|
United Kingdom20285 Posts
This new AMD cpu looks amazing
Why?
You just have to read between the lines, there's a reason there is an 8-core 4ghz CPU parked next to a 4-core 3.4ghz CPU, same price point and everything but nobody on tech forums will touch the 8-core with a 1ft stick
Stock vs stock the 3570k has about a 35-40% single threaded performance lead over the 8350. It's weaker multithreaded (by about 25-30%) but you cant really apply that to games, ever. Even streaming, with current tech, it's very advantageous to give up multithreaded performance for single threaded, for either playability of the game or to keep game FPS above encode FPS (because a stream running at 60fps wont be able to show 60fps if your game is running at 50)
Overclocking increases single threaded performance advantage for intel and decreases multi threaded disadvantage, closes the gap out of AMD's favor some
It sounds great, they shoved all of their "worlds first 8-core desktop processor" and "wowow 3.6ghz" or whatever it was down peoples throats, all over ipl3 and stream ads hoping nobody would realize that an Ivy Bridge core at 4ghz is something like 55% faster than a Bulldozer core at 4ghz, but they really advertised to the wrong market there. Piledriver (newer revision of Bulldozer) is still not nearly fast enough to compare with 3570k in single threaded performance and Haswell is released before Steamroller (newer CPU's) so intel advantage will get bigger.
If the bulldozer architecture can beat intel in single threaded performance while simultaneously having twice as many cores at the same price point, it would be one hell of a victory, it's pretty much impossible and shouldnt be expected to happen, which means that while games are heavily reliant on single threaded performance (starcraft 2 in particular, some newer games not quiite so much, but dont benefit much if at all from going from 4 strong to 8 weak cores) they wont be of much use, if games were released that would heavily load 8 cores then they would be very competitive, but it still wouldnt be a landslide victory
|
What should I look for when building a computer designed to do computations / run regressions?
I use my laptop (i5-2430M, 540M, 8gb RAM) but it just takes too much time, so I'm thinking of building a desktop one.
|
On February 25 2013 07:54 Infernal_dream wrote:Show nested quote +On February 22 2013 09:09 minitelemaster wrote:
SC2 lag issue:
Normally, I get 20up and 4down with 10ms ping to the nearest server, courtesy of speedtest.net.
However, ONLY when 2 or more computers on the lan are in the same SC2 game (for 2v2, 3v3, 4v4), the internet quality suffers terribly. According to speedtest.net, the connection degrades to 500ms ping, 0.3up and 0.2down. This lag affects all 4 computers on the network, even if only 2 of them are in a game. I have no idea why computers not on SC2 would be effected.
Background info:
Recently switched to a linksys-cisco E2500 N600 router to improve wireless range. There are 4 computers on the lan, 2 of which first go through a TrendNet network switch. There was never a problem with the old netgear WNR1000 router working with team SC2 games. I have not tried port forwarding as of yet, as assigning static ip addresses would be a pain, and I don't know if this would fix the issue. Thanks for any help, and let me know if more info is required! Are you sure none of these computers have anything extra on them? Aka viruses or trojans that would leech your internet? Could just be your router is shitty at handling packets when the extra 2 come online.
Pretty sure there is no malware, and I don't think that the router is the problem, as the problem persists with 2 different routers. I've come to think it's just something weird with the new SC2 patch, since my connection bugs out ONLY when there are 2 or more lan connected computers in the exact same starcraft game. I don't know what to do except wait until the next patch to see if the problem goes away...
|
On February 28 2013 02:48 xN.07)MaK wrote:
What should I look for when building a computer designed to do computations / run regressions?
I use my laptop (i5-2430M, 540M, 8gb RAM) but it just takes too much time, so I'm thinking of building a desktop one.
An overview of the situation with Intel:
Your i5-2430M has two cores and four threads, desktop i5 has four cores and four threads, desktop i7 has four cores and eight threads, so the desktop i7 has everything exactly double of your current CPU. The desktop CPU also runs at much higher GHz, your mobile CPU goes up to 3 GHz with turbo, desktop goes up to 3.8 GHz or something like that.
Price difference between i5 and i7 is about 80 €. You need to research what the eight vs. four threads between i7 and i5 does for what you want to do. Both internally have the same number of cores, double of your mobile i5.
Without overclocking, you can look at motherboards with H77 chipset. With overclocking, you need at least another 50 €, as you can only look at Z77 boards, the CPU needs to be the "K" version, and you need an aftermarket cooler.
About memory, you need to know that you have to use a kit with two sticks, not one single stick, as this enables dual channel mode on the CPU.
About the PSU, choose something from a reputable manufacturer, if you look at something suspiciously cheap, you have to ignore the Watts it says it has when comparing, as it's fake.
An SSD just all around makes sitting at the PC more enjoyable, but doesn't really help with saving you time overall.
You also need to research what's up with your computing stuff and the GPU. If you don't use it, you can skip buying a graphics card.
About AMD instead of Intel, while the CPUs have eight cores, two cores share one FPU unit, so it'll not be better than an Intel with four cores.
|
On February 28 2013 02:48 xN.07)MaK wrote:
What should I look for when building a computer designed to do computations / run regressions?
I use my laptop (i5-2430M, 540M, 8gb RAM) but it just takes too much time, so I'm thinking of building a desktop one.
I think I can give you some help regarding this.
The most important question is 1) what program are you using to preform the computations/regressions, 2) the size of your data and 3) if you are manipulating data in addition to data analysis (merging datasets together and etc...)
|
Hello TL.net!
I'm looking to buy a new laptop. Preferably an ultrabook that has some gaming capabilities. As far as I can tell my best option is an ASUS UX51Vz-DH71 with Intel Core i7 3612QM, SSD and GT650 in the price range of 2000$.
My question is this: Is there any important developments being made regarding laptops that I should be aware of?
As in should I wait 6 months or so for something new and/or something cheaper?
Cheers!
|
On February 28 2013 05:44 Ata wrote:Show nested quote +On February 28 2013 02:48 xN.07)MaK wrote:
What should I look for when building a computer designed to do computations / run regressions?
I use my laptop (i5-2430M, 540M, 8gb RAM) but it just takes too much time, so I'm thinking of building a desktop one. I think I can give you some help regarding this. The most important question is 1) what program are you using to preform the computations/regressions, 2) the size of your data and 3) if you are manipulating data in addition to data analysis (merging datasets together and etc...)
Agreed, but it might be more relevant and practical at first to just look at per-core CPU usage (task manager, or similar), RAM usage, disk usage when doing the computations and regressions. That should tell us what improvements are required, what the bottleneck is, at least for the exact calculations being run now.
If you're writing your own code and could possibly optimize it better or use different resources, then the situation is less clear.
On February 28 2013 06:13 Arctyrus wrote:
Hello TL.net!
I'm looking to buy a new laptop. Preferably an ultrabook that has some gaming capabilities. As far as I can tell my best option is an ASUS UX51Vz-DH71 with Intel Core i7 3612QM, SSD and GT650 in the price range of 2000$.
My question is this: Is there any important developments being made regarding laptops that I should be aware of?
As in should I wait 6 months or so for something new and/or something cheaper?
Cheers!
In four months or so, Intel's next-gen processors will be out. Supposedly the top-end configuration will have integrated graphics with roughly comparable performance as GT 650M (err... according to Intel, so maybe they're comparing to a DDR3 model), so you might find an Ultrabook that's similar in performance for possibly a lower price and lower heat. Sticking ~75W+ worth of parts in an Ultrabook chassis is really stretching the system, for a current i7-3612QM + GT 650M kind of setup. CPU performance should be somewhat better, regardless of whether or not you get discrete or integrated graphics, but not by an amount that matters in most games.
Battery life could be improved as well, but probably not a lot.
There's always going to be something new and better later. If the current high price and heat / low battery life under load are no issue to you, then UX51VZ-DH71 should be fine.
|
On February 28 2013 06:14 Myrmidon wrote:Show nested quote +On February 28 2013 06:13 Arctyrus wrote:
Hello TL.net!
I'm looking to buy a new laptop. Preferably an ultrabook that has some gaming capabilities. As far as I can tell my best option is an ASUS UX51Vz-DH71 with Intel Core i7 3612QM, SSD and GT650 in the price range of 2000$.
My question is this: Is there any important developments being made regarding laptops that I should be aware of?
As in should I wait 6 months or so for something new and/or something cheaper?
Cheers! In four months or so, Intel's next-gen processors will be out. Supposedly the top-end configuration will have integrated graphics with roughly comparable performance as GT 650M (err... according to Intel, so maybe they're comparing to a DDR3 model), so you might find an Ultrabook that's similar in performance for possibly a lower price and lower heat. Sticking ~75W+ worth of parts in an Ultrabook chassis is really stretching the system, for a current i7-3612QM + GT 650M kind of setup. CPU performance should be somewhat better, regardless of whether or not you get discrete or integrated graphics, but not by an amount that matters in most games. Battery life could be improved as well, but probably not a lot. There's always going to be something new and better later. If the current high price and heat / low battery life under load are no issue to you, then UX51VZ-DH71 should be fine.
If the next gen processors have integrated graphics with roughly the same capabilities as GT 650M, won't that be a huge benefit? It would eliminate the need for a dedicated GPU which takes up space, battery life and costs extra.
As long as the integrated graphics can run SC2 I'll be satisfied.
|
They had a demo showing the top-config integrated graphics (most shipping will be slower) looking similar to GT 650M in some game. Probably fps is worse, and they probably picked a game that was more favorable to them. And Intel's probably going to be charging extra for the versions with the fasted integrated graphics configurations.
In terms of system complexity and maybe cost—for something constrained and premium like an Ultrabook—it might be a big deal being able to ditch the dedicated graphics for something not embarrassingly worse, but I wouldn't expect too much otherwise. To get discrete-level performance, they need to ramp up power consumption of the integrated parts to some level not too much different than the discrete parts. The performance doesn't come for free.
It's not like you'll get any kind of similar performance from one of the low-voltage chips (suffix U or Y) that are used in most Ultrabooks now.
Current HD 4000 in a non-low voltage part can run SC2 on low, just maybe not perfectly at 1920x1080. I guess if SC2 is the goal, then upcoming Haswell integrated graphics should be plenty for low, possibly for medium too, depending. Hopefully the CPU performance gains translate into something meaningful for SC2.
|
On February 28 2013 05:10 Ropid wrote:
An overview of the situation with Intel:
Your i5-2430M has two cores and four threads, desktop i5 has four cores and four threads, desktop i7 has four cores and eight threads, so the desktop i7 has everything exactly double of your current CPU. The desktop CPU also runs at much higher GHz, your mobile CPU goes up to 3 GHz with turbo, desktop goes up to 3.8 GHz or something like that.
Price difference between i5 and i7 is about 80 €. You need to research what the eight vs. four threads between i7 and i5 does for what you want to do. Both internally have the same number of cores, double of your mobile i5.
Without overclocking, you can look at motherboards with H77 chipset. With overclocking, you need at least another 50 €, as you can only look at Z77 boards, the CPU needs to be the "K" version, and you need an aftermarket cooler.
About memory, you need to know that you have to use a kit with two sticks, not one single stick, as this enables dual channel mode on the CPU.
About the PSU, choose something from a reputable manufacturer, if you look at something suspiciously cheap, you have to ignore the Watts it says it has when comparing, as it's fake.
An SSD just all around makes sitting at the PC more enjoyable, but doesn't really help with saving you time overall.
You also need to research what's up with your computing stuff and the GPU. If you don't use it, you can skip buying a graphics card.
About AMD instead of Intel, while the CPUs have eight cores, two cores share one FPU unit, so it'll not be better than an Intel with four cores.
Thx! Have 2 RAM sticks, SSD, but I guess still is not enough. Not a fan of overclocking, but can consider it.
On February 28 2013 05:44 Ata wrote:
I think I can give you some help regarding this.
The most important question is 1) what program are you using to preform the computations/regressions, 2) the size of your data and 3) if you are manipulating data in addition to data analysis (merging datasets together and etc...)
1) Stata 12
2) Around 5 gb
3) Yes, but this is not usually the problem as i just need to run it once or twice. But I change regressions all the time.
On February 28 2013 06:14 Myrmidon wrote:
Agreed, but it might be more relevant and practical at first to just look at per-core CPU usage (task manager, or similar), RAM usage, disk usage when doing the computations and regressions. That should tell us what improvements are required, what the bottleneck is, at least for the exact calculations being run now.
If you're writing your own code and could possibly optimize it better or use different resources, then the situation is less clear.
Usually RAM usage is 100%, but will check at the per-core CPU usage (I believe it wasn't as used as the RAM). Yes, I'm writing my own code so there is a high chance I'm the bottleneck in this case :D
|
United Kingdom20285 Posts
Not a fan of overclocking, but can consider it.
Im not sure of prices in spain, so il use UK prices.
Going from £190 to £280 on CPU (i5 to i7) to add hyperthreading gives about a 20-25% performance increase in heavily multithreaded applications, but nothing else.
By overclocking the £190 CPU, you can push it to 140% of the clock speed 24/7, maybe a touch less if you are not advanced (or quite a bit more if you are really advanced), and get that translated pretty much 1:1 into performance on anything CPU bound at all, multithreaded or not.
|
On February 28 2013 07:59 xN.07)MaK wrote:Show nested quote +On February 28 2013 05:10 Ropid wrote:
An overview of the situation with Intel:
Your i5-2430M has two cores and four threads, desktop i5 has four cores and four threads, desktop i7 has four cores and eight threads, so the desktop i7 has everything exactly double of your current CPU. The desktop CPU also runs at much higher GHz, your mobile CPU goes up to 3 GHz with turbo, desktop goes up to 3.8 GHz or something like that.
Price difference between i5 and i7 is about 80 €. You need to research what the eight vs. four threads between i7 and i5 does for what you want to do. Both internally have the same number of cores, double of your mobile i5.
Without overclocking, you can look at motherboards with H77 chipset. With overclocking, you need at least another 50 €, as you can only look at Z77 boards, the CPU needs to be the "K" version, and you need an aftermarket cooler.
About memory, you need to know that you have to use a kit with two sticks, not one single stick, as this enables dual channel mode on the CPU.
About the PSU, choose something from a reputable manufacturer, if you look at something suspiciously cheap, you have to ignore the Watts it says it has when comparing, as it's fake.
An SSD just all around makes sitting at the PC more enjoyable, but doesn't really help with saving you time overall.
You also need to research what's up with your computing stuff and the GPU. If you don't use it, you can skip buying a graphics card.
About AMD instead of Intel, while the CPUs have eight cores, two cores share one FPU unit, so it'll not be better than an Intel with four cores. Thx! Have 2 RAM sticks, SSD, but I guess still is not enough. Not a fan of overclocking, but can consider it. Show nested quote +On February 28 2013 05:44 Ata wrote:
I think I can give you some help regarding this.
The most important question is 1) what program are you using to preform the computations/regressions, 2) the size of your data and 3) if you are manipulating data in addition to data analysis (merging datasets together and etc...) 1) Stata 12 2) Around 5 gb 3) Yes, but this is not usually the problem as i just need to run it once or twice. But I change regressions all the time. Show nested quote +On February 28 2013 06:14 Myrmidon wrote:
Agreed, but it might be more relevant and practical at first to just look at per-core CPU usage (task manager, or similar), RAM usage, disk usage when doing the computations and regressions. That should tell us what improvements are required, what the bottleneck is, at least for the exact calculations being run now.
If you're writing your own code and could possibly optimize it better or use different resources, then the situation is less clear.
Usually RAM usage is 100%, but will check at the per-core CPU usage (I believe it wasn't as used as the RAM). Yes, I'm writing my own code so there is a high chance I'm the bottleneck in this case :D
Thats kinda wierd, I mostly use SAS but im thinking that you shouldnt have any issues running a regression on your laptop. I would say clean your laptop to avoid any issues with cpu overheating and close any RAM hungry programs that you do not need. This could potentially save you some money.
I just cant imagine the program needing more than 5gb of RAM to run your regressions. :S I want to recommend to simple upgread your laptop RAM to 16gb for ~80$ but Im not sure if that will fix your issue.
Finally, I tend to not print my results in SAS as that could take a long time (5 digit amount of regressions), since I need them to be outputed into another dataset anyways. So maybe spend ~1-2 hour and research/learn how to optimize the program in terms of memory usage and etc.
If you dont care about money and just want to get a desktop, make sure to get 16gb of ram and maybe a 3570k. (Dont think anything more would help)
And out of curiosity, what type of regression are you running and how long is it taking you.
Good luck
|
Thx for the help guys, I'll try to use some of your advices.
And out of curiosity, what type of regression are you running and how long is it taking you.
So far, just simple OLS, 10 million observations. Got like ~300+ regressions done (some of them very naive) in about 6h, which is surprisingly fast compared to other projects I've done in the past (bit more complex tho)
Today I discovered that Stata has a weird way to set the maximum amount of RAM to be used, under similar circumstances sometimes it uses all (8gb), and other like 5/6gb. While running continuously.
Maybe optimizing the code is best thing I can do to speed up the process, or just not bitch that much about it and let the comp do the job :D
|
|
|
some basic RCA audio cable pins broke off in my subwoofer. I used pliers to remove most of them, but one broke off too deep, and I can't get it the same way. Any ideas on how to get it out?
|
United Kingdom20285 Posts
"Specifications
- 17.3" Full-HD non-glare 1920x1080 LED Backlit screen
- Genuine Windows 8
- 3rd Generation "Ivy Bridge" Intel Core i7-3630QM Quad Core Processor
- Optional Memory Upgrade - Minimum 4GB DDR3 1600MHz, 4 slots, Max: 32GB (Choose more memory from drop downs)
- Nvidia GeForce GTX680M with 4GB of memory
- 2x 128GB SSD in RAID 0"
Kidding me? This is one of the best spec laptops in existance =P
Being a laptop CPU you will have like half of the performance of a highly overclocked desktop CPU so half the FPS in sc2, but its pretty much impossible to build significantly better.
Go with 8gb RAM
|
You can run it maxed out, but that doesn't mean the performance is going to be ideal.
GPU is fast enough, but you're not going to get more than 3.2-3.3 GHz on the CPU, which is not good enough for smooth SC2. Actually, it could be less if you're using a GT70. GT70 only has a single fan, probably not a great idea if you're trying to make use of the parts. A heavy load would cause throttling (at least according to reviews, but that was using a GTX 670M, which is a Fermi part that probably uses more power than the GTX 680M), but a game might not.
If you have more funds you can get desktop CPUs in "laptops", or just a i7-3940XM with a BIOS that allows for overclocking. Err, in a system that can handle the heat, maybe with external cooling as well.
But if you're looking at monstrosities or if you really want or need the performance, is there much of a point to a laptop form factor?
|
Is there a way to check how much power your computer is using?
I built my desktop back in 2006, and it held up remarkably well, but the graphics card fried itself last week. I bought a new one that I'm reasonably sure my gently aging PSU can handle, but I'd like to monitor it for the first few days to make sure. For reference, I have an Antec earthwatts 430w that will be powering a C2D 2.6GHz, 2 hdds, a dvd drive, and a new Radeon 6670.
|
On March 02 2013 06:31 Iranon wrote:
Is there a way to check how much power your computer is using?
I built my desktop back in 2006, and it held up remarkably well, but the graphics card fried itself last week. I bought a new one that I'm reasonably sure my gently aging PSU can handle, but I'd like to monitor it for the first few days to make sure. For reference, I have an Antec earthwatts 430w that will be powering a C2D 2.6GHz, 2 hdds, a dvd drive, and a new Radeon 6670.
If your PC stays turned on while gaming, the PSU is handling it, as a baseline rule. If you want to be more stringent, you combine a GPU stress test with a CPU one.
Probably the best blend would be the Combined Physics/Graphics test from one of the recent 3DMarks.
|
Earthwatts -> it's probably fine nm, from 2006, I'd be wary but would still expect it to work, especially since it currently works and an additional HD 6670 isn't much of a burden. I wouldn't be surprised to see some popped capacitors in there, depending on how hard it's been working over the years, but it could still manage.
You don't need to measure anything to know that your computer would be pulling well shy of 200W from the PSU. As long as the power supply is a non-defective sample, it should be okay.
If you want to measure power drawn by the entire system including the power supply (including losses from the conversion in the power supply; note that the 430W rating is for output power delivered to a computer and does not include those losses), get some kind of power meter. Some type of cheap Kill-A-Watt style meter may be like $20. To get more accurate results, you can spend more, even thousands.
|
|
|
|
|
|
EPT
