Technical Discussion
  >> Gaming Forum


Register (or login) on our website and you will not see this ad.


  Print Thread
Standard User Chrysalis
(legend) Mon 23-Apr-18 14:21:39
Print Post

ryzen2


[link to this post]
 
So some reviews have this en par or even in some cases beating intel CL chips.

Here is my thoughts since some people are discrediting the tests.

First I think the way reviews are currently done is flawed, people like adoredtv are trying to correct the industry but he is a lone fish in a big pond.

These guys should be buying retail hardware as anonymous customers, and as such not only not have favoured kit but also not be bound by "review guidelines". With that been said here is my view on whats going on.

I think with meltdown+spectre patched it seems logical to me intel's advantage would be noticeably reduced. As far as I am aware on Windows 10 the intel microcode update is distributed by microsoft so is expected to be in play on modern intel systems, but I believe for AMD a bios update is required so its inconclusive if the AMD systems are actually spectre patched. Meltdown does not apply to AMD. This is basically if I am right dependent on if the shipping ryzen 2 bios has spectre mitigation's.

In terms of how the chips were overclocked is the bit I am not happy with. It seems the reviewers with the exception of OC3D decided to not put intel manually overclocked chips on their data graphs, given intel chips overclock very well over the shipped configuration this obviously hurts the intel results. I do understand the merit to a degree, the merit been that AMD are willing to have an official turbo clock very close to the chips max capability and, intel not, this decision by intel seems to be hurting them as the reviewers are only putting intel clocks on that are within spec of the chip. Given I think its a reasonable expectation that a owner of a K series intel chip would manually overclock the chip I do think the manual OC data should be on these graphs (alongside the stock data).

So where is this data? well the reviewers will always supply it on intel reviews, the problem is in the "review guidelines" they will generally be asked to do things to make the product they reviewing look in the best light possible, which in this case is comparing AMD stock to intel stock. The data exists as it was done on the intel reviews, its just been omitted.

Even with this issue tho and even if the data was put in the graphs its clear the ryzen2 is a very good chip from AMD.

My new thoughts are this.

If you just want per core performance and threaded performance isnt a priority then I would consider intel i5 chips. Ryzen is much closer now but I think intel still has the edge.
However if you value overall threaded performance I think th ryzen 2 is better buy over the i7. So basically I now consider i7 in no mans land now. I expect the next i7 generation will be an 8 core chip, but unless its per core performance increases I think it will still have a problem as it will be priced higher than the r7.

Personally I think for best gaming chip also its really the i5 8600k vs the 2600x, the media is in a frenzy with 8700k vs 2700x, but the reality is the 8600k is just about as good as a 8700k for gaming but much cheaper, and likewise the 2600x is just about as good as a 2700x and £80 cheaper.

Sky Fibre Pro BQM - IPv4 BQM - IPv6

Edited by Chrysalis (Mon 23-Apr-18 14:24:05)

Standard User epyon
(fountain of knowledge) Mon 23-Apr-18 22:58:16
Print Post

Re: ryzen2


[re: Chrysalis] [link to this post]
 
The reviewers guidelines have been the same for a decade or more

it really needs to change but i would imagine that intel wouldn't like that so much and withhold review samples kinda like what happened to kyle at hard OCP and AMD


If you have first gen ryzen i don't see a reason to upgrade

heck even if you have a system 3-4 years old pc on an intel based platform there isn't a reason from a purely gaming stand point

the real ass beating will come on 7nm Ryzen 3000 series

intel is stlll struggling to get 10nm yields to an acceptable level while GF and TSMC both have 7nm on track for the end of this year if i remember correctly

i will probably jump on the 3700 or whatever they decide to call it

I've just been using a 2200G and x370 while waiting for the 3000 series since i only play at 1080p i really haven't experienced anything too bad my cpu is pin at 99% when playing most games but no stutters or random 0.1% lows.

hope amd keep their word and my x370 board will support the 300 series with a bios update.

Plusnet FTTC - 80/20 *TP-Link W9980*
Three - 4G *Huawei Honor 8*

Edited by epyon (Mon 23-Apr-18 23:20:09)

Standard User Chrysalis
(legend) Tue 24-Apr-18 20:49:11
Print Post

Re: ryzen2


[re: epyon] [link to this post]
 
witholding samples is not an issue if you buy your parts anonymously at retail which is what reviewers should be doing.

If you buy what you reviewing at retail, then you got no obligations to follow any guide of "review guide", you get retail parts the same as consumers, not engineering samples, or cherry picked of the production line.

Buying retail also you may get late enough to get a part that has had bait and switch of components happen, which if it affects consumers it should be reflected in reviews.

There has been a ton of problems I observed lately in reviews, as you say this practice has gone on for a while but its just something I have picked on more lately.

1 - Most Vega reviews witholding 1080ti from results, even tho vega 64 cards were in 1080ti price range, I argued with kyle on hardforum about this, his reasoning was that the retail price was been price gouged and that isnt AMD's fault. I said reviews should reflect reality for consumers, and his response was in lala land, we never came to agreement. Gibbo (ocuk staff member in control of stock) also responded to some of these claims saying if he sold at the prices hardocp claimed were what should be sold at then his company would be selling at a loss, suggesting this price gouging is not just at retail level.
2 - Overclocked intel CPU's witheld from most ryzen 2 test results, in addition features like MCE disabled for intel systems but equivelent features like PE enabled for ryzen 2.
3 - On the other side, reviewers claiming they testing intel at stock but actually testing with MCE enabled, giving inflated results.
4 - Also on intel, reviewers testing 8400 intel CPUs on high end ROG boards with TDP limits removed on a non overclocking chip.
5 - GN deciding to test gaming and streaming performance on a ryzen 2600x and compare it to the 8600k., this one really got my interest, mainly because its the first time I seen such a thing tested by anyone, secondly he was testing using uncapped game framerate and software based encoding, which is a highly unusual config that game streamers would use, I concluded GN did this test because he knew it would put the 2600X in the best possible light vs the 8600k, its the ultimate perfect workload for it. The problem I have with this is the reviewer went out his way to find a test to put the product he is testing in the best light possible, meaning that the reviewer feels he has a job to do that and hence bias. There is absolutely no doubt GN will not do these software based streaming tests when he next reviews an intel i5 chip, if he did he may get put on intels blacklist.
6 - When intel gets reviewed the intel OC performance results will be on the graph to help out intel, when AMD gets reviewed, they will be omitted to help out AMD.
7 - when the Nvidia maxwell cards got reviewed, people observed there was a very limited set of other products they were put alongside on the graphs, it was as if nvidia only wanted that visual comparison to be made, turns out people were right, pcper confirmed it was at nvidia's request.
8 - Certain products never get reviewed, simply because the company decides not to send review samples out of those products, ok this is not strictly true, smaller rviewers who actually have to pay for what they review, might buy these products, but generally they get skipped over by the well known reviewers, an example is low end motherboards, especially from asus, 5 years ago the mainstream range of asus boards used to get reviewed, but now they dont, perhaps its not surprising that feature wise they have been significantly downgraded and its not in asus's interest for these boards to get reviewed. OC3D recently discovered the asus z370 strix boards had broken LLC in their bios, he reported it, they ignored it, he waited, nothing happened, he published a video saying he cannot review the boards because they broken and all hell broke loose. Asus eventually fixed the bios, but it turns out all the asus prime boards had the same bug, however they did not get a bios fix for this reason stated by raja @ asus.
"The boards dont have the hardware spec for the feature to work" So basically asus sold overclocking boards that cannot utilise standard overclocking features because they were under spec'd.
9 - The games tested on reviews are games that are optimised for high core cpu's, the problem is only 8% of games (that have been analysed) on steam can use more than 2 cpu cores at once, yet 96% of games reviewed in the past 12 months can. This is clearly to push a message that cpu's with high thread counts are great for gaming, intel and AMD both want this pushed so it gets done, e.g. intel want people to buy i7's instead of i5's and AMD wants people to buy their thread heavy cpu's as well. However even with this bias on the reviews, 8600k's keep up with 8700k's and the 2600x keeps up with the 2700x.

There is more but I feel I have made my point.

Adored did an excellent video analysing the chances of getting a "golden sample" 8700k at retail based on the silicon lottery website statistics, silicon lottery is a website that tests loads of cpus, rates them for a specific speed, and sells tham for that speed. It gives "golden sample" something like a 3% chance of finding one. Adored analysed the cpu's reviewed by a bunch of reviewers and found more than half had "golden samples".

The reviewer in the past 12 months I have found with the most integrity is bizarrely OC3d. He talks to his viewers like they kindergarden But he is the only reviewer that did all of the following.

1 - Reported reviewers were using MCE on stock intel tests, and that he even tried to get the default changed to disabled (unsuccessfully). On his intel cpu graphs he has 3 results, stock, MCE and OC, so all 3 scenarios covered, including on AMD reviews.
2 - He had the 1080ti on his vega review graphs, he didnt talk about it, but it was at least on the graph.
3 - He came out with the broken strix motherboard video, and also done another recently for a gigabyte board.
4 - He said clock speed is king in games on his AMD review, I dont think AMD would have liked that very much, but its an accurate statement.

Sky Fibre Pro BQM - IPv4 BQM - IPv6

Edited by Chrysalis (Tue 24-Apr-18 21:00:07)


Register (or login) on our website and you will not see this ad.

Standard User Chrysalis
(legend) Tue 24-Apr-18 21:11:43
Print Post

Re: ryzen2


[re: epyon] [link to this post]
 
Back to ryzen.

Right now I think if you want threaded performance and your budget is in the i7 price range, then buy a ryzen r7.
If you just want gaming performance, and your budget reflects it then get an 8600k, or if budget not quite there a 8400 or 2600.

This leaves the i7 sort of lost.

Intel are very likely going to bring out an 8 core chip next. I expect given the way most reviewers ignored the advantage of high overclocking on intel chips, we may also see higher stock supported speeds from intel, they releasing a promotional chip soon, which on paper is basically an i7, but with a supported turbo clock of 5.1ghz. So in stock it will clock to 5.1ghz, thats mighty.

Like you said intel do seem to have a yield problem, OC3D reported the reason intel is not soldering their chips is because its lowering yields, the fact they selling chips with huge buffers on clock speeds as well suggests a low confidence in making those speeds stock, possibly also related to yields.

AMD however also have things up their sleeve, their next fabric process will support a target speed of 5ghz, they may not reach it but they should at least get much closer to it. If we see Ryzen 3 chips at say 4.6-5ghz then intel have a big problem. I posted my single core cinebench score on ocuk forum earlier it was 207, a ryzen 2 owner managed to get after tuning in his bios 186. Thats getting close, my i5 is at 4.8ghz, his single core clock is 4.5ghz. If his chip was at 4.9ghz a 10% boost in clocks then his score would be extremely close to mine meaning the IPC is very similar. The intel advantage now is mostly down to clock speeds.

Sky Fibre Pro BQM - IPv4 BQM - IPv6

Edited by Chrysalis (Tue 24-Apr-18 21:12:16)

Standard User epyon
(fountain of knowledge) Wed 25-Apr-18 01:18:59
Print Post

Re: ryzen2


[re: Chrysalis] [link to this post]
 
I pretty much agree with everything you said

when it comes to reviews i'm always ignoring "big" youtube reviews like jay2cents,linus

jay is probably one of the more incompetent people i've seen attempting to overclock anything AMD based.

i prefer people like techyescity usually and hardware unboxed but i now do have some doubts after jims video on the golden samples then the full analyst over steves reviews really made him look like a foolm so much consistencies


I'm pretty hyped for Ryzen 3000 if we can get 4.6ghz then intel is actually in big trouble

as clock for clock ryzen and coffee lake are 5% different if that.
Standard User Chrysalis
(legend) Sat 28-Apr-18 14:04:07
Print Post

Re: ryzen2


[re: epyon] [link to this post]
 
its getting more blatant.

https://www.youtube.com/watch?v=yBYBMQrK7XE

Calling it the ultimate 6 core showdown and proceeds to not use the ultimate 6 core intel chip

8400 vs 2600x a chip that costs £70 more. Even a plain 2600 is priced higher than a 8400.

My comment on GN's video about the test been a highly unusual workload, has apparently left people confused as well about what I meant.

Sky Fibre Pro BQM - IPv4 BQM - IPv6
Standard User Chrysalis
(legend) Fri 11-May-18 03:23:53
Print Post

Re: ryzen2


[re: epyon] [link to this post]
 
HARDOCP been cutoff by nvidia for attacking the GPPP, they have warned they may now go out of business as a result as nvidia partners may also blacklist them.

More proof of how messed up the review industry is.

Someone just needs to setup a site and run on donations, in return they promise the userbase to only review retail parts without review guideline constraints. The costs wouldnt be that high as parts could be sold after testing completed.

Sky Fibre Pro BQM - IPv4 BQM - IPv6

Edited by Chrysalis (Fri 11-May-18 03:24:24)

  Print Thread

Jump to