v3 is basically $150 at this point, can take ECC RDIMMs, and overclocked to 4 GHz or so the 8 Haswell-E cores will absolutely cream the 12 Westmere cores on the X5650. Even if they aren't "dedicated cores", it can still task swap and perform better than the 12 dedicated cores on Westmere. Or you can see if there are still any deals on those dual-socket Intel server boards and do a dual 2650v3 build or whatever. Paul MaudDib fucked around with this message at 23:15 on May 27, 2020 |
# ? May 27, 2020 23:01 |
Paul MaudDib posted:like usual, the top Westmere SKUs have a very steep price increase for a pretty small increase in performance. Last I checked the X5670 was $20 and the X5690 was still like $90. I think that's not money well-spent there, the X5670 absolutely makes sense for a cheap performance boost on existing hardware but if you're going to drop an additional $140 (for two CPUs) for basically 300 MHz more clock, that's a signal that you actually do care about performance/efficiency/etc and I think you're better off trying to move to something newer. I hadn't even looked at the one-rung-down CPUs, at all. Yeah, a X5670 will do me nicely too. The current server is a dual-CPU system on a QPI link (ie. it's a 20-bit bus at 5.7GT; ~1.7Gbps-114Gbps) between the CPU, so getting a octocore with SMT would also remove that bottleneck, but I kinda like the idea that an entire CPU, a third of the memory with RAIM, and up to 3 disks can fail and the system can still keep on chugging - and getting something equivalent is going to be difficult, because the used server market in Denmark is loving weird. I can get Westmere with ECC and RAIM for the price of picking it up, whereas even Sandy Bridge (which can be upgraded to Ivy Bridge CPUs by just replacing the CPU if the motherboard has a new enough revision) will set you back half the price of a good desktop build and a four year old Broadwell costs the full price of a newest-generation i5-i7 workstation build. BlankSystemDaemon fucked around with this message at 23:46 on May 27, 2020 |
|
# ? May 27, 2020 23:34 |
|
How do you OC a Xeon?
|
# ? May 28, 2020 00:36 |
|
Shaocaholica posted:How do you OC a Xeon? You go into the BIOS and change the multipliers Xeon E5 1650, 1660, and 1680 v1, v2, and v3, 1620 v1, and W-3175X, are multiplier-unlocked. As well as certain high-end W35xx, W36xx, and X56xx models are unlocked as well although that situation is kinda vague. W3680 and W3690 seem to be unlocked. https://forums.servethehome.com/index.php?threads/xeons-with-an-open-multiplier.15983/ You do need to use an X58/X79/X99 chipset, not a server/workstation chipset, of course, but there are certain boards (ASUS X99 WS) that support ECC including RDIMMs and LRDIMMs if the processor does (see the memory QVL). As well as chinese boards in theory, although I haven't heard great things about the overclocking options there. There is considerably more wiggle room in the "Xeon" branding than most people realize. In the Nehalem days, Intel didn't really care if you overclocked one of the top one or two xeons - you were already paying top dollar for one of the most expensive processors they offered. And quietly, they kept overclocking support in the Xeon equivalents to the i7 HEDT processors (1620, 1650, 1660, and 1680) up through the Haswell days. Definitely not widely advertised but I wouldn't be surprised if some of the sorts of HFT customers who bought things like EVEREST or BLACKOPS quietly made use of it. As with many things, it's not really until Broadwell that they got greedy and started clamping down, people just forget about the Before Time before BK started searching for every penny to bilk. They also brought the unlocked multiplier back on the W-3175X again though. Although that processor has other restrictions on total memory even though it has overclocking and ECC/RDIMM support, but it's really the least-restricted processor Intel has released since the Nehalem days. It's actually less restricted than AMD's options too, Epyc can't be overclocked (or, der8auer did some stuff but it basically involved severe BIOS hacks) and Threadripper has locked off RDIMM support, W-3175X can do both overclocking and ECC RDIMMs together. Paul MaudDib fucked around with this message at 04:31 on May 28, 2020 |
# ? May 28, 2020 01:04 |
|
Farmer Crack-rear end posted:I still think the Intel and AMD threads should be collapsed into a single x86 thread. No, you can't sit with us. Get off the 14nm process first loser
|
# ? May 28, 2020 03:31 |
|
Then we could have a thread title including both PLATFROM and COCK SPEED; but apart from that.. I kinda like the two threads.
|
# ? May 28, 2020 03:34 |
|
start a new combined thread and close the two legacies with a post each that simply says “thread er”
|
# ? May 28, 2020 03:42 |
|
Zedsdeadbaby posted:No, you can't sit with us. Get off the 14nm process first loser lol I'm still on 22nm myself, some of us don't compulsively upgrade their computers my next one's probably gonna be AMD though
|
# ? May 28, 2020 04:05 |
|
I’m lucky if my daily(in rotation) is smaller than 45nm. I’ll probably be happy with 45nm for a long time yet.
Shaocaholica fucked around with this message at 07:12 on May 28, 2020 |
# ? May 28, 2020 07:01 |
|
Zedsdeadbaby posted:No, you can't sit with us. Get off the 14nm process first loser Excuse me that's 14++++
|
# ? May 28, 2020 08:32 |
|
Malcolm XML posted:Unless you're samsung and dgaf https://link.medium.com/arrb9RG8u6 you can't even guard it with asm capability check (because the OS lied), you could get switched onto another core after the check EoRaptor posted: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.
|
# ? May 28, 2020 16:33 |
|
JawnV6 posted:lol incredible This is not the first time Samsung has done things like this and created very dangerous situations. The Note 7 had different cache line sizes per core: https://www.theregister.co.uk/2016/09/13/arm_biglittle_gcc_bug/
|
# ? May 28, 2020 16:37 |
|
Twerk from Home posted:This is not the first time Samsung has done things like this and created very dangerous situations. The Note 7 had different cache line sizes per core: https://www.theregister.co.uk/2016/09/13/arm_biglittle_gcc_bug/ oh my god it's like an exam question about self-modifying code, caches, and false sharing come to life
|
# ? May 28, 2020 16:40 |
|
GRINDCORE MEGGIDO posted:Then we could have a thread title including both PLATFROM and COCK SPEED; but apart from that.. I kinda like the two threads. Has the current thread title been cock speed for awhile and I just naturally read it as clock and never noticed?
|
# ? May 28, 2020 17:53 |
|
latinotwink1997 posted:Has the current thread title been cock speed for awhile and I just naturally read it as clock and never noticed? A mod updated it recently, to bring us in line with: A typo in a presentation doc they made.
|
# ? May 28, 2020 19:18 |
|
Really cocked that one up, didn't they?
|
# ? May 28, 2020 19:22 |
|
Definitely pulled a boner
|
# ? May 28, 2020 19:24 |
|
GRINDCORE MEGGIDO posted:A mod updated it recently, to bring us in line with: Ha, even better than I could have imagined.
|
# ? May 29, 2020 05:07 |
|
https://www.youtube.com/watch?v=qQ_AETO7Fn4 Good to see that the motherboard makers are still up to their regular tricks. Reminder that "stock", "auto OC", or "AI Tuner" settings on Z series boards are nearly always a disaster and you need to manually tune them. Giga feeds the 10900k 25% over stock voltage out of the box...
|
# ? May 29, 2020 21:45 |
|
|
# ? Apr 18, 2024 04:30 |
|
I think it's too easy to say "hot and power hungry is all on motherboard manufacturers", Intel is setting the limits and could enforce them but for some reason chose not to. Instead, they're pushing even more MCE-style OC they're not calling OC through BFB. In the end, as long as mainboard manufacturers do somewhat reasonable +voltage OC (so, probably not what Gigabyte is doing with that board he tested ) they're probably happy to take more "longer bar = more better" wins.
|
# ? May 29, 2020 23:48 |