Personal computing discussed

Moderators: renee, Flying Fox, morphine

1700x or 8700k

Poll ended at Mon Oct 09, 2017 12:20 pm

8700k
28 (78%)
1700x
8 (22%)
 
Total votes: 36
 
just brew it!
Administrator
Posts: 54500
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: Which cpu to get?

Sat Oct 14, 2017 12:30 pm

Redocbew wrote:
Yeah I suppose that's true. There are good reasons to dislike SMT, but in general it seems like a good thing to me. The effects of SMT(namely having virtual cores which otherwise wouldn't be there) are much more easily visible than that of branch prediction or many other features which I guess is part of the problem. If it just sorted of faded away into the background, then it probably wouldn't have received so much attention. The segmentation games of using SMT to help define the differences between an i3, i5 and i7 probably haven't helped with that either, but those are only quasi-technical things. As a developer this is all stuff way outside my specific field, so I imagine I might be a bit more measured in my opinion if I had to deal with the corner cases where SMT is problematic. Overall it seems like most people designing chips these days have decided that the pros of SMT outweigh the cons, and since I'm not one of them I'll just have to take their word for it. :P

For me, the biggest headache with SMT is that it makes performance more difficult to predict for threaded workloads, since the contribution of those extra virtual cores is very dependent on the code being executed, cache utilization, etc.

A non-trivial part of my day job these days is performance analysis and optimization of server code for large-scale data storage systems, just to give you an idea of where I'm coming from here...

I guess we're getting pretty far off-topic WRT the question asked in the OP. :lol:
Nostalgia isn't what it used to be.
 
Topinio
Gerbil Jedi
Posts: 1839
Joined: Mon Jan 12, 2015 9:28 am
Location: London

Re: Which cpu to get?

Sun Oct 15, 2017 4:28 am

Krogoth wrote:
Topinio wrote:
Krogoth wrote:
Already did it and still came to the same conclusion. Outside of a small minority of mainstream applications and games, the benefits of going from dual-core to quad-core are still kinda small. The majority of the reviewers out there are cherry-picking games that are known to use more than two-threads (Doom (2016), Witcher 3, GTA 5, Battlefield 4/1). It is no shock that you see considerable returns, however they are the exceptions not the rule as far mainstream applications are concerned. The benefits of going beyond quad-core are even smaller with the aforementioned games unless you do CPU streaming at the same time.

From 2010, from arguably the most mainstream game of the decade.
[...]

WoW is still dual-threaded like the majority of gaming titles out there in the field. The only reason you see a small jump from gaining an extra gain is because of OS and other overhead in an "artificial" dual-core setup.

Even if you were correct that the majority of games including WoW were dual-threaded (and you are not) we still see a jump going to 4 cores. It is not a "kinda small" jump.

The GPU driver thread, and all the other threads on the system, have to compete with the (supposed) 2 game threads on a 2C CPU and so a 4C CPU produces kinda not small gains.

Dude, it is not 2006 any longer.

Krogoth wrote:
Just notice how there's no returns after going beyond three threads and the 6-core Phenom is only faster when it is unlocked because it has full access to L2/L3 cache pool at its disposal. "Artificially" gimping silicon's core-count can possibly restrict is access to its cache pool depending on CPU topology.

My emphasis, because

Krogoth wrote:
the benefits of going from dual-core to quad-core are still kinda small

is disproven by the bit where you admit there was benefit in going to 3 cores even in 2010 even for WoW.

Also, plenty of reviews show modern 2C chips getting stomped by old 4C ones, e.g. https://www.techspot.com/article/1313-i ... page2.html where the 2015'Q4 i3-6100 is comprehensively beaten by the 5 years older 2011'Q1 i5-2500K, though it does manage to beat the 10 years old C2Q's...

Krogoth wrote:
The reason why reviewers pick the outliers is because they want stuff that stresses the CPU. The reviews would be awfully boring if they stuck what the masses are playing these days (LOL, TF2, COD, CS:GO, DOTA, MMORPG stuff) not stuff that taxes current CPUs. The listed 99th percentile FPS ratio graph only includes the titles that do take advantage of more then two threads. If you were to throw in content that is still stuck in two-threaded world (What the masses still play to this day). All of the quad-core CPUs and beyond would be around a much tighter cluster distancing themselves away from dual-core chips by a small gap. The differences between the quad-core chips and beyond would be just price.

Reviewers review in reviews, news at 10. PUBG? The devs have optimized that for 6-core or higher CPUs, they obviously see a reason. Doom? Quake?? &c. &c.

You're not wrong that reviews are somewhat artificial, though. One thing reviewers typically do not do is replicate normal user behaviour, as they don't want loads of random crap running on the systems ruining reproducibility. IRL, though, having a 4C or higher CPU helps normal people with running their games, even those ones that have only 2 threads, because the average person does have all sorts of rubbish going on with their PC while gaming.
Desktop: 750W Snow Silent, X11SAT-F, E3-1270 v5, 32GB ECC, RX 5700 XT, 500GB P1 + 250GB BX100 + 250GB BX100 + 4TB 7E8, XL2730Z + L22e-20
HTPC: X-650, DH67GD, i5-2500K, 4GB, GT 1030, 250GB MX500 + 1.5TB ST1500DL003, KD-43XH9196 + KA220HQ
Laptop: MBP15,2

Who is online

Users browsing this forum: No registered users and 1 guest
GZIP: On