AMD launches Zen+ 12nm Ryzen and X470 motherboards

juanrga said:
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.
Click to expand...

Told you the problem was mainly on the Intel side.

As for the anomalous 22% result, it was thrown out of their calculations anyway (appropriately), so I don't really care about that. Remainder of their Ryzen 2700X results were correct. It was their 8700k results that were screwed up by the HPET problem. Now Anand's gaming results are broadly inline with other reviewers, showing somewhere between a 5% and 10% gaming gap for 2700X vs 8700k.

https://www.anandtech.com/show/12678/a-timely-discovery-examining-amd-2nd-gen-ryzen-results/5
 
DuronBurgerMan said:
Told you the problem was mainly on the Intel side.

As for the anomalous 22% result, it was thrown out of their calculations anyway (appropriately), so I don't really care about that. Remainder of their Ryzen 2700X results were correct. It was their 8700k results that were screwed up by the HPET problem. Now Anand's gaming results are broadly inline with other reviewers, showing somewhere between a 5% and 10% gaming gap for 2700X vs 8700k.

https://www.anandtech.com/show/12678/a-timely-discovery-examining-amd-2nd-gen-ryzen-results/5
Click to expand...

LOOL , Now You Read my Post 564 , I said it before first Anandtech review might be due to HPET.

Poor juanrga , at least wait then attack !

Anandtech problems are on the AMD part, not in the Intel part
Click to expand...
 
Great follow up article on AT's part, goes to show there are often many different definitions of a valid measurement. I'm really curious if they will publish more in depth findings of the impact of HPET with and without the spectre patches. If Windows ever begins to rely on more precise timings and starts defaulting HPET to on if it's available this would be an enormous change.

It doesn't quite sit well with me how the "extreme" overclocking results will have one set of numbers with HPET forced on, where you can't really scale those numbers down to more sane clocks in an average review. I guess those numbers only ever lived to get high scores anyway but it's one more caveat.
 
DuronBurgerMan Xuper Yes, the HPET issue crippled the gaming performance of Intel chips and invalidated the gaming results. That is correct in your side. But I didn't focus on criticizing the wrong gaming results, because the review was already wrong in the first part: the part about cache latencies, IPC and application performance. My comments are about that first part. That first part is wrong because the flaw is on the AMD side. Cache latencies for the AMD are incorrect, several IPC values for AMD are incorrect, and overall non-gaming performance is higher than in other reviews.

I understand your point about HPET and gaming. Do you understand mine? You can eliminate all the Intel chips from the review and all those flaws and incorrect vañues I mentioned remain.
 
Last edited:
juanrga said:
DuronBurgerMan Xuper Yes, the HPET issue crippled the gaming performance of Intel chips and invalidated the gaming results. That is correct in your side. But I didn't focus on criticizing the wrong gaming results, because the review was already wrong in the first part: the part about cache latencies, IPC and application performance. My comments are about that first part. That first part is wrong because the flaw is on the AMD side. Cache latencies for the AMD are incorrect, several IPC values for AMD are incorrect, and overall non-gaming performance is higher than in other reviews.

I understand your point about HPET and gaming. Do you understand mine? You can eliminate all the Intel chips from the review and all those flaws and incorrect vañues I mentioned remain.
Click to expand...


Reviewer activates a feature that makes AMD slower: STUPID AMD! AMDSUXLOLOLOL

Reviewer activates a feature that makes Intel slower: STUPID REVIEWER! AMDSUXLOLOLOLOL
 
Yea I mean I expected Intel to rule in gaming. No doubt about it. So if you plan to upgrade to purely game just get a 8700k. But if you game AND do production work, the 2700k is the way to go.
 
Brackle said:
Yea I mean I expected Intel to rule in gaming. No doubt about it. So if you plan to upgrade to purely game just get a 8700k. But if you game AND do production work, the 2700k is the way to go.
Click to expand...

They are differentiated but close enough in either that it goes like this:

If you Game, get an Intel.
If you Game primarily and do some productivity, still get Intel.
If you focus on productivity and do some gaming, get AMD.
If you focus only on productivity, get AMD.
 
KazeoHin said:
They are differentiated but close enough in either that it goes like this:

If you Game, get an Intel.
If you Game primarily and do some productivity, still get Intel.
If you focus on productivity and do some gaming, get AMD.
If you focus only on productivity, get AMD.
Click to expand...

Yea man I do a lot of productivity for work, but I use the work laptop for that....Otherwise I would get the AMD. Since i mostly game now when im home on my main PC, Been really leaning toward the 8700k, but it really does depend how well this Xeon X5675 does at 4k...If i go 1440p/144hz then for sure I am going 8700k Going to see if I can get it to 4.6ghz on the Xeon X5675, if it does well im stickin to what I got.
 
KazeoHin said:
They are differentiated but close enough in either that it goes like this:

If you Game, get an Intel.
If you Game primarily and do some productivity, still get Intel.
If you focus on productivity and do some gaming, get AMD.
If you focus only on productivity, get AMD.
Click to expand...

well you should change that to if you game at 1080p actually. Not much of a difference on higher resolutions. I honestly cant say don't get amd if you game. 100 fps vs 110 fps is not much. Still a smooth experience at 1080p.

I have intel but my next system will likely be AMD. I do mild gaming but honestly if I gamed more I would still get AMD. But it will likely be zen 2.

I got my 7600k running at 4.9ghz. So Hard to justify buying anything right now lol.
 
  • Like
Reactions: N4CR
like this
NKD said:
well you should change that to if you game at 1080p actually
Click to expand...

If and only if you're only going to play the games you're already playing with the supporting GPU you already own until you next upgrade your CPU.

I think most people will want to play newer games and upgrade their GPU between CPU upgrades, and that's why the 8700k gets the nod.

Hell, many will probably grab higher refresh rate monitors and/or VR headsets before they upgrade their CPUs as those components increase in production rate and decrease in cost.
 
juanrga said:
AMD SMT implementation isn't more efficient.
Click to expand...

I am not going to get into most of this, but this statement is correct. SMT has always been about remediating utilization issues in chips. Being "more efficient" at best would imply it is able to fix more shortcomings, most commonly in the decoder. CPU threading is, and has always been, a ploy to increase execution unit utilization when a single thread would not do so normally.

To whit: you actually do not want SMT to show huge gains, ideally. This is a resource balancing issue. The bigger gains SMT shows, the bigger disparity between your actual workload and the execution units in your chips. Some of this imbalance can be recovered with highly-threaded workloads, and that's great (and is the point of SMT). But don't call it out as being "more efficient". It usually means your single-threaded throughput is not keeping up with a single threaded workload, for one reason or another. Reasons can be the decoder, memory subsystem, procunit starvation, etc. SMT throws (currently) double the uOps to the procunits, and thus increases utilization for the instruction streams present, and mitigates stalls in a single thread.

For clarity, I am not trying to crap on AMD's implementation. This is an attempt at a layman's explanation of SMT's design goals and role in a CPU.
 
You must log in or register to reply here.
Back
Top