AMD launches Zen+ 12nm Ryzen and X470 motherboards

Xuper said:
power consumption is based on quality of the silicon .according to Stilt , He thinks various Power consumption's Ryzen 2700X on all review is due to "SIDD (static leakage) mostly"

https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-73#post-39393777

on other hand New node is not completely matured.Hopefully a couple months later , New Ryzen 2700x with better quality silicon launch.
Click to expand...

I found that to be really cool. I need to reread it, but he was also toying around in the bios disabling power protections in that section. I can't recall if he tested power usage with those off or at default. AMD has a pretty long record of using leaky cores for their high end. Like he mentioned, the 2700 non-x is likely binned in the opposite direction to be tighter and not soak itself in amperage. A direct comparison will be interesting to see.

The part about motherboards being the deciding factor of how much power it finally gives the CPU with XFR enabled was also a good read. If the board has high quality power delivery and has those parameters set properly, it should give different results than a lower quality board which wouldn't be able to handle the current delivery over the long haul. This platform is really trying to be very smart about clocking and power delivery.

Maybe to the detriment of traditional old school tweakers, but it makes things pretty awesome to simply drop in and get 95% of the way to maximum performance (save memory tweaks) without any effort. Gone are the days of the 300A and 50% overclocks, you more or less get exactly what you pay for. You only have to decide where you want to sit in the power/performance curve and buy into that range.
 
i feel like everyone is missing the point that it costs more to run intel vs AMD. It just cost more! you gotta spend more on cooling the chips cost more the motherboards are more expensive. the 2nd gen ryzen and x470 you can get away with cheaper ram kits with better compatibility. This is a big problem for intel.
 
New bench , HPET really killed Intel performance , so maybe this explains anandtech's stuff:

http://hardwarebg.com/44332-ryzen-7-2700x-ryzen-5-2600x/6/

another Photoshop CC 2018 CPU Performance Test.:

https://www.pugetsystems.com/labs/a...erformance-AMD-Ryzen-2-vs-Intel-8th-Gen-1136/

I guess Extremely heavy single thread.Photoshop does not use more than 1/2 cores which is shame.

also after effect CC 2018 Test :

https://www.pugetsystems.com/labs/a...Comparison-AMD-Ryzen-2-vs-Intel-8th-Gen-1137/
 
123Lanoix said:
i feel like everyone is missing the point that it costs more to run intel vs AMD. It just cost more! you gotta spend more on cooling the chips cost more the motherboards are more expensive. the 2nd gen ryzen and x470 you can get away with cheaper ram kits with better compatibility. This is a big problem for intel.
Click to expand...
Maybe improved memory compatability will help drag the memory prices down. Keep thinking there is an artificial price bubble there.
 
Xuper said:
New bench , HPET really killed Intel performance , so maybe this explains anandtech's stuff:
Click to expand...

Anandtech problems are on the AMD part, not in the Intel part. The results for Zen+ are wrong. The reported 22% IPC gain in Cinenbench is pure nonsense. The claimed 1 cycle latency reduction in L1 is pure hallucionation,...
 
juanrga said:
Anandtech problems are on the AMD part, not in the Intel part. The results for Zen+ are wrong. The reported 22% IPC gain in Cinenbench is pure nonsense. The claimed 1 cycle latency reduction in L1 is pure hallucionation,...
Click to expand...

Source?
 
Shadowed said:
Source?
Click to expand...

Sources aren't really needed. Just use logic reasoning. If the nonsensical 22% IPC gain that Anandtech reports for CB multithread was correct, then the CB score would be about 30% higher (IPC+ higher-clocks) than the score of the 1800X and Anandtech gives 1801 and 1626 respectively, which is only 10% difference.

Same comments about the caches. Zen L1 latency is 4 cycles, if the latency was reduced by one cycle (i.e. a 25%) as Anandtech claims, then the cache would be about 40% faster (cycle-reduction + higher clocks) and again this is not happening.

Do you still need sources?

AMD claims IPC gain is ~3%. The Stilt measured 2.4% IPC gain and reviewers measured 2.7% and 2.4% gains

15241006322z6uo19egw_4_4.png

aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9FL1AvNzY1MTY5L29yaWdpbmFsL2ltYWdlMDA4LnBuZw==.jpg


AMD claims L1 cache latency reduction of 13% (1.10ns vs 0.95ns) is due to higher clocks. Reviews confirm the same L1 latency when 2700X and 1800X are set at same clocks

getgraphimg.png


1.34 ns for both, because both have the same 4 cycle L1 cache

4 cycle / 3 GHz = 1.33 ns

The Stilt also confirms the L1 cache is unchanged in Zen+ cores aka it has 4 cycles latency.

The Anandtech review is a piece of nonsense. The problem is not only on the nonsensical numbers, and the internal contradictions on the review, but Ian inventing crazy theories such as 3 cycle L1 cache, 11 cycle L2 cache, massive SMT improvements,... to justify his numbers.
 
Last edited:
Just a note, ipc is workload dependent and isn't a simple and consistent figure as it is sometimes thrown around. If a workaround depends heavily on a specific instruction and that instruction was made more cycle efficient then the IPC for that workload will be greatly increased.

That so many reviewers are measuring IPC increase so close to the marketing claim is a strong positive as clearly the gains were made such that many common workloads will benefit.
 
sover said:
Just a note, ipc is workload dependent and isn't a simple and consistent figure as it is sometimes thrown around. If a workaround depends heavily on a specific instruction and that instruction was made more cycle efficient then the IPC for that workload will be greatly increased.

That so many reviewers are measuring IPC increase so close to the marketing claim is a strong positive as clearly the gains were made such that many common workloads will benefit.
Click to expand...

The marketing claim is for CB only.
 
juanrga said:
The marketing claim is for CB only.
Click to expand...

Point still stands, unless they're providing a specific project file for testing. Either way as I said, that reviewers are consistently finding between two and three percent IPC gain is a positive.
 
juanrga said:
Anandtech problems are on the AMD part, not in the Intel part. The results for Zen+ are wrong. The reported 22% IPC gain in Cinenbench is pure nonsense. The claimed 1 cycle latency reduction in L1 is pure hallucionation,...
Click to expand...

No, they aren't. I mean it's possible they are wrong about a conclusion (22% IPC gain - but where did you find that claim? I looked around for it and didn't see it - but it's possible I'm overlooking it). But their benchmark results for AMD are perfectly in line with what Kyle got, and what I'm getting right now. Their Cinebench results were actually a bit lower than Kyle's and my own results.

Point is, if Anandtech had a problem with their benchmark results, it would be on the Intel side. The AMD results are in line with what we're seeing.
 
DuronBurgerMan said:
No, they aren't. I mean it's possible they are wrong about a conclusion (22% IPC gain - but where did you find that claim? I looked around for it and didn't see it - but it's possible I'm overlooking it). But their benchmark results for AMD are perfectly in line with what Kyle got, and what I'm getting right now. Their Cinebench results were actually a bit lower than Kyle's and my own results.

Point is, if Anandtech had a problem with their benchmark results, it would be on the Intel side. The AMD results are in line with what we're seeing.
Click to expand...

He is looking at the lone single synthetic benchmark result with cinebench (CB15nT result) from the graph on this page: https://www.anandtech.com/show/12625/amd-second-generation-ryzen-7-2700x-2700-ryzen-5-2600x-2600/4 , but ignoring the text that follows where anandtech states that it is a 2% to 5% gain in the others synthetic benchmarks. There is a small paragraph that talks about the 22%, in which they state what it appears is the reason,where they also state more testing needs to be done, which means they are unsure of the results of that lone synthetic benchmark. They then go on to state that it appears to have a 10% increase over all. He is just trying to cherry pick one single benchmark to argue about, rather than using all the reverent data. BTW there are people on reddit getting similar results in those synthetic benchmarks as anandtech, and goes along with the AMD results being inline with what we are seeing.
 
Last edited:
Do remember that AMD themselves touted Cinebench, and that Cinebench was and is used as an arch comparison point widely.

It's unfortunate that there seems to be an issue here; it's even more unfortunate that anyone uses Cinebench at all, as IIRC it's pretty outdated and only really addresses base float performance.
 
DuronBurgerMan said:
No, they aren't. I mean it's possible they are wrong about a conclusion (22% IPC gain - but where did you find that claim? I looked around for it and didn't see it - but it's possible I'm overlooking it). But their benchmark results for AMD are perfectly in line with what Kyle got, and what I'm getting right now. Their Cinebench results were actually a bit lower than Kyle's and my own results.

Point is, if Anandtech had a problem with their benchmark results, it would be on the Intel side. The AMD results are in line with what we're seeing.
Click to expand...

this from Anandtech :
For this graph we have rooted the first generation Ryzen 7 1800X as our 100% marker, with the blue columns as the Ryzen 7 2700X. The problem with trying to identify a 3% IPC increase is that 3% could easily fall within the noise of a benchmark run: if the cache is not fully set before the run, it could encounter different performance. Shown above, a good number of tests fall in that +/- 2% range.

However, for compute heavy tasks, there are 3-4% benefits: Corona, LuxMark, CineBench and GeekBench are the ones here. We haven’t included the GeekBench sub-test results in the graph above, but most of those fall into the 2-5% category for gains.

If we take out Cinebench R15 nT result and the Geekbench memory tests, the average of all of the tests comes out to a +3.1% gain for the new Ryzen 2700X. That sounds bang on the money for what AMD stated it would do.

Cycling back to that Cinebench R15 nT result that showed a 22% gain. We also had some other IPC testing done at 3.0 GHz but with 8C/16T (which we couldn’t compare to Bristol Ridge), and a few other tests also showed 20%+ gains. This is probably a sign that AMD might have also adjusted how it manages its simultaneous multi-threading. This requires further testing.
Click to expand...

as you can see , There is nothing problem.They claim 3.1% with Margin Error.

hell even He went on this

https://www.realworldtech.com/forum/?threadid=176004&curpostid=176020
 
Last edited:
NWRMidnight said:
He is looking at the lone single synthetic benchmark result with cinebench (CB15nT result) from the graph on this page: https://www.anandtech.com/show/12625/amd-second-generation-ryzen-7-2700x-2700-ryzen-5-2600x-2600/4 , but ignoring the text that follows where anandtech states that it is a 2% to 5% gain in the others synthetic benchmarks. There is a small paragraph that talks about the 22%, in which they state what it appears is the reason,where they also state more testing needs to be done, which means they are unsure of the results of that lone synthetic benchmark. They then go on to state that it appears to have a 10% increase over all. He is just trying to cherry pick one single benchmark to argue about, rather than using all the reverent data. BTW there are people on reddit getting similar results in those synthetic benchmarks as anandtech, and goes along with the AMD results being inline with what we are seeing.
Click to expand...

Ah I see. Yeah, that's weird - and they say that the result is odd and are going to look into it.

But, of course, that doesn't effect the benchmark results further down, where Anand got a 1801 Cinebench MT score. Kyle got 1827 stock. I got 1802 in first pass stock. I've since posted a few higher than that. Point being, low 1800s is normal for a stock 2700X.

I also figured out what was going with variable Cinebench MT scores. Kyle ran a 4GHz test, and got 1812. Stock he got 1827. When I got 1802, I was watching Ryzen Master. Clocks were at 3975 across all cores. But I've seen all core boosting as high as 4200 - and quite commonly at 4100.

Depending on your cooling and/or power delivery, the all core boost can be quite high. Me at 1802 @3975 is in line with Kyle's 1812 @ 4GHz and 1827 stock (probably a hair over 4 GHz. Maybe 4025 or 4050 or something). Kyle's would be higher because liquid cooling and probably better motherboard power delivery. Precision Boost 2 detects this and expands the margins a bit.
 
sover said:
Point still stands, unless they're providing a specific project file for testing. Either way as I said, that reviewers are consistently finding between two and three percent IPC gain is a positive.
Click to expand...

Some reviewers are finding average IPC between 2% and 3%. HFR got 2.7% average. Others aren't. I think [H] got an average between 1% and 2%. The Stilt review quotes 1.5% gain average.
 
juanrga said:
Some reviewers are finding average IPC between 2% and 3%. HFR got 2.7% average. Others aren't. I think [H] got an average between 1% and 2%. The Stilt review quotes 1.5% gain average.
Click to expand...

On the workload they tested. Did AMD provide the workload that they're using to quote the 3% increase? If not then all you're really talking about is the performance delta on some other workload. The fact of the matter is there's a consistent gain in performance on a wide array of software, and that's what actually matters. Maybe it's time to just leave the windmills alone.
 
DuronBurgerMan NWRMidnight Xuper

(i) I didn't wrote that Anandtech claims 22% IPC average. What I wrote is

juanrga said:
The reported 22% IPC gain in Cinenbench is pure nonsense
Click to expand...

(ii) Mentioning glaring mistakes in a review is not to "cherry pick".

(iii) One of the things one learns in a lab is to check the validity/consistency of data and to reject invalid data. We know that

Performance = IPC * Frequency

If I measure 22% higher IPC in CB15, and next I measure a 10% higher performance in CB15, and I know frequency is higher, then there is a problem somewhere in the data. The problem is in the IPC measurement. 2700X does not have 22% higher IPC than 1800X. The IPC gap is about 2.5% in CB15. So there is a huge measurement error of one order of magnitude.

Instead eliminating that invalid datapoint and investigating how he could do such giant measurement error, Ian added the result to the IPC table used to compute the average IPC. Not only that, but he latter claims obtained similar 20% IPC gains in other benches that he didn't published, and further he tries to justify those values by appealing to improvements in SMT, which adds more fun to the whole history because he got that 22% higher IPC in Cinebench with SMT disabled (at least that is what he claims).

As mentioned this is not the only issue with his review. Not only the measurement methodology is clearly flawed, but the lack of critical review of the findings, and the attempt to justify some of the incorrect measurements puts the whole review on hold, even if we ignore the discrepancies with other reviews.
 
Last edited:
My own testing (I have both a 1700X and a 2700X) leads me to believe a 2 to 3% IPC increase is about accurate. At any even, the highest I've seen quoted is 3.1%, and the lowest around 1.5%. So somewhere in between those.

The thing worth noting is that latency-sensitive tasks benefited the most.
 
sover said:
On the workload they tested. Did AMD provide the workload that they're using to quote the 3% increase? If not then all you're really talking about is the performance delta on some other workload. The fact of the matter is there's a consistent gain in performance on a wide array of software, and that's what actually matters. Maybe it's time to just leave the windmills alone.
Click to expand...

As I wrote in #572, the "~3%" marketing claim is using CB only. It is not an average over different workloads.
 
juanrga said:
DuronBurgerMan NWRMidnight Xuper

...and the attempt to justify some of the incorrect measurements puts the whole review on hold, even if we ignore the discrepancies with other reviews.
Click to expand...

No, it doesn't "put the whole review on hold." It puts that particular data point up to scrutiny. The other results, including his actual Cinebench MT result on page 10 are perfectly in line with what other reviewers got out of the 2700X.
 
Also, when computing his 3.1% average, Ian threw out the 22% result anyway:

"If we take out Cinebench R15 nT result and the Geekbench memory tests, the average of all of the tests comes out to a +3.1% gain for the new Ryzen 2700X. That sounds bang on the money for what AMD stated it would do."

This kind of thing is why good testing methodology often throws out the high and the low when computing the average. It's not showing Ian to be a shitty tester, it's showing that he got one result which is weird/inconsistent with the others, and which he hasn't been able to explain (others may be able to), and he threw it out anyway.

Anandtech has always been very credible, and willing to admit error. They said they were doubling back and reviewing their data again, and that should be good enough for you. If you don't like it, you are welcome to start your own review site.
 
juanrga said:
As I wrote in #572, the "~3%" marketing claim is using CB only. It is not an average over different workloads.
Click to expand...

CB isn't an end unto itself, it doesn't perform exactly the same on every project, because every project isn't the same.

What you're saying is like stating that Intel is x% faster than Ryzen at Grand Theft Auto V, but not specifying any specific benchmark or segment. If you haven't noticed, different workloads for the same engine sometimes produce different results. That you're so fixated on a .5% delta and ignoring that you're comparing oranges to tangerines is concerning.
 
DuronBurgerMan said:
No, it doesn't "put the whole review on hold." It puts that particular data point up to scrutiny. The other results, including his actual Cinebench MT result on page 10 are perfectly in line with what other reviewers got out of the 2700X.
Click to expand...

I already know that the Cinebench MT results on page 10 are correct, I already mentioned the scores of 1801 and 1626, in my post #570, which are correct values that give the expected 10% performance gap in that benchmark. It doesn't invalidate my argument a bit.

One has to know first which is the correct value for a given workload and then go to Ian's review to check if the value he gives is in the expected range or not.
 
DuronBurgerMan said:
Also, when computing his 3.1% average, Ian threw out the 22% result anyway:

"If we take out Cinebench R15 nT result and the Geekbench memory tests, the average of all of the tests comes out to a +3.1% gain for the new Ryzen 2700X. That sounds bang on the money for what AMD stated it would do."
Click to expand...

As mentioned before, the problem is not only that his R15 nT result is glaring wrong, but he even tries to justify that 22%. The other results could be also wrong and no one notice by the eye (his POV-Ray results looks suspicious to me). We have to be checking one for one. Also we have to trust him that the average is that. He doesn't give raw data for we can check. And the graph is prepared more in a cosmetic shape than on a scientific shape.

Same happen in the cache section, not only his measurements are invalid, but he invents microarchitectural improvements to justify his data. There is no such improvements, the Zen+ cores are identical to the Zen cores one can find in EPYC, ThreadRipper, and Raven Ridge. The L1 and L2 latencies are 4 cycles and 12 cycles on all them. His claim that L1 cache in Zen+ is 3 cycles is something Ian invents to justify his invalid measurements.

DuronBurgerMan said:
Anandtech has always been very credible, and willing to admit error. They said they were doubling back and reviewing their data again, and that should be good enough for you.
Click to expand...

The site quality has reduced since founder left. Anandtech has published several flawed reviews since then. We have discussed several of those at RWT, including the recentest reviews of ARM, IBM, Intel, and AMD servers. Below an excerpt of the DK comment on the Anandtech review of EPYC vs Xeon

So I'm pretty sure that something is horribly wrong in Johan's testing.

David
Click to expand...

I will not insist more. I just shared my opinion and proved the mistakes in that Pinnacle Ridge review. That is all.
 
Last edited:
juanrga said:
As mentioned before, the problem is not only that his R15 nT result is glaring wrong, but he even tries to justify that 22%. The other results could be also wrong and no one notice by the eye. We have to go checking one for one. Also we have to trust him that the average is that. He doesn't give raw data for we can check. And the graph is prepared more in a cosmetic shape than on a scientific shape.
Click to expand...

Um, no, you don't have to trust him on that. He tells you which ones he threw out of the average. Run the numbers yourself after throwing those two out. I did, and got *my shocked face* 3.1%. That's evidence that he threw the 22% result out of his calculations.
 
sover said:
That you're so fixated on a .5% delta and ignoring that you're comparing oranges to tangerines is concerning.
Click to expand...

That's very true. The obsession with a very small delta is odd. The IPC difference is somewhere between 1.5% - 3.1%. That should be good enough for anybody.
 
You guys should really just put him on ignore. He will never give AMD kudos for putting out a good product even though he clearly prefers Intel and will always see what he wants to see in the benchmarks.

I think there's a proverb about this: "Do not answer a fool according to his folly, or you yourself will be just like him."
 
kirbyrj said:
You guys should really just put him on ignore. He will never give AMD kudos for putting out a good product even though he clearly prefers Intel and will always see what he wants to see in the benchmarks.

I think there's a proverb about this: "Do not answer a fool according to his folly, or you yourself will be just like him."
Click to expand...

I don't mute him, because he is occasionally insightful. But yes, he has a very clear pro-Intel/anti-AMD bias.

I'm about as neutral on this as I can be, really. I've switched back and forth between AMD and Intel several times as one or the other had a clear lead. Machines I've built: i486 SX-33 -> Cyrix 6x86 "PR133" (lol a shitty chip) -> Pentium II 400 -> Athlon (K7) 650 -> A bunch of overclocked (some fried) Durons (hence my name) -> Athlon Tbird 1400 -> Athlon 64 3000+ -> Athlon X2 4800+ -> Core 2 Quad Q6600 (still running my arcade cabinet) -> Sandy Bridge 2600k (still running my home theater) -> Ryzen 1700X -> Ryzen 2700X
 
kirbyrj said:
He will never give AMD kudos for putting out a good product even though he clearly prefers Intel and will always see what he wants to see in the benchmarks.
Click to expand...

Kudos to AMD for releasing a good product as the RyZen 2000 series.

DuronBurgerMan said:
I don't mute him, because he is occasionally insightful. But yes, he has a very clear pro-Intel/anti-AMD bias.
Click to expand...

I am pro-truth and anti-hype. That most of the hype is coming from the AMD camp is what has some people confused about me. If the Intel camp was predicting/hyping 6GHz on air or 20% higher IPC for next gen, if the Intel camp was relabeling the cores in the i7-8700k Skylake++ core (pretending there is new scheduler, new branch prediction, 1 cycle-faster caches,...) or if the Intel camp was relabeling the 14nm++ node as 11nm, then I would be here, fighting all the false information and the hype. It just happen most of the hype and misinformation is coming about AMD products.

I have critized Anandtech reviews multiples times. I criticized their Intel vs ARM servers review (because it was pro-Intel). I criticized their Intel vs IBM servers review (because it was pro-Intel). I criticized their Intel vs AMD servers review (because it was pro-AMD).

It is fascinating how some few people is able to invent lots of crazy things about me like I am an Intel worker with number of contract 4004, that I am Mexican, that I am typing this sitting at Starbucks and using an iPad...
 
juanrga said:
It is fascinating how some few people is able to invent lots of crazy things about me like I am an Intel worker with number of contract 4004, that I am Mexican, that I am typing this sitting at Starbucks and using an iPad...
Click to expand...


/me chuckles

Bet you wear a turtleneck.
 
juanrga said:
Kudos to AMD for releasing a good product as the RyZen 2000 series.



I am pro-truth and anti-hype. That most of the hype is coming from the AMD camp is what has some people confused about me. If the Intel camp was predicting/hyping 6GHz on air or 20% higher IPC for next gen, if the Intel camp was relabeling the cores in the i7-8700k Skylake++ core (pretending there is new scheduler, new branch prediction, 1 cycle-faster caches,...) or if the Intel camp was relabeling the 14nm++ node as 11nm, then I would be here, fighting all the false information and the hype. It just happen most of the hype and misinformation is coming about AMD products.

I have critized Anandtech reviews multiples times. I criticized their Intel vs ARM servers review (because it was pro-Intel). I criticized their Intel vs IBM servers review (because it was pro-Intel). I criticized their Intel vs AMD servers review (because it was pro-AMD).

It is fascinating how some few people is able to invent lots of crazy things about me like I am an Intel worker with number of contract 4004, that I am Mexican, that I am typing this sitting at Starbucks and using an iPad...
Click to expand...


Looks like Kyle took care of that for you. I guess he thinks your full of shit. He never responds to you so take that for what it's worth.
 
https://www.anandtech.com/show/12678/a-timely-discovery-examining-amd-2nd-gen-ryzen-results

Something about HPET skewing the results. Turning it off brings Anandtech's results in line with what we've been seeing from other reviews, at least it seems so to me. A portion of their conclusion I feel should be highlighted, although you should read the article in its entirety.

First and foremost, we have decided that force-enabling HPET is not how we want to test systems, as this is non-default behavior. While it has an important role in extreme overclocking, to verify accurate timing, ultimately it was akin to taking a sledgehammer to cracking an egg for our testing - we need to be testing systems at stock. So all further CPU testing going forward will be using HPET's default behavior, and we have even put checks in our scripts to ensure this is now the case.

As a result we are retracting our existing results for all of the processors we used in the Ryzen 2000-series review. This goes for both the review and for Bench. All of these products will be updated with revised results using the default HPET behavior just as soon as the updated data is available over the course of the next week. In fact we're already the process of running this updated testing, which we've used for this article and uploaded to Bench.
Click to expand...

Also, I'm glad to see that Anandtech is gonna update the suite of games it uses to 2018 standards (FC5, Shadow of War, F1) , to be frank, using 2013 released games rose my eyebrows way too often for it to be pleasant.
 
SighTurtle said:
https://www.anandtech.com/show/12678/a-timely-discovery-examining-amd-2nd-gen-ryzen-results

Something about HPET skewing the results. Turning it off brings Anandtech's results in line with what we've been seeing from other reviews, at least it seems so to me. A portion of their conclusion I feel should be highlighted, although you should read the article in its entirety.



Also, I'm glad to see that Anandtech is gonna update the suite of games it uses to 2018 standards (FC5, Shadow of War, F1) , to be frank, using 2013 released games rose my eyebrows way too often for it to be pleasant.
Click to expand...


Well that is annoying. They really should know better. That was an issue for some in testing Ryzen 1. But at least they are correcting it. Better to admit the mistake and fix it than excuse it.

And I agree on the games - even though I tend to prefer older games anyway.
 
SighTurtle said:
https://www.anandtech.com/show/12678/a-timely-discovery-examining-amd-2nd-gen-ryzen-results

Something about HPET skewing the results. Turning it off brings Anandtech's results in line with what we've been seeing from other reviews, at least it seems so to me. A portion of their conclusion I feel should be highlighted, although you should read the article in its entirety.
Click to expand...

HPET can explain the huge up to 76% crippling in the gaming performance for Intel chips, but it doesn't explain the anomalous latency/IPC results for RyZen. Moreover, HPET is not to blame for those microarchitectural improvements invented by Ian to justify numbers in the original review.

Xuper said:
hell even He went on this

https://www.realworldtech.com/forum/?threadid=176004&curpostid=176020
Click to expand...

pendragon1 said:
^ screams desperation
Click to expand...

I criticized the review because it was plain wrong. It was so obvious... Kudos to Anandtech for rectifying (at least partially):

When we first published our Ryzen-2000 series review, with HPET forced as the timer in the operating system, our results were broadly showing that the new processors leading the pack. In light of the audit, especially with the way that the Intel gaming results have changed, paint a different picture.
Click to expand...
 
You must log in or register to reply here.
Back
Top