Need advice on this build.

[Bill_Bright]

Notice how low efficiency is at power loads below 20%.

Low? That is a very deceiving statement!

Look at the scales used on that graph. According to that graph, that PSU's efficiency at 20% is sill above 88% @ 110VAC and above 89% @ 230VAC. That is still amazingly high! And there is less than a 4% spread from 20% to 100% load. That is fantastic!

Understand a cheap PSU is likely to be just 70% efficient around 50% load and drop off significantly on either side from there - that is, a cheap PSU will have very steep "bell" shaped curve, with just one peak efficiency value that is not very good to start.

That CS450M has a very nice, relatively "flat" efficiency across the full range of expected loads that starts out very high above 88%, and never drops below that value.

 

[thebluebumblebee]

Low? That is a very deceiving statement!

Look at the scales used on that graph. According to that graph, that PSU's efficiency at 20% is sill above 88% @ 110VAC and above 89% @ 230VAC. That is still amazingly high! And there is less than a 4% spread from 20% to 100% load. That is fantastic!

Understand a cheap PSU is likely to be just 70% efficient around 50% load and drop off significantly on either side from there - that is, a cheap PSU will have very steep "bell" shaped curve, with just one peak efficiency value that is not very good to start.

That CS450M has a very nice, relatively "flat" efficiency across the full range of expected loads that starts out very high above 88%, and never drops below that value.

I wasn't clear. Think about how low the efficiency is for a larger PSU at that same low load. How low is the efficiency of a 750 watt PSU at 80 watts?

 

[Bill_Bright]

I wasn't clear. Think about how low the efficiency is for a larger PSU at that same low load. How low is the efficiency of a 750 watt PSU at 80 watts?

This is true - but not in all cases and it is getting better all the time as being "green" becomes more important and accepted. And for this reason, the 80 PLUS Certification has been expanded to include Titanium which adds a 10% load efficiency requirement of 90% efficiency.

So to your point that buying to big a PSU is a waste of money due to poorer efficiencies a lower load levels (which is true), IMO, what buying too big really does with a quality PSU is waste your money at the time of purchase.

 

[rubenclavs]

@Roshan

Here is my suggestion

Intel Core i5-4690K 3.5GHz Quad-Core Processor - $229.99
ASRock Z97 PRO3 ATX LGA1150 Motherboard - $80.98
Cooler Master Hyper 212 EVO - $26.98
G.Skill Ripjaws X Series 8GB (2 x 4GB) DDR3-1600 - $49.99
Sandisk Solid State Drive 128GB 2.5" Solid State Drive - $52.89
Western Digital Caviar Blue 1TB 3.5" 7200RPM - $58.88
Zotac GeForce GTX 970 4GB - $310.95
NZXT S340 - $64.99
EVGA 500W 80+ Bronze - $37.99

Total: $913.64

 

[Roshan]

@Roshan

Here is my suggestion

Intel Core i5-4690K 3.5GHz Quad-Core Processor - $229.99
ASRock Z97 PRO3 ATX LGA1150 Motherboard - $80.98
Cooler Master Hyper 212 EVO - $26.98
G.Skill Ripjaws X Series 8GB (2 x 4GB) DDR3-1600 - $49.99
Sandisk Solid State Drive 128GB 2.5" Solid State Drive - $52.89
Western Digital Caviar Blue 1TB 3.5" 7200RPM - $58.88
Zotac GeForce GTX 970 4GB - $310.95
NZXT S340 - $64.99
EVGA 500W 80+ Bronze - $37.99

Total: $913.64

Thank you! Ill do a side by side with my current build and see if I can make any improvements!

 

[nsdp]

First off , you are spending your money in the wrong place if you are gaming.
1.Try this for $550 through the 11th. http://www.newegg.com/Product/Produ...&PID=361116&SID=ebs2d-1d9344d1433622005726sbe Tomshardware says it is still the fastest card out there by 33% over the GTX980 Spec.org and Tech Powerup say it is twice as fast as the 980 Tech powerup 5,733 Gflops x2 for the 295 vs 4,616 Gflops for the plane vanilla GTX980 and 5,632Gflops for GTX980TI. The 970 comparison is a real laugher at 3,494Gflops or 1/3 of the 295.
2. You don't need more than a FX6350 to feed the 295 graphics card all the data it can handle. I know I am running Flownex http://www.flownex.com/ in 3D modeling gas turbine temperatures burning pure hydrogen and oxygen checking hot spotting on metals and it makes games look like Lego blocks. You won't be CPU limited. There are several I3 cpus that cost around a $120 that will do equally well if you want Intel and you could get by with a AMD 4350. . Read the review at Toms on the power supply wattage that the 295 needs.
3.OS Windows 7 at $70 tomorrow 6/7/2015 with code EMCATAT37 http://www.newegg.com/Product/Produ...60615-index-_-OperatingSystems-_-32416806-S0B If you need to upgrade go to windows 10 for free not 8. Software compatibility is better with W7.

4. For the motherboard any AMD 970 board will do. the suggested Intel boards are fine also.

Rest of your part choices are fine. Newegg's sale on the 295 makes anything else look bad and mismatched. I do very high tech gas turbine thermo models for a living. I don't waste computer time or my money.

 

[nsdp]

Roshan here is a direct comparison of the FX6350 and the I7 4790 Devil's Canyon http://www.cpu-world.com/Compare/453/AMD_FX-Series_FX-8350_vs_Intel_Core_i7_i7-4790K.html If you are gaming look about 3/4 of the way down the page at the gaming benchmarks 3Dmark 03,3DMark06 and 3D Mark 01 benchmarks and the PCMark 05 graphics. The other benchmarks are irrelevant for your purposes (and very questionable given Intel paying developers to use Intel biased compilers ). WITH a Radeon 6670 they are within 5% of each other. Both are GPU bottlenecked. as is shown by the difference which has a 100% correlation to CPU speed. 4000mhz/3913mhz or 5%. The PCMark 05 isn't measuring anything bottlenecked by the card. Basically a push in cpu performance for a $200 difference in price. That same $200 will buy you a 2X-3X improvement in game performance with the R9 295. This is a clearance price for AMD as they have new cards coming out the end of the month and dealers are selling/clearing inventory ahead of the announcement and release of the new cards. . Past history shows once the new models are out, clearance sales are over.

Here is a paper from Lawrence Berkeley National Labs http://crd.lbl.gov/assets/pubs_presos/jpdc09-lbmhd.pdf This paper won a best in track award at the IPDPS conference in Milan. Notice the difference in performance between the dual core AMD Santa Rosa and the Intel Cloverton Quad core using IEEE754 standard 64 bit compilers. The dual core Santa Rosa scored 7.4 gflops vs 5.6glops for the Intel Cloverton in the LBNL benchmarks. Compare that with the Passmark scores of 1415 https://www.cpubenchmark.net/cpu.php?cpu=AMD+Opteron+2214+HE&id=1615 vs 2917 http://www.cpubenchmark.net/cpu.php?cpu=Intel+Xeon+E5345+@+2.33GHzCan you say some one fudged the numbers?

The only benchmarks that are independently verified for accuracy are the spec.org benchmarks. The LBNL benchmarks are indicative of how much time and money are spent manipulating benchmarks in reviews on line.

 

[Roshan]

Roshan here is a direct comparison of the FX6350 and the I7 4790 Devil's Canyon http://www.cpu-world.com/Compare/453/AMD_FX-Series_FX-8350_vs_Intel_Core_i7_i7-4790K.html If you are gaming look about 3/4 of the way down the page at the gaming benchmarks 3Dmark 03,3DMark06 and 3D Mark 01 benchmarks and the PCMark 05 graphics. The other benchmarks are irrelevant for your purposes (and very questionable given Intel paying developers to use Intel biased compilers ). WITH a Radeon 6670 they are within 5% of each other. Both are GPU bottlenecked. as is shown by the difference which has a 100% correlation to CPU speed. 4000mhz/3913mhz or 5%. The PCMark 05 isn't measuring anything bottlenecked by the card. Basically a push in cpu performance for a $200 difference in price. That same $200 will buy you a 2X-3X improvement in game performance with the R9 295. This is a clearance price for AMD as they have new cards coming out the end of the month and dealers are selling/clearing inventory ahead of the announcement and release of the new cards. . Past history shows once the new models are out, clearance sales are over.

Here is a paper from Lawrence Berkeley National Labs http://crd.lbl.gov/assets/pubs_presos/jpdc09-lbmhd.pdf This paper won a best in track award at the IPDPS conference in Milan. Notice the difference in performance between the dual core AMD Santa Rosa and the Intel Cloverton Quad core using IEEE754 standard 64 bit compilers. The dual core Santa Rosa scored 7.4 gflops vs 5.6glops for the Intel Cloverton in the LBNL benchmarks. Compare that with the Passmark scores of 1415 https://www.cpubenchmark.net/cpu.php?cpu=AMD+Opteron+2214+HE&id=1615 vs 2917 http://www.cpubenchmark.net/cpu.php?cpu=Intel+Xeon+E5345+@+2.33GHzCan you say some one fudged the numbers?

The only benchmarks that are independently verified for accuracy are the spec.org benchmarks. The LBNL benchmarks are indicative of how much time and money are spent manipulating benchmarks in reviews on line.

Thank you for all this information!

 

[nsdp]

Fry's has a week long sale that is pretty good. for two items.
http://www.frys.com/search?query_string=8259456+8259466&site= 120gb SSD $59 and the 24gb SSD for $79 after promocode.
http://images.frys.com/art/email/060715_sun587mvc/sun_web.html?site=cemail060715 2x4gb Patriot 1866 for $45 after promo. Link is broken so you might call and see if you can get the memory. Fry's web site doesn't always work well. You might also price checka 6350 while you are at it and motherboard. Fry's won't match Neweggs price on the R9295.

I won't use it so I will send you my promo code via email.

 

[Caring1]

Here is a paper from Lawrence Berkeley National Labs http://crd.lbl.gov/assets/pubs_presos/jpdc09-lbmhd.pdf This paper won a best in track award at the IPDPS conference in Milan. Notice the difference in performance between the dual core AMD Santa Rosa and the Intel Cloverton Quad core using IEEE754 standard 64 bit compilers. The dual core Santa Rosa scored 7.4 gflops vs 5.6glops for the Intel Cloverton in the LBNL benchmarks. Compare that with the Passmark scores of 1415 https://www.cpubenchmark.net/cpu.php?cpu=AMD+Opteron+2214+HE&id=1615 vs 2917 http://www.cpubenchmark.net/cpu.php?cpu=Intel+Xeon+E5345+@+2.33GHzCan you say some one fudged the numbers?

The only benchmarks that are independently verified for accuracy are the spec.org benchmarks. The LBNL benchmarks are indicative of how much time and money are spent manipulating benchmarks in reviews on line.

Nice story, but you conveniently left out the happy ending for Intel where they have higher compute cores and IPC and all lived happily ever after.
Those Passmark scores take in to account those compute cores.

 

[nsdp]

I am curious where does Intel have higher computer scores? I play in the big leagues and until this deal http://www.anandtech.com/show/9151/...contracts-for-2-dept-of-energy-supercomputers Intel has been shut out of the DOE DARPA world. Look at the HPC Challenge between 2005 and 2014. Intel is 0 for 108, AMD is 27 for 108, SPARC is 22 for 108 and the rest, 59 for 108 are IBM. If Intel was championship caliber they would have shown at least once. http://www.hpcchallenge.org/ Top 500 benchmarks declared obsolete in 2008 by original code writer Jack Dongarra. nsidehpc.com/2013/07/replacing-linpack-jack-dongarra-announces-new-supercomputing-benchmark/ Is IPC the computer version of TBall? I have been at this since 1963 and was a charter member of the IEEE computer Society in 1975. Was on the Spec.org committee in 1988. Never heard of it in professional computer science circles. How about BenchMarKing instead of BenchMarKETing?

 

[Roshan]

I am curious where does Intel have higher computer scores? I play in the big leagues and until this deal http://www.anandtech.com/show/9151/...contracts-for-2-dept-of-energy-supercomputers Intel has been shut out of the DOE DARPA world. Look at the HPC Challenge between 2005 and 2014. Intel is 0 for 108, AMD is 27 for 108, SPARC is 22 for 108 and the rest, 59 for 108 are IBM. If Intel was championship caliber they would have shown at least once. http://www.hpcchallenge.org/ Top 500 benchmarks declared obsolete in 2008 by original code writer Jack Dongarra. nsidehpc.com/2013/07/replacing-linpack-jack-dongarra-announces-new-supercomputing-benchmark/ Is IPC the computer version of TBall? I have been at this since 1963 and was a charter member of the IEEE computer Society in 1975. Was on the Spec.org committee in 1988. Never heard of it in professional computer science circles. How about BenchMarKing instead of BenchMarKETing?

I'm currently a Computer Engineering Major and we have a IEEE organization at my campus (MST)! Just thought that was cool. I will keep all this in mind when I build my friends computer here at the end of june, so thank you on the info!

 

[Caring1]

You're comparing Opterons and Xeons.
Try looking at what the average home user has at home.
p.s. I don't give a shit how long you have been in any field, those details only impress you, they bore everyone else.

 

[CrAsHnBuRnXp]

Personally, id hold out until Broadwell/Skylake. But if he needs a computer now, you have some good options already to go with.

 

[Roshan]

Personally, id hold out until Broadwell/Skylake. But if he needs a computer now, you have some good options already to go with.

Do you know when they will be released?

 

[CrAsHnBuRnXp]

Do you know when they will be released?

Broadwell in Q2 and Skylake in September/Octoberish. Skylake boards I think are coming out in august and are socket 1151 and support ddr4. Broadwell works on socket 1150 boards. And windows 10 is coming out July 29th and supports dx12 which improves performance by as much as 300% depending on the card and curent dx11 video cards will natively support dx12

I am curious where does Intel have higher computer scores? I play in the big leagues and until this deal http://www.anandtech.com/show/9151/...contracts-for-2-dept-of-energy-supercomputers Intel has been shut out of the DOE DARPA world. Look at the HPC Challenge between 2005 and 2014. Intel is 0 for 108, AMD is 27 for 108, SPARC is 22 for 108 and the rest, 59 for 108 are IBM. If Intel was championship caliber they would have shown at least once. http://www.hpcchallenge.org/ Top 500 benchmarks declared obsolete in 2008 by original code writer Jack Dongarra. nsidehpc.com/2013/07/replacing-linpack-jack-dongarra-announces-new-supercomputing-benchmark/ Is IPC the computer version of TBall? I have been at this since 1963 and was a charter member of the IEEE computer Society in 1975. Was on the Spec.org committee in 1988. Never heard of it in professional computer science circles. How about BenchMarKing instead of BenchMarKETing?

*yawns* what?

 

[nsdp]

I would not get too excited as Broadwell is the tick in Intel's tick/tock design strategy. http://en.wikipedia.org/wiki/Intel_Tick-Tock Intel introduces major upgrades (Haswell)tock and then releases fixes in(Broadwell)tick. The major improvement in Broadwell is a complete fix of this problem which began with the Core 2 architecture.. http://www.zdnet.com/article/remote-code-execution-through-intel-cpu-bugs/ That is why I would suggest you not waste time on Javascript but learn Python instead. Skylake will offer major improvements if Intel delivers at a premium premium far beyond the improvement . Sandy Bridge was kind of a bust so you don't really know. You run the risk of doing Intel's QC work for them. Broadwell is a die shrink of 33% The biggest percentage shrink in the past that worked well was 25%+/- 65/nm to 45nm. The other thing to note is that the heat and risk of failure go up as a function of the cube of the increase in frequency. Die shrink offsets this but is linear. We had a wonderful example of this phenomena with the Intel Netburst design. The other question is the CPU/GPU design a dead end? http://superuser.com/questions/308771/why-are-we-still-using-cpus-instead-of-gpus The GPGPU solution is clearly where we are headed. The CISC CPU is kind of where the RISC chip was in the 1990's. Road Runner and the Chinese Tianhe have made this pretty clear. The Cpu is now nothing more than a traffic cop directing traffic while the GPU does all the work. For my flow models I am running a S10000 on a Supermicro H8DACi board with two Opteron 290's for detailed double precision ECC calculations and a R9 295 on the same board set up for quick analysis. Memory is Winbond BH-5 which at 2-2-2-5 1T ($10/gb)will chew up your fastest DDR3/4 because of latency. DDR3 would have to be 6-6-6-15 to be as fast at moving a total volume of data. I have tried some of the newer CPU's and the 2-5% improvement they offer isn't worth it. The criteria for a CPU should be that it doesn't bottleneck your GPU. In the case we are talking about where you have $700 to spend on the CPU and GPU the 6350/R9 295 will totally destroy any I7/GTX970 or 980 . Spend your money where you get the most performance for your dollar. At $550 the R9 295 that figures to be about $48/Tflop. The cheapest 970/980 are $90+ /Tflop. In gaming there was a 5% gain for $200 on the I7 over the 6350. That same $200 yields a 300% increase in performance.