Yes, and that standard 5.0 ghz 8700k will be a bottleneck for 4k un "the future". Better run ouy and get that $800 silicon lottery verzion!
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Ryzen vs Coffee Lake
- Thread starter adobian
- Start date
Yes, and that standard 5.0 ghz 8700k will be a bottleneck for 4k un "the future". Better run ouy and get that $800 silicon lottery verzion!
Dayaks
[H]F Junkie
- Joined
- Feb 22, 2012
- Messages
- 9,773
4k is for noobs. Real men run high Hz screens and VR (high hz).
Don't get me wrong. Ryzen has it's place. I got a Ryzen 1200 for a family member and was thoroughly impressed with the build quality and performance.
If you are running 60Hz Ryzen is probably a decent choice. If Ryzen does get a few hundred Mhz in the March launch the differences between AMD and Intel should be neglible.
Yet, my 2500k says otherwise when compared to a 2600k. Which if I'd bought the i7 back then, I doubt I would have upgraded my intel system to haswell i7. Hell, my biggest gripe these days has been newer video cards causing issues with some older mobos (cough amd/sapphire/msi).
Derfnofred
Gawd
- Joined
- Dec 11, 2009
- Messages
- 606
^ I think you're making a totally different point though. I hope you bought your 2500K because it was the best option for you at the time of purchase, not because you speculated it would last you all these years. I just upgraded from a i7-860 (mid-high end Nehalem), so I totally get it.
It's the "I'm buying this CPU/GPU because in the future the drivers/software will take better advantage of the hardware available." mindset that I'm talking about. Buy what is best for you* today and reevaluate over time as needs and finances permit -- predicting future trends in computing needs and resources is a fool's errand.
* Whatever metric of cost-to-performance is your sweet spot.
It's the "I'm buying this CPU/GPU because in the future the drivers/software will take better advantage of the hardware available." mindset that I'm talking about. Buy what is best for you* today and reevaluate over time as needs and finances permit -- predicting future trends in computing needs and resources is a fool's errand.
* Whatever metric of cost-to-performance is your sweet spot.
cjcox
2[H]4U
- Joined
- Jun 7, 2004
- Messages
- 2,946
But humorous when you consider AMD's failure to produce a promised quad core, which cost them what little market share they had for what... 15+ years? AMD hasn't proven themselves yet... still waiting.
ManofGod
[H]F Junkie
- Joined
- Oct 4, 2007
- Messages
- 12,864
They haven't? Then you will be waiting a long time because your perspective seems to think they never could. Oh well, glad they do not have you in charge.
kirbyrj
Fully [H]
- Joined
- Feb 1, 2005
- Messages
- 30,693
I got the 7820X because the 7820X was $449 at Newegg and I got a X299 board (and case and PSU ) for $170AR. So essentially, I'm out the door with brand new parts for $570 after I sold the case and PSU to my brother. A full price $599+motherboard...probably not worth it. But at $570 with a motherboard? I think it's worth the extra $20 (or virtually the same cost depending on the 1151 board you buy) for the more upgradable platform (think multi core off lease Xeons) and the extra 2 cores now. Overclocked to 4.5-4.7Ghz on 8 cores...I doubt the 8700k would be a noticeable "upgrade." Even the 7800X is often at $300 or less (guy had one BNIB in forums for $259). So I wouldn't agree with your "much cheaper" argument if you shop around.
In fair disclosure, I'm a bang for the buck whore, and I'm definitely not paying over MSRP for a mainstream Intel part when the new version 8 core version is looking like Q2'18. The bang for the buck is why I think Ryzen is a good buy for the average user.
Last edited:
I went with the 2500k because why spend the extra on an i7 that wasn't really worth it. Cost wasn't really a big deal, bang for the buck was. Yet, the trends of the past 6-8 years show that that i7 is the better buy long term. May not hold true now but from what I've see in reviews, I think it holds.
Derfnofred
Gawd
- Joined
- Dec 11, 2009
- Messages
- 606
Ranulfo -- I think we're essentially saying the same thing in different words.
* Hyperthreading doesn't really help games, so by and large the 2500k and 2600k are equivalent. You're still good!
* Hyperthreading doesn't really help games, so by and large the 2500k and 2600k are equivalent. You're still good!
In my country it's $700 vs $1100 with no deals available, so the choice was easy.
OrangeKhrush
[H]ard|Gawd
- Joined
- Dec 15, 2016
- Messages
- 1,673
There is a 9% difference in the IPC gap reported by Gur3D and the gap reported by rest of reviews.
1) As mentioned in my post, the problem is that Guru3D obtains higher scores for RyZen chips and lower scores for Intel chips. And they are getting the same anomalous results since RyZen launch. Aren't random variances due to BIOS/AGESA or memory setups. If you read my post, you can check I also mentioned some of the dirty tricks that Guru3D uses to obtain those anomalous results.
2) The problem with your pretension that the 8700k goes above the max turbo and RyZen goes below the base clock, is that it vanishes once all chip are tested at same clocks, and the 15--20% IPC gap remains.
8370 at 4K.
I think your mathematics is severely problematic.
PcPer (feb/march) - 139
Guru3D (August) - 146
% difference = 4.59%
Far cry form 9%, I also find this rich from a guy that once posted that Anandtech 3Ghz showdown bench and while Ryzen got around 119, the Sandy was around 106 and you claimed that 12% + is inside margin of error.
A 4.5% deviation is hardly anything but margin of error and as I stated that is down to various smaller factors like motherboard stability, RAM stability. Stop kicking up a mountain from a mole hill.
If i recall you are also a big proponent of blaming some intel underwhelming results on "new socket" teething, as far as I am concerned what it swings both ways Mr straight down the middle. (kappa)
Ryzen's limitations come from clockspeed off sets, AMD can't aggressively set throttling and boost policies so in gaming there is normally a massive off set in clock speed and some games are benefactors of clockspeed. In games that are not sensitive to clockspeed Ryzen is perfectly keeping honest with Intels latest and greatest.
Hard to track your constant IPC changing differentials, one moment it is 10% now it is 20%, though going on your math, there is a great chance pigs will start to fly. IPC offset clock for clock to Coffee lake is probably in that 10-12% which is certainly a lot closer than the days it was like 60-65% off, that is progress.
OrangeKhrush
[H]ard|Gawd
- Joined
- Dec 15, 2016
- Messages
- 1,673
At this point this is a straight out cherry pick, we now have PC Gamer a binding source, we have some french slide that seems to be in a list of saved to desktop files that repeat depending on the circumstances, and newer benches are dismissed.......Oh and I forgot we now have simple mathematics going out the window to try push agenda
Last edited:
OrangeKhrush
[H]ard|Gawd
- Joined
- Dec 15, 2016
- Messages
- 1,673
I never claimed that did I.
And you can get 8700K for 370$. But it did give big price cuts on Ryzen as I said
Not to mention how far away your IPC claim ended.
1) if you can find a reseller selling for 370 great, from my local list to conversion rate and VAT plus vendor mark up it seems bang on 450-500 right now. No matter how intel want to price it the ultimate price is resellers prerogative, this is business.
2) my cinebench was a lot closer than "he who shall not be named", even had it vetted before the NDA lift as being inside margin of error. 10%~ seems to have been right on the mark.
Juangra IDF attack plan for all Ryzen threads:
1.) Show French graph where day 1 Ryzen has similiar IPC as Sandy Bridge on obscure benchmark (with no system spec listed).
2.) Show completely different benchmark that shows Agesa 1006 is no better than 1004 or others (again, no system specs)
3.) Bash any review that disagrees with above two steps
4.) Ignore Math
1.) Show French graph where day 1 Ryzen has similiar IPC as Sandy Bridge on obscure benchmark (with no system spec listed).
2.) Show completely different benchmark that shows Agesa 1006 is no better than 1004 or others (again, no system specs)
3.) Bash any review that disagrees with above two steps
4.) Ignore Math
kirbyrj
Fully [H]
- Joined
- Feb 1, 2005
- Messages
- 30,693
5). Ignore the cost differences
The example you mention is nice. But both chips have similar single thread performance and the i7 advantage is coming from HT. Single-core performance is not sacrificed for more threads, but the i7 core can use iddle resources to run another thread. The argument of this other people is different. They are pretending that future games will magically hide the huge single-core performance in Zen cores.
It is fair to expect better threaded games in a future. No one is negating that. What is being negated is the fantasy that magic improvements in future games will do that chips like the 1800X, with a 30--40% core performance deficit, will start getting the top scores with faster GPUs.
I have been reading the same fantasy for many years. First with Bulldozer, then with Piledriver, then with PS4/Xbox1, then with Mantle, then with DX12 and Vulkan, now with RyZen...
Historical record of "CPU tests" made by Computerbase.de show that i5 2500k continues being a better gaming CPU than FX-8350. Improvements in engines and APIs have reduced the FX-i5 gap from something as 17% in 2012 to something as 11% today. The FX was slower then and continues being slower today.
So I could expect something like that for RyZen and Kabylake. With better threading 16T RyZen could reduce the performance gap with 8T Kabylake in some few percents in next five years. But there is a huge difference now. One difference you correctly mention... CoffeLake
In the ancient FX-i5 debate, the 8T Piledriver was faster than the 4T SandyBridge in multithreading. But today the 12T CoffeeLake is already so fast (or even faster) than the 16T Zen in multithreaded benches. So better multithreading will not close the performance gap between RyZen and CoffeeLake.
If you are going to comment at least read my posts, because my remark about the 9% was about something different. It wasn't about the PCPer-Guru3d difference.
Also if by mentioning the dates of when the PcPer and Guru3D reviews were performed, you are pretending that the IPC of Zen has increased with BIOS/AGESA updates, I can say you are riding a dead horse. Just compare apples with apples. Compare Guru3D with Guru3D.
Comparing Guru3D (August) data with Guru3D (March) data, the 1800X was 3.2% faster on ST and 1.2% faster in MT. The average is a ridiculous 2.2% faster.
But here comes the interesting part. Guru3D doesn't give any details about the hardware specs used in the August review. We don't know if the August review used the same settings than the March review or if the August review did run overclocked RAM. <<---- hint
Luckily we have Guru3D review of the effect of different AGESA versions on the 1700. Gur3d only gives MT scores in this case:
R7 1700 - Guru3D (April): 1438
R7 1700 - Guru3D (June): 1433
The R7 1700 running the latest AGESA version was slower than with older version of AGESA.
The only way that the 1700 can be now faster than in the initial review is when using overclocked RAM. The R7 1700 with 3200MHZ is 1.3% faster than the 1700 with 2933MHz according to Guru3D. A whole 1.3% faster. Wow!
It is pretty obvious now that the 1800X is 2.2% faster in the last Guru3D review only because the guy changed the hardware configuration. He used slower memory on the launch review (March) and he overclocked the RAM to 3200MHz in the August review. He forgot to mention those details in the review and in the graphs, just as he forgets to mention he is testing engineering samples of Intel chips.
So when comparing scores of different Guru3D reviews and seeing 1% or 2% higher scores than at launch time. It is because they guy is changing the hardware configuration from review to review. He also forget to mention in the graphs that he is not testing stock RyZen chips., but overclocked chips.
It is very funny how reporting facts is considered an attack.
1) I have given HFR reviews of 1300X. And I have given HFR reviews of CoffeLake. I have also provided Ars and PcPer reviews of CoffeLake. No one of those is "day 1 RyZen". At contrary, those reviews have all BIOS/AGESA updates installed.
2) AGESA 1.0.0.6 only provides better overclocking/compatibility support with faster RAM (up to 4000MHz). It doesn't change the IPC of Zen core neither the stock settings performance. Even using the new AGESA 1.0.0.6 to push memory to higher overclocks, the improvements in performance are minimal... 0% in Deux Ex, 5% in Hitman, 3% in TombRaider.... It is very funny some people still continue believing on some magic AGESA update that will solve Zen latency deficits and will close a ~30% gap.
3) Did you heard of the scientific method? Reporting the flaws and technical mistakes in reviews is a must for anyone with a critical thinking. Mentioning how the wrong settings in Hardware Unboxed reviews affect results or how Guru3D is testing engineering samples of Intel chips is part of a scientific approach.
4) The pot calling the kettle black
5) Cost differences are more relevant for people in a budget or for people that doesn't need the performance and get the job made with a cheaper chip. At time of writing this, there is only three RyZen chips among the top ten bestselling list in Amazon. The rest are Intel chips. This proves that cost is not a so relevant metric.
Also many of you only consider the acquisition cost, and ignore any other cost when you introduce "cost" in the discussion.
Last edited:
Oh definetly, expecting that big performance jumps is just silly. There is some untapped potential that can be unlocked with more threading but that cannot create miracles for current Ryzens. They are good CPU's but ultimately they are held back by poor max clockspeed. Even my IRL friend who owns Ryzen 7 1700 and is happy with it admits this. Hell anybody who is denying this is in, well, denial. However the performance they bring with only 4ghz max is good, even you should see that. It really is not that far behind from Intel equivalents that are clocked whopping 1ghz higher. That must mean the design of the CPU itself is sound, just in need of little improving. Ryzen is very unlike Bulldozer and its offspring, their performance absolutely sucked back then no matter what clockspeed! Clock for clock it was slower than Phenoms and even more multithreaded games did not save them because in reality they did not have that high core count. Or more like, in theory the design had a lot of cores but in reality it worked like it had half of them or something like that, I don't remember the details. I think this caused some internet outrage too? I stopped following Dozer and Driver news ages ago, everybody saw what kind of turds they were right from the start. Ryzen however makes me hopeful for AMD's future and I am following where this leads with great interest. But if AMD is unable to break the 4ghz barrier in 2018 and instead goes back to "moar cores!" mode I crawl back into my hole feeling depressed.
I expect 5--10% higher clocks for Zen next year with Pinnacle Ridge.
We'll see. At this point it's all a guessing game.
Dayaks
[H]F Junkie
- Joined
- Feb 22, 2012
- Messages
- 9,773
IMO processors are the only thing you can "future proof". Yes they depreciate, but my 5960x still handily beats AMD's best OC to OC and it came out in 2014. I hope to keep it for another few years... no reason not to.
Video cards have much higher gains and are super easy to change out. It makes less sense there. Processors are pretty stagnant.
Video cards have much higher gains and are super easy to change out. It makes less sense there. Processors are pretty stagnant.
cjcox
2[H]4U
- Joined
- Jun 7, 2004
- Messages
- 2,946
Wow. That was uncalled for.
My point is that AMD was doing well and they totally blew it. They don't have a good track record.
OrangeKhrush
[H]ard|Gawd
- Joined
- Dec 15, 2016
- Messages
- 1,673
If you are going to comment at least read my posts, because my remark about the 9% was about something different. It wasn't about the PCPer-Guru3d difference.
Also if by mentioning the dates of when the PcPer and Guru3D reviews were performed, you are pretending that the IPC of Zen has increased with BIOS/AGESA updates, I can say you are riding a dead horse. Just compare apples with apples. Compare Guru3D with Guru3D.
Comparing Guru3D (August) data with Guru3D (March) data, the 1800X was 3.2% faster on ST and 1.2% faster in MT. The average is a ridiculous 2.2% faster.
But here comes the interesting part. Guru3D doesn't give any details about the hardware specs used in the August review. We don't know if the August review used the same settings than the March review or if the August review did run overclocked RAM. <<---- hint
Luckily we have Guru3D review of the effect of different AGESA versions on the 1700. Gur3d only gives MT scores in this case:
R7 1700 - Guru3D (April): 1438
R7 1700 - Guru3D (June): 1433
The R7 1700 running the latest AGESA version was slower than with older version of AGESA.
The only way that the 1700 can be now faster than in the initial review is when using overclocked RAM. The R7 1700 with 3200MHZ is 1.3% faster than the 1700 with 2933MHz according to Guru3D. A whole 1.3% faster. Wow!
It is pretty obvious now that the 1800X is 2.2% faster in the last Guru3D review only because the guy changed the hardware configuration. He used slower memory on the launch review (March) and he overclocked the RAM to 3200MHz in the August review. He forgot to mention those details in the review and in the graphs, just as he forgets to mention he is testing engineering samples of Intel chips.
So when comparing scores of different Guru3D reviews and seeing 1% or 2% higher scores than at launch time. It is because they guy is changing the hardware configuration from review to review. He also forget to mention in the graphs that he is not testing stock RyZen chips., but overclocked chips.
It is very funny how reporting facts is considered an attack.
1) I have given HFR reviews of 1300X. And I have given HFR reviews of CoffeLake. I have also provided Ars and PcPer reviews of CoffeLake. No one of those is "day 1 RyZen". At contrary, those reviews have all BIOS/AGESA updates installed.
2) AGESA 1.0.0.6 only provides better overclocking/compatibility support with faster RAM (up to 4000MHz). It doesn't change the IPC of Zen core neither the stock settings performance. Even using the new AGESA 1.0.0.6 to push memory to higher overclocks, the improvements in performance are minimal... 0% in Deux Ex, 5% in Hitman, 3% in TombRaider.... It is very funny some people still continue believing on some magic AGESA update that will solve Zen latency deficits and will close a ~30% gap.
3) Did you heard of the scientific method? Reporting the flaws and technical mistakes in reviews is a must for anyone with a critical thinking. Mentioning how the wrong settings in Hardware Unboxed reviews affect results or how Guru3D is testing engineering samples of Intel chips is part of a scientific approach.
4) The pot calling the kettle black
5) Cost differences are more relevant for people in a budget or for people that doesn't need the performance and get the job made with a cheaper chip. At time of writing this, there is only three RyZen chips among the top ten bestselling list in Amazon. The rest are Intel chips. This proves that cost is not a so relevant metric.
Also many of you only consider the acquisition cost, and ignore any other cost when you introduce "cost" in the discussion.
As before differences come from BIOS and setup, no guarantees were made as to hardware used and as for BIOS updates some hurt others improved performance.
Actually the Guru3D chart referred to is the 3.5Ghz chart, ie: that is not an overclock so the point is irrelevant and i'll dismiss as ramblings again.
I like how you accuse Guru, they were not on the "friends of Intel" list and got shipped a board with issues for their 8700K, while "friends with benefits" got the auto overclock ASUS boards, the disparity in performance was very substantial, i didn't see you complaining then about bad reviewing.
I am glad 9% dropped to 1-2% which is basically margin of error accountable in benchmarking. If one assumes a rough 140 @ 3.5ghz a inside margin score to excpect on CB 15 is 160+ which is factually proven by any Ryzen user at 4ghz and is not even worth debating anymore.
Just because you keep ending with that "facts [brain fart]" rubbish doesn't make you right.
Last edited:
OrangeKhrush
[H]ard|Gawd
- Joined
- Dec 15, 2016
- Messages
- 1,673
Well given that Samsungs LP node in present process is not capable of stable overclocking beyond that 4ghz or so mark I guess you can get 4.5ghz out of a 5960X though I guess if you can't then beat an AMD processor then something is truly sad. On a clock / clock basis its pretty much the same, just that the AMD's platform is mainstream while yours is a extremely prices X platform.
OrangeKhrush
[H]ard|Gawd
- Joined
- Dec 15, 2016
- Messages
- 1,673
A little while ago he "expected" 15%, now we are down to 5-10%, safe to say it is just guess work and not founded upon anything more that fanboism gummi juice.
Even Guru3D shows that new AGESA provide zero performance improvements. In fact they found that RyZen is slower with the new AGESA 1.0.0.6.
R7 1700 - Guru3D (April): 1438
R7 1700 - Guru3D (June): 1433
So, not only those magic "BIOS updates" you have been promoting since launch don't exist, but the updates go in the opposite direction. RyZen is slower now than at launch.
The overclock comes from overclocking the RAM, which automatically overclocks the interconnect on RyZen chips. So even at fixed 3.5GHz, the RyZen chips are overclocked because the interconnect is not running at stock speed. I am not against testing overclocked chips, but about not noticing this fact in the graphs. For fairness and rigor, Guru3D would add a "[OC]" to the labels of the RyZen chips when are not running at stock speeds.
Open your eyes then. I have accused Guru3D both when reviewed Intel chips and when reviewed AMD chips. Their reviews are bad and useless in both cases. That is why I avoiud that unprofessional site.
The only reason why the 1800X is a 2% faster now that in the launch review, according to Guru3D, is because he changed the hardware specs. He used a lower memory in the launch review "DDR4 2133 ~ 3000 MHz (Corsair)" and he changed the memory to 3200MHz in the last review that you mention "G-Skill 3200 14-14-14-34 (F4-3200C14D-16GTZR G.Skill RGB)", which means he is now testing an overclocked RyZen, because overclocking memory to 3200MHz automatically overclocks the interconnect. So the 2% gain comes from overclocking the chip, which reduces the CCX-CCX-memory latency.
This is the second time that you misunderstand what the 9% means. My advice is the same than the first time: "If you are going to comment at least read my posts, because my remark about the 9% was about something different."
Give the link to myself "expecting 15%" higher clocks for Pinnacle Ridge.
Should Intel chips get the same OC label with >2666 memory speeds?
OrangeKhrush
[H]ard|Gawd
- Joined
- Dec 15, 2016
- Messages
- 1,673
Even Guru3D shows that new AGESA provide zero performance improvements. In fact they found that RyZen is slower with the new AGESA 1.0.0.6.
R7 1700 - Guru3D (April): 1438
R7 1700 - Guru3D (June): 1433
So, not only those magic "BIOS updates" you have been promoting since launch don't exist, but the updates go in the opposite direction. RyZen is slower now than at launch.
The overclock comes from overclocking the RAM, which automatically overclocks the interconnect on RyZen chips. So even at fixed 3.5GHz, the RyZen chips are overclocked because the interconnect is not running at stock speed. I am not against testing overclocked chips, but about not noticing this fact in the graphs. For fairness and rigor, Guru3D would add a "[OC]" to the labels of the RyZen chips when are not running at stock speeds.
Open your eyes then. I have accused Guru3D both when reviewed Intel chips and when reviewed AMD chips. Their reviews are bad and useless in both cases. That is why I avoiud that unprofessional site.
The only reason why the 1800X is a 2% faster now that in the launch review, according to Guru3D, is because he changed the hardware specs. He used a lower memory in the launch review "DDR4 2133 ~ 3000 MHz (Corsair)" and he changed the memory to 3200MHz in the last review that you mention "G-Skill 3200 14-14-14-34 (F4-3200C14D-16GTZR G.Skill RGB)", which means he is now testing an overclocked RyZen, because overclocking memory to 3200MHz automatically overclocks the interconnect. So the 2% gain comes from overclocking the chip, which reduces the CCX-CCX-memory latency.
This is the second time that you misunderstand what the 9% means. My advice is the same than the first time: "If you are going to comment at least read my posts, because my remark about the 9% was about something different."
At launch with bios updates there were barely any boards capable of running 3200mhz, now after each vendor has pushed stability updates all boards are capable of 3200mhz. I also see you are not objecting to Intel using 3200mhz and "OCing" the IMC, probably not objecting ot ASUS's secret built in overclock either. I don't really see that much in it, it all comes down to perspective, you are pro PCPer who are pro Intel and hate anyone that in anyway objectively views AMD. If you are trying to preach to the choir, I already know that, you don't have to convince me of your hate of AMD.
OrangeKhrush
[H]ard|Gawd
- Joined
- Dec 15, 2016
- Messages
- 1,673
Beat me to it, but the answer is no, Intel can do what they want.
Thanks for repeating to me what I already know and wrote both here and in other threads.
This is not an AMD vs Intel question. It is about being accurate and fair or not.
When you overclock the memory on Intel chip, the chip continues running at stock settings. The interconnect in the chip runs at stock settings, only the memory is overclocked.
When you overclock the memory on Bulldozer/Piledriver/Steamroller/Excavator chips, the chips continue running at stock settings. The interconnect in the chips runs at stock settings, only the memory is overclocked.
However, when you overclock the memory on RyZen /Threadripper chips, the chips no longer run at stock settings. The interconnect in the chips runs overclocked, because the Zen microarchitecture ties the memory speed to the the Infinity Fabric speed. Overclocking memory automatically overclocks the chips.
So when testing Piledriver or Broadwell or Excavator or Kabylake chips with memory overclocked, the chips are on stock settings and would be labelled as stock chips on graphs. But when testing RyZen or ThreadRipper chips with memory overclocked, the chips aren't on stock settings and would be labelled as overclocked chips on graphs.
It is so unfair to run overclocked RyZen and Threadripper chips and label them as stock chips in graphs, as would be to run Skylake-X with the interconnect overclocked and label it as a stock chip. However, whereas we find dozens of examples for the first case, I don't know a single example of the second case. All reviews that overclock the interconnect on Intel chips, mention this fact on graphs. Check this review
In this above case, it is irrelevant to mention if the RyZen chip has the interconnect overclocked or not, because the memory used is mentioned in the graph. So a single look at the graph and we know what is overclocked and what is stock in each case. But in many other reviews, the memory used is not mentioned in the graph. And one has to visit the hardware settings section of the review to check if the Zen chips are being overclocked or not.
Particularly bad reviews as Guru3D even hide this fundamental information to readers, and one has to do computations to infer what memory is being used for the 1800X in the August review so loved by OrangeKrush and others. One has to do computations to infer that the 1800X is running with both memory and interconnects overclocked in the August review, because Guru3D forgets to mention this information either in graphs or in text.
In fact, Guru3D is so unprofessional that tests engineering samples for Intel chips but forget to add a "[ES]" label to graphs.
So when looking at graphs as the next, one initially believes Guru3D is testing retail vs retail
but the reality is that both RyZen and ThreadRipper chips would have an "[OC]" label indicating they aren't running at stock clocks (the interconnect is overclocked by 20%). And the i9-7900X would have an "[ES]" label indicating Guru3D is using an engineering sample instead a retail chip.
Last edited:
It's not that you don't have a legitimate point, but I think you're being overly pedantic on this one as it has been an unwritten rule for years that all CPUs run with higher than specified memory speeds (and everything related to those speeds).
kirbyrj
Fully [H]
- Joined
- Feb 1, 2005
- Messages
- 30,693
Yes, but certain members of the IDF have used this as an explanation of why the z370 is necessary. I mean how could it not be? Z270 only supported 2400mhz
.
OrangeKhrush
[H]ard|Gawd
- Joined
- Dec 15, 2016
- Messages
- 1,673
This is not an AMD vs Intel question. It is about being accurate and fair or not.
When you overclock the memory on Intel chip, the chip continues running at stock settings. The interconnect in the chip runs at stock settings, only the memory is overclocked.
When you overclock the memory on Bulldozer/Piledriver/Steamroller/Excavator chips, the chips continue running at stock settings. The interconnect in the chips runs at stock settings, only the memory is overclocked.
However, when you overclock the memory on RyZen /Threadripper chips, the chips no longer run at stock settings. The interconnect in the chips runs overclocked, because the Zen microarchitecture ties the memory speed to the the Infinity Fabric speed. Overclocking memory automatically overclocks the chips.
So when testing Piledriver or Broadwell or Excavator or Kabylake chips with memory overclocked, the chips are on stock settings and would be labelled as stock chips on graphs. But when testing RyZen or ThreadRipper chips with memory overclocked, the chips aren't on stock settings and would be labelled as overclocked chips on graphs.
It is so unfair to run overclocked RyZen and Threadripper chips and label them as stock chips in graphs, as would be to run Skylake-X with the interconnect overclocked and label it as a stock chip. However, whereas we find dozens of examples for the first case, I don't know a single example of the second case. All reviews that overclock the interconnect on Intel chips, mention this fact on graphs. Check this review
In this above case, it is irrelevant to mention if the RyZen chip has the interconnect overclocked or not, because the memory used is mentioned in the graph. So a single look at the graph and we know what is overclocked and what is stock in each case. But in many other reviews, the memory used is not mentioned in the graph. And one has to visit the hardware settings section of the review to check if the Zen chips are being overclocked or not.
Particularly bad reviews as Guru3D even hide this fundamental information to readers, and one has to do computations to infer what memory is being used for the 1800X in the August review so loved by OrangeKrush and others. One has to do computations to infer that the 1800X is running with both memory and interconnects overclocked in the August review, because Guru3D forgets to mention this information either in graphs or in text.
In fact, Guru3D is so unprofessional that tests engineering samples for Intel chips but forget to add a "[ES]" label to graphs.
So when looking at graphs as the next, one initially believes Guru3D is testing retail vs retail
but the reality is that both RyZen and ThreadRipper chips would have an "[OC]" label indicating they aren't running at stock clocks (the interconnect is overclocked by 20%). And the i9-7900X would have an "[ES]" label indicating Guru3D is using an engineering sample instead a retail chip.
Corona is not really a memory intensive benchmark, but I will point out that it is Dual Channel vs Quad Channel and like the Systemsoft benches where I said a very long time that Ryzen was held back by dual channel bandwidth limitations, it was vindicated by Thread ripper and quad channel adn again per AIDA64 AMD showing higher theoretical bandwith. This is againn shown by TR having a higher score.
You can call guru bad, yet you failed to point out Techreport pulling a quad vs dual channel bandwidth offset. If you choose to live by cherrypicks nobody is going to take you seriously.
4k is for noobs. Real men run high Hz screens and VR (high hz).
Don't get me wrong. Ryzen has it's place. I got a Ryzen 1200 for a family member and was thoroughly impressed with the build quality and performance.
If you are running 60Hz Ryzen is probably a decent choice. If Ryzen does get a few hundred Mhz in the March launch the differences between AMD and Intel should be neglible.
haha talking the man right here (tangoseal) that has a 240hz for the win!
I've owned a 1700x then a Threadripper and both of those can't even come close to the FPS gaming power than Coffee Lake and Skylake X can deliver.
And here you do again. You ignore my point and then answer about something else...
I am not discussing throughput neither memory bandwidth. I am discussing latency and interconnect speeds: Infinity fabric, mesh, and ring.
If you check the corona bench. The RyZen chip with the interconnect overclocked is 7% faster than the same RyZen chip with the interconnect working at stock clocks. Overclocking the memory on the SKL-X chip only improves performance by 3%. The difference comes because the SKL-X chip has the interconnect running at stock clocks in both cases.
If one is going to compare chips, one would compare stock vs stock or overclocked vs overclocked. However, that useless/biased Guru3D site that you promote often is testing overclocked RyZen/ThreadRipper retail chips vs engineering samples at stock clocks.
Using those weird tricks and others, Guru3D provides false IPC gaps between Intel and Zen. Since the IPC gap is false, you cannot use it to explain why 6C CFL beats 8C Zen on multithreaded workloads, and then you have to report to fantasias about how the Intel chip is running above the max turbo frequency, whereas the RyZen chip is running under the base clock...
__hollywood|meow
[H]ard|Gawd
- Joined
- Feb 20, 2006
- Messages
- 1,500
i just want to jump in from my usual lurking to say that i came here to read an interesting technical debate about zen vs coffee lake & instead was greeted with stark raving lunacy presented as even-handed logic by someone who apparently has the biggest chip on his shoulder to ever exist against one of the most reputable & objective hardware sites in existence. juan, if you have to bend over backwards making incoherent clusterfuck posts for days just to get your point across, youre either not expressing yourself well or twisting information to suit a predetermined conclusion.
okay. back to lurking now. but not here, as theres no real technical discussion. see you guys
okay. back to lurking now. but not here, as theres no real technical discussion. see you guys
OrangeKhrush
[H]ard|Gawd
- Joined
- Dec 15, 2016
- Messages
- 1,673
And here you do again. You ignore my point and then answer about something else...
I am not discussing throughput neither memory bandwidth. I am discussing latency and interconnect speeds: Infinity fabric, mesh, and ring.
If you check the corona bench. The RyZen chip with the interconnect overclocked is 7% faster than the same RyZen chip with the interconnect working at stock clocks. Overclocking the memory on the SKL-X chip only improves performance by 3%. The difference comes because the SKL-X chip has the interconnect running at stock clocks in both cases.
If one is going to compare chips, one would compare stock vs stock or overclocked vs overclocked. However, that useless/biased Guru3D site that you promote often is testing overclocked RyZen/ThreadRipper retail chips vs engineering samples at stock clocks.
Using those weird tricks and others, Guru3D provides false IPC gaps between Intel and Zen. Since the IPC gap is false, you cannot use it to explain why 6C CFL beats 8C Zen on multithreaded workloads, and then you have to report to fantasias about how the Intel chip is running above the max turbo frequency, whereas the RyZen chip is running under the base clock...
Eitherway the result difference is negligable, memory doesn't affect performance all that much, and is only noticeable in memory synthetics.
It is trite that a buyer of a Ryzen setup will almost certainly opt for best performance possible and in this case given that Ryzen does benefit more from added bandwidth I don't think anyone other than a casual builder will put anything less than 3000mhz in, for intel it is inconsequential, I have never bought anything faster than the IMC rated.
I have also found that with Ryzen you can actually get better performance from timings, I can make a 2600mhz kit beat a 3000mhz kit. It is something very few overlook, it is not about mhz with memory.
Yes a 6 core CFL beats a 8 core 290 dollar part running stock clocks, and the CFL using ASUS auto OC running 4.7ghz compared to the know stock ryzen 1700 all core around 3.4ghz. 2 cores cannot compensate for 1.4ghz added clock speed.
They used a gigabyte board so the 8700K was running its standard 4.4ghz all core, and it only managed a measly 20pts to a stock 3.4ghz 1700, once overclocked the Ryzen 1700 just flies away.
Another clockspeed vs core count and Ryzen again shines offering copious performance at lower clocks.
With all your fanboying, I would be more concerned that AMD don't stumble upon a high clocking node, if they do, it will be very bad. Going from Bulldozer to Zen the gap has closed a lot.
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
Do you even read the crap you post?
"At stock, the Intel system with fewer cores is faster, but the AMD system gets faster when you overclock it!"
What about when you overclock the Intel system? Or are you not going to compare apples to apples? Is your need to 'counter-troll' so demanding that you cannot present an honest argument?
Further, with respect to memory, it's not about bandwidth with Ryzen- it's all about latency. By increasing memory clockspeed, you overclock the CPU's infinity fabric, which results in lower inter-die latency, and by using lower latency memory, you decrease memory latency on top of that, therefore lowering system latency.
Your best bet for RAM on Ryzen is the DDR4-3200 C14 stuff, which is stupid expensive, when on Intel it mostly doesn't matter, and it'll still be faster than Ryzen. If anything, AMD needs to get a massaged stepping out that isn't so dependent on nice RAM.