Forged Alliance Forever

Re: Please Help - CPU Recommendation

2017-04-20T12:42:21+02:00

High end chips can process more per tick due to architecture cache size ect so frequency maybe similar but performance may vary.

Statistics: Posted by Gr1nder — 20 Apr 2017, 12:42

Re: Please Help - CPU Recommendation

2017-04-19T23:50:18+02:00

Citation needed. To be blunt, I simply don't believe this yet.

Well icedreamer, since you haven't cited anything yourself........ to be blunt I would appreciate you not being a knowitall I am pulling from passmark benchmarks for the results I posted in direct response to you, and for the general statements I made in that post, the information is coming from Wendel's testing and articles I have read.

Statistics: Posted by BRNKoINSANITY — 19 Apr 2017, 23:50

Re: Please Help - CPU Recommendation

2017-04-19T16:03:02+02:00

At the very least ryzen caused a dramatic drop in prices

Statistics: Posted by TheKoopa — 19 Apr 2017, 16:03

Re: Please Help - CPU Recommendation

2017-04-19T12:19:12+02:00

BRNKoINSANITY wrote:
Also, Ryzen overclocking isn't exactly linear, I think we have a long ways to go before we see the peak of ryzen first gen. The way the architecture works, it gets greater gains from frequency increases and specifically ram frequency increases than would be indicated by percentages.

Sorry but the ram thingy is bullshit as ppl are forgetting that you can also put faster ram in I5/i7 rigs also enchacing the perfomance by couple %
https://www.purepc.pl/procesory/test_procesora_amd_ryzen_5_1400_konkurent_intel_core_i5_7400?page=0,42
https://www.purepc.pl/procesory/test_procesora_amd_ryzen_5_1400_konkurent_intel_core_i5_7400?page=0,43
Sorry for it being in polish but this is a site I trust the most when it comes to CPU thingy. Also you should have no problem reading the graphs.

About ryzen overclocking capability, it's like Ice said. The current gen of zen arch won't go faster than 4.2-4.3 ghz per core as it's seems that that's the limit for air and "normal" voltage. But in one year we will see another gen of the zen arch and then we should expect some IPC improvments and maybe clock's of 4.7/4.8 on air but not on the current Gen. In current gen what we cal hope for is better bioses as it was already said that bioses are not even polished half as good as the ones, and not everything is working as supposed to be.

Also when it comes to the best single threaded perf per clock I would place my bet's on the Broadwell arch

Statistics: Posted by Endranii — 19 Apr 2017, 12:19

Re: Please Help - CPU Recommendation

2017-04-19T10:50:08+02:00

BRNKoINSANITY wrote:
I understand that the high end i5s and i7s can beat the xeons if clocked very very high, but why is it then that the kaby lake e-1280v6 xeon chip at 3.9 has faster single thread performance than the i3 7300 at 4 ghz, and the 7700 and 7300, an i7 vs i3, the i7 has .4 ghz less frequency and the exact same single thread score?

I definitely overstated the difference, and I am sorry for that, but to say that all CPUs in the same brand in the same generation have exactly the same architecture and therefore the same speed at the same frequency is flat out wrong. Maybe it has more to do with cache available, or something else that is different between the chips, but there is DEFINITELY a difference.

Citation needed. To be blunt, I simply don't believe this yet. All testing I have seen indicates that in single-thread benches, one architecture is uniform across the board clock for clock. The one conceivable difference is the one you point out, that both an i7 and a Xeon have higher cache amounts which will have a small impact in a limited number of synthetic test scenarios. Such differences will not have any impact in games, which tend not to be cache-limited. Additionally, cache-driven differences in benchmarks are usually within margin of error for testing (I really wish review sites would use proper standard deviation bars...).

BRNKoINSANITY wrote:
If I am not totally mistaken the fastest times ever tested in supcom (done on the forum some years ago to test all the rigs we possibly could in the same simulation and test real time completion) was on a xeon chip, and not at "that" impressive of a frequency. I could be remembering wrong though,and I don't feel like digging up that long dead thread xD

The Xeon in question was at my university at the time, a pet project of one of my professors. If I recall, it was an Ivy Bridge Xeon clocked at 5.9GHz on LN2 . Not a standard chip by any measure haha, but by no means at all would I expect it to have done any better than an IB i7 at that kind of clock.

BRNKoINSANITY wrote:
Also, Ryzen overclocking isn't exactly linear, I think we have a long ways to go before we see the peak of ryzen first gen. The way the architecture works, it gets greater gains from frequency increases and specifically ram frequency increases than would be indicated by percentages. I think that 5ish percent is going to shrink in the next few months.

Agreed that it will shrink, but not by any core clock increases, by other methods that you've mentioned. Zen 1 is hitting the process limit at 4.2GHz on the very, very best chips. People have pushed beyond that, but only on LN2.

Statistics: Posted by IceDreamer — 19 Apr 2017, 10:50

Re: Please Help - CPU Recommendation

2017-04-19T07:34:20+02:00

"the fastest times ever tested in supcom (done on the forum some years ago to test all the rigs we possibly could in the same simulation and test real time completion) was on a xeon chip, and not at "that" impressive of a frequency. I could be remembering wrong though,and I don't feel like digging up that long dead thread xD"

ah lol yes you mean the guy who claimed to have oc-ed his schools xeon server on h2o to 5 ghz, i thought everyone realised that was obvious bs.

Statistics: Posted by ZOB — 19 Apr 2017, 07:34

Re: Please Help - CPU Recommendation

2017-04-19T02:40:17+02:00

Also, Ryzen overclocking isn't exactly linear, I think we have a long ways to go before we see the peak of ryzen first gen. The way the architecture works, it gets greater gains from frequency increases and specifically ram frequency increases than would be indicated by percentages. I think that 5ish percent is going to shrink in the next few months.

Statistics: Posted by BRNKoINSANITY — 19 Apr 2017, 02:40

Re: Please Help - CPU Recommendation

2017-04-19T02:38:24+02:00

I understand that the high end i5s and i7s can beat the xeons if clocked very very high, but why is it then that the kaby lake e-1280v6 xeon chip at 3.9 has faster single thread performance than the i3 7300 at 4 ghz, and the 7700 and 7300, an i7 vs i3, the i7 has .4 ghz less frequency and the exact same single thread score?

I definitely overstated the difference, and I am sorry for that, but to say that all CPUs in the same brand in the same generation have exactly the same architecture and therefore the same speed at the same frequency is flat out wrong. Maybe it has more to do with cache available, or something else that is different between the chips, but there is DEFINITELY a difference.

If I am not totally mistaken the fastest times ever tested in supcom (done on the forum some years ago to test all the rigs we possibly could in the same simulation and test real time completion) was on a xeon chip, and not at "that" impressive of a frequency. I could be remembering wrong though,and I don't feel like digging up that long dead thread xD

Statistics: Posted by BRNKoINSANITY — 19 Apr 2017, 02:38

Re: Please Help - CPU Recommendation

2017-04-18T23:59:43+02:00

RocketRooster wrote:
/SNIP

Oh what a fantastic post! You, Sir, have no idea how nice it is for me to see there's someone else around who really knows his stuff now. Great response

BRNKoINSANITY wrote:
I think a common misconception is that the frequency of the processor (ghz) has everything to do with single thread performance. While this was true years ago, and even between brands, it is no longer the case at all. A high end xeon processor at 2.2 ghz can stomp the single threaded crap out of a lot of i5s at 75% higher frequency.

Wish I could say the same here While the rest of your post is completely on point, this bit is a bunch of crap.

- Intel's CPUs from i3 all the way to Xeon, in a given generation, run the same architecture. A Sandy Bridge Xeon @ 2.2GHz runs identically in single threaded performance to a Sandy Bridge Core i3 @ 2.2GHz.
- Intel's single thread/IPC has improved only marginally over the past 5 years. This is why the 2700K is still considered a good chip by many.
- There is no sensible situation in which a Xeon will outperform an i5 at single threaded tasks. Situations where this may occur are limited to: Programs requiring Xeon-specific instruction sets/hardware features (You are not running these at home, ever), and a comparison of the very latest Xeon (Kaby Lake, 4C8T) to a much, much older i5, and even then the Xeon is running higher clocks, not lower. In fact I cannot think of ANY situation in which a 2.2GHz Xeon of any generation would beat a 3.85GHz i5 of any generation.

Summary - Since across Intel, platform IPC hasn't moved much, frequency IS in fact a very good measure of relative single thread performance, even across a sensible number of generations. Additionally, AMD's new Ryzen core appears to perform roughly on par, to within about 5%, clock for clock, losing single-thread benchmarks because it doesn't clock as high as Kaby Lake. So yeah, we're now back at a point where frequency is a good indicator of single thread performance, across both generation and vendor.

Statistics: Posted by IceDreamer — 18 Apr 2017, 23:59

Re: Please Help - CPU Recommendation

2017-04-18T21:30:19+02:00

I think a common misconception is that the frequency of the processor (ghz) has everything to do with single thread performance. While this was true years ago, and even between brands, it is no longer the case at all. A high end xeon processor at 2.2 ghz can stomp the single threaded crap out of a lot of i5s at 75% higher frequency. On the same note, an FX AMD cpu running at 5 ghz will still lose in single threaded to an intel CPU at 4.2 ghz.

As to the wattage of a CPU, the mobile U series chips are no longer binned desktop chips, they are specifically manufactured extremely low power chips, and are not capable of the speeds (generally) or core counts (mostly) of desktop chips. They are very impressive in their own rights, but can not compare to a "real" cpu.

Supcom is a single threaded task, not locked to a single core. A hyper threaded dual core is NOT the same as a quad core CPU. Basically, with more cores you are not necessarily gaining speed directly on SupCom, but you ARE giving your PC more places to stick random crap. Yeah, your game is basically taking up two cores (grossly oversimplified) but your operating system, windows update, that webpage in the background, etc are not also fighting for those same two cores. The experience generally gets smoother the more cores you have. (Ryzen 7 is fantastic for workstations for this reason).

Long story short, Dual cores are worthless moving forward in the computer world, Higher watts generally give you a higher clock/performance threshold, and you should probably also look for something with dedicated graphics in the vein of a 1050 if you are going to be gaming on this computer.

Statistics: Posted by BRNKoINSANITY — 18 Apr 2017, 21:30

Re: Please Help - CPU Recommendation

2017-04-18T20:32:12+02:00

One thing you have to understand is that processor capability for each individual chip of a given model is a crapshoot. The CPU makers have clever people and algorithms that optimize a given cpu node and stepping in terms of cost/performance. They will try and make as many chips at this optimum as they can, but not all chips make the grade. Some cpu dies may have defective transistor or interconnect here and there, so they are able to switch off sections of the CPU and sell it as a lower spec model.

Sometimes a specific die needs higher voltage to run stably, and runs hotter. It all depends. The chipmakers bin their cpus according to performance capability - the crap ones become celerons or are sent back to be recycled.

In the past it used to be the case that mobile cpus were the ones that were the most stable at a given performance level at the lowest possible voltage and wattage. I'm not sure if the binning process works that way anymore, since these days CPUs run turbo boost and thermal throttling - the clock speed will boosted if the cpu is kept cool enough or throttled below stock speed if it becomes too hot.

Laptops are notorious for having relatively poor cooling given the form factor constraints and you won't get the benefits of boosting as much, if at all. It's impossible to say how well your chosen model of laptop will be engineered or how well fabbed the cpu is, but it is possible that a "slower" CPU with the same boost clock as a higher tier item might perform the same. The only way to know is to spend money and try it...

Where core count and IPC is concerned, remember that the sim speed will largely be determined by the slowest PC taking part in a team game.

And people have the wrong idea about core counts. The operating system has dozens, if not hundreds of threads running concurrently. Even if your game is only single-threaded, the operating system must still timeslice and context switch between processes and their threads. This switching has a performance cost and the more cores you have, the easier it is for the scheduler to give all threads a slice of the pie and the less it is necessary to context switch away from the game running in the foreground, even if it is single-threaded. Yes, it isn't a substantial increase compared to dedicated multi-threaded use by the game, but there is a noticable reduction in stuttering and everything just feels smoother.

It's anecdotal, I know, but my previous PC was a 2.4ghz Kentsfield Core 2 Quad. Vanilla supcom gave it a hard time, and it was worse on a core 2 duo yorkie at 2.8GHz I'd swapped it out with as a test.

Don't waste your cash on dual core. That ship has sailed. If it must be a laptop, get the fastest Intel quad core you can afford.

Statistics: Posted by RocketRooster — 18 Apr 2017, 20:32

Re: Please Help - CPU Recommendation

2017-04-18T17:17:50+02:00

Thanks for all of the recommendations and tips - they are very helpful.

I forgot to mention that I am considering the T570 because of the larger screen size, longer battery life, USB-C port, and lower price.

But I am quite confused on the watts & power envelope: I thought that if it is lower, than it is more efficient at processing a certain task. However, the comments mention that higher is better? Is that only true if I have a powerful cooling solution? (It will be a laptop, so I am not certain to how cool I can keep it). And any recommendations on what will keep a laptop running cool will be highly appreciated.

I have also read from other forums how important Single Thread performance is for this game, and that FA is bound to one core (maybe up to two). If this is the case, will either the i7-7600U or the i7-7820HQ make much a difference as their Single Thread performance is also the same? (both have a Frequency from around 2.8GHz to 3.9GHz).

Once again, I thank everyone for the great advice and tips.

Statistics: Posted by JakobeBeef — 18 Apr 2017, 17:17

Re: Please Help - CPU Recommendation

2017-04-18T02:38:29+02:00

He said it was going to double as a work laptop, so...... desktop is probably not an option.

Definitely the HQ processor of any type. Like was said above, the power/thermal limit is low, and the core count is low on the U series processor. It is entirely possible to play FAF on a U series. I have played an entire, 1+ hour setons game on my surface pro 4, and was not the slowest person (235ish cpu score and never thermal throttled). However, there are noticeable stuttering problems if you have multiple processes running, encoding or heavy work tasks run like crap, and the processor doesn't feel "snappy" due to the fact that it only has two cores, and the i5 part is 12 watts max (i7 may be 15? idk).

Long story short, get the HQ. My current laptop has one the 6300hq in it and I am completely happy.

Statistics: Posted by BRNKoINSANITY — 18 Apr 2017, 02:38

Re: Please Help - CPU Recommendation

2017-04-16T22:14:46+02:00

TheKoopa wrote:
Why not build a pc for 1/2 the price

I was thinking the same, but some guys like to be able to tote their rigs around. Plus he needs it for work I suppose so there's that.

Personally I'd get a cheap celeron crapola jobbie for graft and build a killer rig for the fun stuff.

Statistics: Posted by RocketRooster — 16 Apr 2017, 22:14

Re: Please Help - CPU Recommendation

2017-04-16T22:09:30+02:00

Why not build a pc for 1/2 the price

Statistics: Posted by TheKoopa — 16 Apr 2017, 22:09