New build advice for student

Tetras said:
I don't know what they used (because I haven't checked the test methodology), but there are different ways of getting this information.

For example, you can:
- Read it directly from the power connectors with your own equipment (I think gamers nexus use this method) and isolate the CPU from the board.
- Use software readings like from hwinfo, that report from the cores and package power.
- Use a plugin wall meter that takes it from the plug and subtract from this reading (e.g. PSU efficiency, motherboard and graphics card idles).

In reference to the question: why does the 13700K have higher idle power in the guru3d chart when the CPU idle draw is lower in tweaktown's chart, I would guess a few reasons:
- The nature of the readings and how they were measured (as mentioned above).
- The motherboard used (since it can make a big difference, especially a high-end one, which is what reviewers tend to use).
- How the BIOS / firmware was optimised.

I'd also reiterate again, most hardware reviewers just don't care about idle power and they don't worry about the test conditions. For example: they might even use a different graphics card, or PSU between reviews and this is not stated in these charts.

Some of the more reliable websites I've found for power figures are:

Computerbase, they usually have quite a few numbers in their reviews and their CPU/GPU numbers seem more accurate than many I've seen (example).
HWCooling (they have some interesting articles on rarely covered topics, like this)
Igorslab, though they're maybe more interested in graphics cards (exa
Tetras said:
I don't know what they used (because I haven't checked the test methodology), but there are different ways of getting this information.

For example, you can:
- Read it directly from the power connectors with your own equipment (I think gamers nexus use this method) and isolate the CPU from the board.
- Use software readings like from hwinfo, that report from the cores and package power.
- Use a plugin wall meter that takes it from the plug and subtract from this reading (e.g. PSU efficiency, motherboard and graphics card idles).

In reference to the question: why does the 13700K have higher idle power in the guru3d chart when the CPU idle draw is lower in tweaktown's chart, I would guess a few reasons:
- The nature of the readings and how they were measured (as mentioned above).
- The motherboard used (since it can make a big difference, especially a high-end one, which is what reviewers tend to use).
- How the BIOS / firmware was optimised.

I'd also reiterate again, most hardware reviewers just don't care about idle power and they don't worry about the test conditions. For example: they might even use a different graphics card, or PSU between reviews and this is not stated in these charts.

Some of the more reliable websites I've found for power figures are:

Computerbase, they usually have quite a few numbers in their reviews and their CPU/GPU numbers seem more accurate than many I've seen (example).
HWCooling (they have some interesting articles on rarely covered topics, like this)
Igorslab, though they're maybe more interested in graphics cards (example).

In all honesty, if your PC is spending 90% of the time idle and you really care about idle, you'd likely be better off buying a NUC or something like this and then just using the PC when you really need the power.
Click to expand...

mple).

In all honesty, if your PC is spending 90% of the time idle and you really care about idle, you'd likely be better off buying a NUC or something like this and then just using the PC when you really need the power.
Click to expand...

Thanks for the info. Regarding the Guru3D PC system idle power usage, assuming they were using equivalent motherboards for Intel and AMD, maybe Intel motherboards are more power hungry than AMD ones, therefore, overall Intel system idle power consumption is higher.
 
Last edited:
Nevara said:
Should the PC system power usage data be compared and not just the CPU power consumption, as this is a fairer comparison due to the differences in the CPU architecture? Assuming that the other components on the system are kept the same, but they must have done this to conduct a fair comparison test
Click to expand...

The problem with this question is that they're both meaningless, since the idle power of a system is so dependent on so many things that just the change of motherboard alone is enough to make most results meaningless. You can't test an AMD AM5 CPU and an Intel 1700 CPU in the same conditions, because you can't buy an identical motherboard, or optimise them both in exactly the same way.

All we can do is take ballpark figures, from looking at many different sources. Most hardware reviews are barely useful at all when trying to make a decision on idle power, because you need to isolate everything and be able to compare it. For example: the difference between two AM5 motherboards might be 10 watt and that's enough to move a CPU from the bottom of an idle chart to the top (or vice versa). As an example, take this review:

www.tomshardware.com

ASRock Z790 Steel Legend Review: Legendary for Budget Builds

An affordable, well-equipped and good-looking Z790 option around $230.
www.tomshardware.com www.tomshardware.com

ASRock's B760M Sonic reports an idle of 55 watt, the Z790 Steel Legend is 65 watt and the Gigabyte Z790 Aorus Master is 84 watt. If in an Intel CPU review, they used the Z790 Aorus Master, it would probably make the 13th gen CPU reviewed the worst CPU in the test. My guess is that this result is not even correct and the Gigabyte board had an early/poorly optimised BIOS. I will say though, that ASRock are usually one of the best when compared against other brands and this is consistent across many reviews.

The best source is often a forum thread on NAS/server forums, like the kind I posted above, where many users post their results and put them in a table, so that you can see the real world differences between CPUs, motherboards and other components.
 
To clarify, I never leave my PC on with no programs running, I always place the PC in hibernation or shut it down when it is not in use.
Would any of the following situations be classed as the PC being idle or under light load?
  • Around 10 tabs open in a web browser.
  • Several tabs open in a web browser, plus a few documents open in one or two office programs.
  • Visual Studio or MATLAB or a circuit simulation program open, plus an office program, and some tabs open in a web browser.
Plus antivirus and OneDrive running.

Many thanks
 
Last edited:
Nevara said:
Would any of the following situations be classed as the PC being idle or under light load?
  • Around 10 tabs open in a web browser.
  • Several tabs open in a web browser, plus a few documents open in one or two office programs.
  • Visual Studio or MATLAB or a circuit simulation program open, plus an office program, and some tabs open in a web browser.
Many thanks
Click to expand...

All of them are idle/light load, except if you're actually using the CPU to do something, e.g. compiling.
 
It seems like my best option is the 13700K as I don't use any software that would require the better multi-core performance of the 7950X. I also read that the 13700k has better RAM compatibility than the 7950X.
I just need to decide whether to wait and see what the price of the 14700K is in October, or if the 13700K is reduced much when the 14700K is launched. Information about the 14700K, showed that the multi-core performance will be very close to the 7950X because of the extra E-cores, plus single core performance is slightly improved.

Is this comment correct from here:
https://nanoreview.net/en/cpu-compare/intel-core-i7-13700k-vs-amd-ryzen-9-7950x#comments
"13700k way better purchase. Better for gaming, overclocking, and ram support. Beats 7950x in most single-threaded work with a much better igpu too. In fact, in most multi-threaded work that uses 8 cores or less, the 13700k should smack the 7950x."
 
Last edited:
Nevara said:
Is this comment correct from here:
https://nanoreview.net/en/cpu-compare/intel-core-i7-13700k-vs-amd-ryzen-9-7950x#comments
"13700k way better purchase. Better for gaming, overclocking, and ram support. Beats 7950x in most single-threaded work with a much better igpu too. In fact, in most multi-threaded work that uses 8 cores or less, the 13700k should smack the 7950x."
Click to expand...

Better for gaming: they're pretty similar. RAM optimisation alone is probably enough to swing it either way. I consider 13th gen (13600K and up) and Zen 4 to be equivalent. The X3D CPUs are a step above both.
Overclocking: I don't know.
RAM support: I don't know, but DDR5 compatibility/support is changing all the time, so what's true last month might be wrong now.
Single-threaded work: well, kind of. For example: if you look at Puget's benchmarks, the 13th gen CPUs won in Adobe (known for being single-thread oriented), but the RAM used was only 4800 for both platforms and AMD generally loses more performance than Intel with slow RAM. Their single core Passmark score gives the 13700K a slight (1.7%) lead, but like with games, I suspect RAM optimisation could swing this either way.
Multi-threaded work: true, the 13700K does beat the 8-core AMD equivalent (7700/7700X), but what heavily-threaded software only uses 8 cores? I find it a strange statement. When all the cores are used, the 7950X usually wins and is more efficient (at least, at stock settings).
IGP: as far as I'm aware, their media capabilities are pretty similar (in terms of codecs), but Intel supports AV1 encode and RDNA 2 does not. I've never seen any benchmarks which compare Intel's QuickSync with AMD's IGP. Gaming performance: I haven't seen enough benchmarks to say, but I do know they're both bad (struggle to maintain playable frame rates @ 1080p/low in newer games) and are both convincingly beaten by a GTX 1630. I don't know anything about their overclocking capabilities or how they're affected by RAM speed.
 
Tetras said:
Multi-threaded work: true, the 13700K does beat the 8-core AMD equivalent (7700/7700X), but what heavily-threaded software only uses 8 cores? I find it a strange statement. When all the cores are used, the 7950X usually wins and is more efficient (at least, at stock settings).
Click to expand...

I found this information about multi-core workloads from here:
https://www.cgdirector.com/multi-core-performance-cpu-list/#Which_Workloads_are_dependent_on_high_Multi-Core_performance

"Here are some popular workloads that can be well-parallelized, meaning they scale extremely well across any amount of cores your throw at them:
  • CPU 3D Rendering
  • Video Encoding
  • Batch Processing
  • Machine learning
And here are workloads that can be partly parallelized but don’t usually scale well across more than 4-8 or so cores:
  • PC Gaming
  • Physics Simulation
  • General Productivity
  • Active work (interacting with a Software’s GUI)"
Do you think the software I need to use would fall into the 4-8 core category?
 
Last edited:
Nevara said:
I found this information about multi-core workloads from here:
https://www.cgdirector.com/multi-core-performance-cpu-list/#Which_Workloads_are_dependent_on_high_Multi-Core_performance

"Here are some popular workloads that can be well-parallelized, meaning they scale extremely well across any amount of cores your throw at them:
  • CPU 3D Rendering
  • Video Encoding
  • Batch Processing
  • Machine learning
And here are workloads that can be partly parallelized but don’t usually scale well across more than 4-8 or so cores:
  • PC Gaming
  • Physics Simulation
  • General Productivity
  • Active work (interacting with a Software’s GUI)"
Do you think the software I need to use would fall into the 4-8 core category?
Click to expand...
Most software has a main thread and a GUI thread, the main thread is the bottle neck where single core performance is king. Engineering software will probably use more threads: main, GUI and workers. Renderers typically use all the cores available, so more cores get the job done faster.

Visual Studio will run fine on anything, the only time I have seen it use more than 50% is when I do a release build for a MAUI mobile app, then it hammers the CPU/RAM but only for a minute or so.

Any midrange or above CPU will be fine and probably a massive upgrade from what you are using now. For AMD, a 7700(non-X) or 7900(non-X) would be more than enough and you could do a drop-in CPU upgrade two years later for a nice easy BIG upgrade if you feel the need. For Intel, probably not going to be any big drop-in upgrades. For a work PC, I would not get the fastest RAM as it costs more and does not give much, 5200/5600 seems to offer the best value and most CPU’s can run it without problems.
 
Nevara said:
Do you think the software I need to use would fall into the 4-8 core category?
Click to expand...

My experience is that if you're running heavily multi-threaded software that needs as many as cores as you can throw at it, then you will know, because it will be a very painful (best case: make a cup of tea, worst case: watch game of thrones from start to end) kind of scenario.

In theory, some of what you have mentioned could be very heavy indeed, but if we're talking about undergraduate level work, I'm inclined to say that it's unlikely they will expect you to work with models/software that take many hours or even days to complete. More likely they'll be small examples that just allow you to apply the principles. If you know anyone that has done the course recently (of if they have student forums), I'd ask.

In terms of things like Visual Studio, I agree with FredFlint, most of the time it is idle apart from a few minutes while you're compiling or searching.
 
Tetras said:
In theory, some of what you have mentioned could be very heavy indeed, but if we're talking about undergraduate level work, I'm inclined to say that it's unlikely they will expect you to work with models/software that take many hours or even days to complete. More likely they'll be small examples that just allow you to apply the principles. If you know anyone that has done the course recently (of if they have student forums), I'd ask.
Click to expand...

^^^ This. So much this.

Ignoring all the power usages etc ... my first thought on reading the specs was that you're possibly shooting for too high a spec unnecessarily, and you could realistically wind it back a lot, save a chunk of money and still achieve plenty of performance sufficient for what your degree work likely needs.

As above, I'd be surprised if a degree required full works station spec machine to complete the work ... they need students to be able to complete the course work, and to expect them to have high-end machines all round is not realistic I would say, thus the work needs to be able to completed on non-top end machines.

If you looked at previous generations of processors, you could probably get the work done in a similar timeframe range, for far far less. Wait a few minutes and save £££ would be worth it for me.
 
You must log in or register to reply here.
Back
Top Bottom