Ira Wechsler
Posts: 24 +12
AMD has stated it expects approximately a 15% IPC gain. All the demos and benchmarks are pointing to that and this delusional clown has zero to back up his prognostications.
Zen 2 might not offer the IPC increase you expect
- Thread starter mongeese
- Start date
wiyosaya
Posts: 9,773 +9,662
They found the perfect home!
There is so much wrong with this article, starting from AMD talking about the IPC increase to the server processors and consumer processors using the same chiplet design which would likely end up with similar latency across all cores and across all models. Also without having actual game benchmarks you have to try to draw conclusions from any place you can which would be the leaked benchmarks and any real ones should give some idea of performance differences, but will never translate directly because the workload is very different for each game and each benchmark. I would also point out that the current ryzen processors are plenty fast enough for gaming and the new ones are supposed to have a significant IPC increase (based on what AMD has said) they might also be able to hit higher clocks which would all point towards quite a bit better gaming experience. What I have seen in official announcements would indicate a 20% improvement, but realistically I expect more 10-15% and will wait for the real benchmarks to get confirmation on the actual improvement.
bluetooth fairy
Posts: 235 +166
Because we talk about IPC, we need to compare apples to apples. Quotations were taken from TS article, it's not mine, btw.
2600X just cant compete against 8700, thats why the prices are diff. Shall it could compete, it would be more expensive. Remember the times of Athlon 64 X2.
I don't think most AMD fanboy could count to 10 when 64 was even released. I have no idea what the author of this article was even trying to say. I mean Steve who writes for the same website did a same clock speed benchmark and this guy completely ignores his research. Where the hell did he get 3%?
Soulburn74
Posts: 144 +78
Here let me set the record straight..... I mean, Let me speculate about the already mentioned speculations.........
In what world is AMD in front of Intel in productivity and scientific tasks? There are certain benchmarks and apps, they are faster at but in the absolute majority, the Intel is faster on ringbus and dominating them with the mesh /SkylakeX/ without AVX. With AVX, it is not even competition, Intel is far ahead.
Get your facts together.
Get your facts together.
Evernessince
Posts: 5,469 +6,158
1. Actually yes the current 2700X trades blows with the 9900K, let alone next gen products.
2. The 2700X already goes up to 4.35 GHz out of the box.
3. AMD is already more power efficient the Intel, they don't need 7nm for this. 7nm will just widen the gap.
4. Chicken or the Egg
5. Good thing Zen 2 will have maximum performance per core. Mind you AMD is not only increasing clock speed and IPC, they are also adding 2 more max boost XFR CPUs with their 16 core product. That's 4 cores that can boost to the max frequency. Intel's boost tech can only do a single core at max frequency.
6. Well they already compete on a per core basis as there is only a 3% IPC difference in general applications.
Perhaps you should research before you post. You didn't know that AMD has already cracked 4 GHz and you didn't know that AMD already has better power efficiency. This appears more like a drunken rant.
The 2700X clocks at 4.35 GHz out of the box. If you are going to bash someone for having incorrect numbers make sure you aren't doing the same.
You forget to mention the part when the AMD midrange product beat the Intel Flagship 8700K in nearly every productivity and office benchmark.
Depends on your use case. The fact that Cinebench is used by most all reviews clear
https://www.techspot.com/article/1616-4ghz-ryzen-2nd-gen-vs-core-8th-gen/
Looks like you used your time stone to move yourself to a world in which AMD is in fact better in nearly all productivity and office apps.
Your amd bias is a bit too much don't you think? Intel's pricier for a reason. Even at 4ghz the 9900k is better than an oc'ed 2700x - oh right it doesnt oc much at all. Can you even call it an oc?
1. Actually yes the current 2700X trades blows with the 9900K, let alone next gen products.
2. The 2700X already goes up to 4.35 GHz out of the box.
3. AMD is already more power efficient the Intel, they don't need 7nm for this. 7nm will just widen the gap.
4. Chicken or the Egg
5. Good thing Zen 2 will have maximum performance per core. Mind you AMD is not only increasing clock speed and IPC, they are also adding 2 more max boost XFR CPUs with their 16 core product. That's 4 cores that can boost to the max frequency. Intel's boost tech can only do a single core at max frequency.
6. Well they already compete on a per core basis as there is only a 3% IPC difference in general applications.
Perhaps you should research before you post. You didn't know that AMD has already cracked 4 GHz and you didn't know that AMD already has better power efficiency. This appears more like a drunken rant.
The 2700X clocks at 4.35 GHz out of the box. If you are going to bash someone for having incorrect numbers make sure you aren't doing the same.
You forget to mention the part when the AMD midrange product beat the Intel Flagship 8700K in nearly every productivity and office benchmark.
Depends on your use case. The fact that Cinebench is used by most all reviews clear
https://www.techspot.com/article/1616-4ghz-ryzen-2nd-gen-vs-core-8th-gen/
Looks like you used your time stone to move yourself to a world in which AMD is in fact better in nearly all productivity and office apps.
About Adored he's a well known bullshiter I don't know how long you've been watching him but there's a reason why his content is banned on some subreddits. Lots of viewers are so brainwashed it's absurd. Like one guy put it a few years ago he's the tech version of keemstar.
144hzGamer
Posts: 214 +131
Your amd bias is a bit too much don't you think? Intel's pricier for a reason. Even at 4ghz the 9900k is better than an oc'ed 2700x - oh right it doesnt oc much at all. Can you even call it an oc?
1. Actually yes the current 2700X trades blows with the 9900K, let alone next gen products.
2. The 2700X already goes up to 4.35 GHz out of the box.
3. AMD is already more power efficient the Intel, they don't need 7nm for this. 7nm will just widen the gap.
4. Chicken or the Egg
5. Good thing Zen 2 will have maximum performance per core. Mind you AMD is not only increasing clock speed and IPC, they are also adding 2 more max boost XFR CPUs with their 16 core product. That's 4 cores that can boost to the max frequency. Intel's boost tech can only do a single core at max frequency.
6. Well they already compete on a per core basis as there is only a 3% IPC difference in general applications.
Perhaps you should research before you post. You didn't know that AMD has already cracked 4 GHz and you didn't know that AMD already has better power efficiency. This appears more like a drunken rant.
The 2700X clocks at 4.35 GHz out of the box. If you are going to bash someone for having incorrect numbers make sure you aren't doing the same.
You forget to mention the part when the AMD midrange product beat the Intel Flagship 8700K in nearly every productivity and office benchmark.
Depends on your use case. The fact that Cinebench is used by most all reviews clear
https://www.techspot.com/article/1616-4ghz-ryzen-2nd-gen-vs-core-8th-gen/
Looks like you used your time stone to move yourself to a world in which AMD is in fact better in nearly all productivity and office apps.
About Adored he's a well known bullshiter I don't know how long you've been watching him but there's a reason why his content is banned on some subreddits. Lots of viewers are so brainwashed it's absurd. Like one guy put it a few years ago he's the tech version of keemstar.
He works for amd, is our only explanation loool the dude is everywhere, everyday preaching it. Sad part is that he does not realize how fool he seems when he tries to make ppl believe amd trades blows with intel.
9900k completly obliterates 2700x on everything and it also can be used with 4133mhz ram wich improves even more. 9700k at 5ghz is almost as fast as 2700x and hella faster in games or single threaded.
Also amd isnt all about ipc, there are latencies going on.
If zen 2 beats Intel in gaming I will delete my account. You read it here first.
Adoredtv is a scammer.
Evernessince
Posts: 5,469 +6,158
Do you actually have something substantial here or are you just going to hurl insults?
bluetooth fairy
Posts: 235 +166
I'm not for fanboizm here. My point was to point on what was incorrect. How on Earth could the author say 3 and have a link where it is 7 to 14?
You could use the same link to find where exactly 8700 was beaten:
https://www.techspot.com/article/1616-4ghz-ryzen-2nd-gen-vs-core-8th-gen/page2.html
2600 became the midrange product because it can't compete in higher price segment. You probably lack of product placement understanding due to some obvious fanboizm.
Evernessince
Posts: 5,469 +6,158
looncraz
Posts: 9 +6
About 4.0~4.1GHz all-core judging by how 7nm behaves for power and the assumption that the uncore (IO die) power remained unchanged and how 2700X scales.
Cinebench should have a pretty healthy increase, but we don't know how well SMT will behave with the changes AMD made - it's possible that relative scaling will be worse as the single threaded performance is improved, thereby utilizing core resources for one thread more efficiently.
looncraz
Posts: 9 +6
I agree with your methods - comparing the 2600X to the 8700k at the same clocks is a proper architectural comparative match - but there's a factor that needs to be considered:
When you overclock Ryzen's memory, that gap narrows ever more. The DDR4-3200 CL14 memory they used, though, is quite good, but they were probably seeing 65ns memory latency or higher. Dropping that to the 60ns region can gain another 2~3% in performance, which is just insane.. and illustrates where Ryzen's bottleneck resides.
Zen 2 has doubled the L3 caches, we know this for a fact. We also know they did this without sacrificing latency. They doubled the LSU and widened the front end. These are improvements that are directly aimed at games, no way around it. Unfortunately, they also moved the memory controller further away, adding memory latency... though we don't know how much.
All of this is to say: AMD's IPC doesn't even need to increase much to reach Intel in games... just their memory throughput does. The large L3 and other improvements should mostly see to that. The IPC is the icing on top, but Zen's current IPC capabilities aren't being used by games because they exhibit significant random access over too large of a memory area. This is the worse area for Zen's performance, which is why games do comparatively worse than most other tasks (AVX-256 notwithstanding... something Zen 2 will soon beat Intel at, I believe).
Ira Wechsler
Posts: 24 +12
This article is full of conjecture and totally lacking in any convincing facts since AMD itself said zIPC gains would make it equal to the current Intel offerings like the 9900K as far as IPC. This is a trash aricle probably from someone who has some material benefit in slanting things against AMD. AMD interviews indicated an IPC improvement would be bnetween 10 and 15%. AMD tends to be conservative in their projections. At CES they ran a Zen2 engineering sample at the same base clock speed as an Intel 9900K. And clock for clock it performed slightly better so that is IPC when you run clock to clock and compare results.The cache, the logic, latency and other improvements have contributed to the improvement in IPC. There is no need to pay attention to the complete bs of this article.
looncraz
Posts: 9 +6
The Zen 2 CPU was likely running 500~700MHz SLOWER than the 9900k, actually, but we don't have any official word on that. The closest I've been told is that the CPU had a turbo of 4.5 and a base of 3.8GHz and was a 65W part. The person had no clue how it was actually configured at CES, but suspected it was not strictly stock since they wouldn't want to chance it not winning on stage and they didn't want to win by too much and kill Ryzen 2000 sales, so it was probably some odd clock like 4.025 or 4.075GHz.
A 2700X at the same frequency as the 9700k would already win in Cinebench, BTW, so this only represents a 10~12% increase in performance per clock, which isn't very hard to believe.
IPC by adding better instructions is useless if the software doesn't use the better instruction set. We've seen this over and over again with AMD vs Nvidia, where the advantages of AMD cards weren't used in games, so they looked slower. Or when games use just one thread, so that Intel CPUs look faster. If your software doesn't know how to use the hardware, don't blame it on the hardware manufacturer.
bluetooth fairy
Posts: 235 +166
My bad I ran into the same issue again thanks to that the article is talking mostly about IPC. They at AMD had already had good steps in Zen+ in terms of cache and memory latency decrease compared to original Zen. They know their microarch, die and package design pros and cons. It would be good shall they gain some % here and there, incl. clock boost, probably mostly the latter.
Arbie
Posts: 531 +898
Wrong. Ryzen is close on per-core and offers more cores for less money. In fact most enthusiasts (and all general users) shouldn't even consider an Intel chip. That's been true for a couple of years now.
What you meant to claim was that AMD cannot (yet) quite match Intel on a per-core basis. Please think about what you're saying. Even this will probably change next month.
And don't forget that if AMD hadn't pulled off a miracle, your vaunted Intel would still be giving you 2017 performance. I buy AMD even if they don't (yet) win in every metric. They are so close it's the only sensible thing to do. My 1800X purchase - besides it being an amazing chip - helped AMD proceed with Zen 2 and beyond. We will all benefit. Thank you AMD!
Last edited:
For Zen 2 used on demo, clock speed is unknown, so is IPC improvement. However we do know that i9-9900K all core turbo max is 4.7 GHz. On low AVX-loads, it should run at 4.7 GHz. So if Zen2 is equally fast as i9-9900K, then 4.025 GHz clock speed would mean 16% IPC improvement vs i9-9900K. If IPC improvement was 10%, then Zen2 clock speed was 4.27 GHz.
As said before, Zen+ has exactly same architecture Zen has.
GreenNova343
Posts: 454 +336
It's always hard to compare IPC in processors, even between different generations of the same manufacturer, let alone between different manufacturers, because we really aren't able to actually say, "CPU X has Y instructions per cycle per core". Simply put, certain instructions can be processed more quickly by a CPU than other instructions. This is also why you don't always see equal performance in applications (gaming or non-gaming), even among games that supposedly use the same gaming/graphics engine: you can see trends & maybe similar performance gaps/issues, but you won't see identical performance.
Now, we can look at real-world performance levels to say, "CPU X performs Y% better/worse than CPU Z"...but even then we're not looking at anything that is easily converted into any sort of IPC equivalent. Mainly this is because when you're comparing multiple CPUs, you're not only looking at different generations of processors & different manufacturers, you're also looking at differing numbers of cores...& also looking at different core frequencies.
And that points out the weakness of the supposed "king CPU", the i9-9900K: Techspot's own review showed it providing almost identical performance to the i7-9700K, a cheaper (release MSRPs were $488USD for the i9 vs. $374USD for the i7) & only slightly slower (max Turbo for all cores/1 core is 5.0Ghz/4.7GHz for the i9, 4.9GHz/4.6GHz for the i7). And even worse, there was little to no performance gain over the earlier i7-8700K, a chip with fewer cores/threads (6/12 vs. the 8/16 of the 9700K & 9900K) & slower frequencies (max Turbo 4.3GHz all cores/4.7Ghz with a single core). That's not IPC "improvement"; if anything, that's IPC devolvement (same performance at a faster clock speed = fewer instructions being performed per clock rate).
When you start comparing the Ryzen to those figures, since the 9700K & 9900K apparently had IPC reductions, you have to go back & look at the 8700K vs. the Ryzen chip...which is significant given that, aside from their core differences (6C/12T for the 8700K, 8C/16T for the 2700X), their frequencies are pretty close (3.7GHz base for both, 4.3GHz Turbo for all cores with the 8700K vs. 4.3GHz maximum Performance Boost 2 for the 2700X)...& since we're looking at CPU performance, not GPU performance, we use the 1080p test results Yes, in some games the 8700K showed ~20% better performance (implying that the 2700X has 0.8 IPC per 1 IPC of the 8700K)...but in other games the difference was only 7% (I.e. 0.93IPC for the 2700X vs. the 8700K), & even identical performance in Forza Horizon 4 (I.e. identical IPC for each one).
What I wouldn't mind seeing is another round of CPU testing, where not only do we see near-identical hardware but we also see all of the CPUs limited to the same per-core frequencies (I.e. 4.0GHz, or something that all testable CPUs can reach without resorting to overclocking). Then we would have comparable numbers to be able to truly say which CPU has more IPC than another.
Now, we can look at real-world performance levels to say, "CPU X performs Y% better/worse than CPU Z"...but even then we're not looking at anything that is easily converted into any sort of IPC equivalent. Mainly this is because when you're comparing multiple CPUs, you're not only looking at different generations of processors & different manufacturers, you're also looking at differing numbers of cores...& also looking at different core frequencies.
And that points out the weakness of the supposed "king CPU", the i9-9900K: Techspot's own review showed it providing almost identical performance to the i7-9700K, a cheaper (release MSRPs were $488USD for the i9 vs. $374USD for the i7) & only slightly slower (max Turbo for all cores/1 core is 5.0Ghz/4.7GHz for the i9, 4.9GHz/4.6GHz for the i7). And even worse, there was little to no performance gain over the earlier i7-8700K, a chip with fewer cores/threads (6/12 vs. the 8/16 of the 9700K & 9900K) & slower frequencies (max Turbo 4.3GHz all cores/4.7Ghz with a single core). That's not IPC "improvement"; if anything, that's IPC devolvement (same performance at a faster clock speed = fewer instructions being performed per clock rate).
When you start comparing the Ryzen to those figures, since the 9700K & 9900K apparently had IPC reductions, you have to go back & look at the 8700K vs. the Ryzen chip...which is significant given that, aside from their core differences (6C/12T for the 8700K, 8C/16T for the 2700X), their frequencies are pretty close (3.7GHz base for both, 4.3GHz Turbo for all cores with the 8700K vs. 4.3GHz maximum Performance Boost 2 for the 2700X)...& since we're looking at CPU performance, not GPU performance, we use the 1080p test results Yes, in some games the 8700K showed ~20% better performance (implying that the 2700X has 0.8 IPC per 1 IPC of the 8700K)...but in other games the difference was only 7% (I.e. 0.93IPC for the 2700X vs. the 8700K), & even identical performance in Forza Horizon 4 (I.e. identical IPC for each one).
What I wouldn't mind seeing is another round of CPU testing, where not only do we see near-identical hardware but we also see all of the CPUs limited to the same per-core frequencies (I.e. 4.0GHz, or something that all testable CPUs can reach without resorting to overclocking). Then we would have comparable numbers to be able to truly say which CPU has more IPC than another.
looncraz
Posts: 9 +6
You're confusing performance and IPC - no worries, insanely easy to do.
Ryzen already outperforms Coffee Lake, per clock, by about 3% in Cinebench. This is mostly thanks to its superior SMT scaling.
Zen 2 would only need about a 10~12% IPC improvement, in Cinebench, for 4.0~4.1GHz with the Zen 2 sample.
At 4.6Ghz, the 2700X would match the 9900k at 4.7GHz.
https://www.techspot.com/news/79970-amd-navi-based-rx-3080-xt-rumored-offer.html
Hey... if I take Rob Thubron description saying that speculation piece based on credible leaks are synonyms of "Bogus", in the case of AdoredTV, than an opinion piece, based on nothing but assumptions, should fall even lower by Techspot standards.
For our own good, Techspot should remove this article since it is not based on anything relevant.
Hey... if I take Rob Thubron description saying that speculation piece based on credible leaks are synonyms of "Bogus", in the case of AdoredTV, than an opinion piece, based on nothing but assumptions, should fall even lower by Techspot standards.
For our own good, Techspot should remove this article since it is not based on anything relevant.
Similar threads
