The TIM isn't that big of a deal. If you are into overclocking and tweaking and buy the K series you should be delidding the CPU anyway. I helped build three 7700k machines in the past 4 months and two of those were delidded. There was no difference in how far we could overclock and the temps between machines were with 5-8c. The only big difference I saw was when running prime 10-12c difference but, you don't buy a 7700k to run prime lol.
Funny you should say that, exactly when I was looking at this:
http://www.pcgameshardware.de/Coffe...8700K-i5-8600K-i5-8400-Test-Review-1240339/2/
Witcher 3 frame times:
RoTR DX12:
HT seems to make a difference even with 6 cores, or so it seems.
Funny you should say that, exactly when I was looking at this:
http://www.pcgameshardware.de/Coffe...8700K-i5-8600K-i5-8400-Test-Review-1240339/2/
Witcher 3 frame times:
RoTR DX12:
HT seems to make a difference even with 6 cores, or so it seems.
Would you call these 'statistical outliers' though? The vast majority of gaming benchmarks I've seen comparing the 8600K and 8700K don't show such scaling with HT - in fact sometimes there is slight performance regression from HT (not sure if its a BIOS turbo bug for the 8700K?)
Also keep in mind the 8700K boosts to 4.3GHz compared to 4.1GHz for the 8600K so there is ~5% frequency discrepancy, so its not a completely 'apples to apples' comparison.
Yeah, different clocks and different cache size. Games can be very sensitive to cache.
I'd like to see 8700K locked speeds with and without HT to really see the difference with the confounding factors removed.
What Intel would need to do is make sure that single core turbo is >= 4.7GHz so that people with beefy cooling are "guaranteed" all-core 4.7GHz on 8 cores. I don't think people like making the "more cores" or "high per core performance" trade off (though I could be wrong about what "people" in general want).
This will be possible, but Intel really has to do something about the terrible TIM they've got on all its parts. It really took the wind out of Skylake X's sails (the silicon has good frequency potential, but good luck cooling it at high frequencies; it's tough), and it could mask the true potential of the 8 core CFL part.
Nobody runs Prime95 daily...except maybe those number theorists looking for new prime numbers
No, but the purpose of an HT chip is for the gains in large parallel processing. If your using it effectively the chip is going to use programs that make it only a slight bit less toasty. 4.8ghz doesn't remind me of ease (especially with cases of throttling above that point based on performance anomalies.). 4.8ghz is likely the max OC for most due to heat wall and less than golden chips without a delid. To get 5ghz you'll need a mix of three things: delid, golden chip, or Top of the line cooling. A careful look at the more prolonged benchmarks show they used AIOs like 115 and 110, some krakens, and for air only nh-d15. Those all carry a premium to your chip purchase.
Kaby lake when it was released showed 5ghz as a GUARANTEE in early benchmarks. Temps were lower at 5ghz, and it was easy to reach. However, a look at run of the mill DIYers showed that when paired with fast ram many only felt safe at 4.8ghz due to temps. 8700k will be similar. I wouldn't be surprised to see that 4.7ghz multicore is where most DIYers stop. 8600k should reach 5.1ghz or 5.2ghz with ease, which is why its the chip to get for gaming and OCing not the 8700k.
So I now have more reasons to stand by my point that 8700 non-k, 8600k and 8400 are the coffee lake chips to get. 8350k and 8100 might be ok due to price, but being kaby lake on coffee platform means that more data is necessary to determine if they will be worthwhile.
---------------------------------------
On another note:
From a tangible standpoint, for the few who state otherwise: These chips are good. However, they do not obsolete all previous intel architectures or ryzen. It will eat into some of skylake-x and some sets of kabylake. However, a lack of difference in standard tasks mean many average persons won't have a preference. Theoretical numbers mean nothing if it doesn't show differences in mainstream use. Due to many apps being bottlenecked, or that show mixed loads where low core count is best, means there is no visual/interactive gain over kabylake.
8700 will give the best MT for its price of Coffeelake or skylake-x, however comparisons to ryzen are difficult. The best chips from ryzen are the 1600, and 1700(which incidentally can also be OC to match their more expensive counterparts). These are cheap chips, which can be equipped on super cheap boards(as low as $50), which provide a power envelope that is untouchable at stock, while maintaining good performance. While many build servers and render farms using server chips, a massive low power draw, low cost farm built purely from said 1600 and 1700 already have a power/efficiency/cost envelope that presents more value in unison than even the larger workstation and server chips from intel and amd. As a result if you don't need large amounts of ram on a single board, ECC ram guaranteed to work, or more specialized work made to use resources only on a local board -> it would be better to make farms from primarily the 1700 than any other chip that currently exists.
Coffeelake does not change this variable. The board/cooling/chip/power costs versus performance cannot beat that chip when purchased in bulk running at stock for unattended parallel tasks.
Where coffeelake slots in for sales is in Mixed workloads. It will provide both good ST and MT. Before you either had to pay a premium for intel workstation, choose the 7700k for Good ST and ok MT, or buy ryzen for ok ST and Good MT.
Good chips
For intel: 8700, 8600k, 8400, 7700k(on sale, or still an OK purchase), several Pentiums, and 7940x are the good chips. For AMD: 1700, and 1600 are still good (1600x is OK expense).
Affected Chips
7700, 7600k, 7500, 7400, 7350k, 7300, 7100, 7800x, 7820x, 1700x, and 1800x are primarily the chips which will have pressure placed upon them. AMD, could still drop 1700x and 1800x to be a mere $25 and $50 over 1700 and they can escape a fair amount of the negative effects, as 1700 is already creeping under $300.
Also as note: over at HARDOCP which has been probably the most gungho of enthusiasts for coffee lake up to this point are all crying foul. They are probably the most pissed at the coffee lake release, with many sharing negative comments. As a result, I'm not entirely sure how the architectures perception actually is in all enthusiast circles. I don't see the release as negative like them, but I'm also trying to be a realist and consumer first about each chip.
But the reality is, Coffee Lake is THE CPU for enthusiasts.
None of the major sites do this, not the English press at least.Lol no...you don't speak for enthusiasts...
Has any review site done a clock for clock comparison yet?
So much negativity in one post.
I read the forums at HARDOCP as well, and I don't see any big wave of negativity. It is essentially the same reactions as here. Some minor griping about needing new MBs or paying over MSRP.
But the reality is, Coffee Lake is THE CPU for enthusiasts.
Lol no...you don't speak for enthusiasts...
Has any review site done a clock for clock comparison yet?
Versus what?
But some people DO run 100% load 24/7/365 (like me)Nobody runs Prime95 daily...except maybe those number theorists looking for new prime numbers
Does your F@H software use AVX?But some people DO run 100% load 24/7/365 (like me)
The CPU client is BOINC, and I have no idea about AVX.Does your F@H software use AVX?
Clock for clock co
mparisons...lol. That would be important if one was not running ~25% higher clock.
The CPU client is BOINC, and I have no idea about AVX.
My point was, that some people do use their CPU's@100%, and as such, running over like 75c for long periods or there abouts is NOT good.
I knew what you were saying...I was just curious. I was thinking about Folding for fun. Maybe I will fire it up and see what kind of sustained temps my setup runs.The CPU client is BOINC, and I have no idea about AVX.
I did say thereabouts. Just for me, anything over 75c makes me nervous, and I would never run anything 24/7@80c or higher. I think this started with a mention of 86c (I could be wrong) and thats WAY too hot. Also I have heard mention about throttling at these temps or close to it.I'm actually interested in how you came to the 75C figure? I mean, I agree that personally, I wouldn't be comfortable with my CPU running at 80C+ for extended periods, but is there actually any evidence that running CPUs at high temps (but under TMAX specs) would severely shorten the lifespan of a CPU? Will a CPU really last *that* much longer running at say, 70C rather than 90C?
I'm too chicken to delid a CPU
Something better than anecdotal evidence, a review by a trusted source: https://www.techpowerup.com/reviews/Intel/Core_i5_8400/17.html
I would be too, but on GamersNexus video, they said the Rockit tool was pretty easy to use.
EDIT:
I'll probably buy the tool and test it on some older 1150 CPUs. I've got a Pentium G4400 that I won't shed too many tears over if I break it. It's only $40-50 to get a new one anyway. (Far better than a back-ordered $380 processor. )
Its interesting from an intellectual point of view, but yes, it's not a very practical comparison, as real world frequencies are approximately 25% higher on the Intel chips, both at stock or overclocked form.
Agreed. That would be the only way to objectively measure the impact (or lack thereof) of HT for gaming with full accuracy.
However, judging by the 12 game average on Techpowerup, there is almost no difference in overall gaming performance between the 8600K and 8700K:
https://www.techpowerup.com/reviews/Intel/Core_i5_8600K/18.html
That 3% difference is probably more down to the fact that the 8700K turbos 200MHz higher than any advantages HT provides, but again, as you said, the best way to know for sure is to toggle HT on/off on a 8700K and measure the differences.