In leaked benchmarks, AMD Zen falls short of Intel Haswell, but shows promise nonetheless
- Thread starter Shawn Knight
- Start date
I can't stand how sites that are supposed to cover tech news either get things so wrong or are influenced (somehow) to make things look better or in this case, worse than they really are.
If you consider that throughout history, engineering samples generally perform much worse than the eventual production part, which historically perform anywhere from 20-50% better than the ES's then we can calculate some very favorable results from this one leaked, potentially real benchmark. At 2.8Ghz, it performs between an i5 and i7 (those numbers aren't much different than the Skylake units). We need to understand that ES's are sent to OEM's to provide feedback for additional tweaking prior to production launch. Vishera, for example sampled at anywhere from 1.8GHz to 3.3GHz while the production part came out at 4.0GHz. So, if we take just the frequency, and place it at 4.0GHz, we can basically (obviously not actually, because we don't know the performance/clock curve of the architecture) calculate that performance would exceed that of the i7 4790 and place it on par with most likely the 6700k or perhaps even the Kabylake update i7 version. That is with frequency alone. With additional tweaks, it could outperform the i7 by even more. Also keep in mind that AoTS doesn't seem to scale much past 6 cores, so this Zen CPU can outperform the i7's even more with multithreaded apps/games that can take advantage of the additional threads.
So long as it is priced well (~$300), it will be a fantastic product for the money. Sites that report the way this article has, lose credibility. Anandtech would never report in this manner. Way to shoot yourself in the foot Techspot. Whereas I had respect for you in the past I have less of it now after reading the quality lacking article this is.
If you consider that throughout history, engineering samples generally perform much worse than the eventual production part, which historically perform anywhere from 20-50% better than the ES's then we can calculate some very favorable results from this one leaked, potentially real benchmark. At 2.8Ghz, it performs between an i5 and i7 (those numbers aren't much different than the Skylake units). We need to understand that ES's are sent to OEM's to provide feedback for additional tweaking prior to production launch. Vishera, for example sampled at anywhere from 1.8GHz to 3.3GHz while the production part came out at 4.0GHz. So, if we take just the frequency, and place it at 4.0GHz, we can basically (obviously not actually, because we don't know the performance/clock curve of the architecture) calculate that performance would exceed that of the i7 4790 and place it on par with most likely the 6700k or perhaps even the Kabylake update i7 version. That is with frequency alone. With additional tweaks, it could outperform the i7 by even more. Also keep in mind that AoTS doesn't seem to scale much past 6 cores, so this Zen CPU can outperform the i7's even more with multithreaded apps/games that can take advantage of the additional threads.
So long as it is priced well (~$300), it will be a fantastic product for the money. Sites that report the way this article has, lose credibility. Anandtech would never report in this manner. Way to shoot yourself in the foot Techspot. Whereas I had respect for you in the past I have less of it now after reading the quality lacking article this is.
Last edited:
Squid Surprise
Posts: 5,596 +5,227
So you slam Techspot for their article, but decide to give completely unsubstantiated numbers to project that this CPU will be on par with a Kaby Lake i7?!?!
I can understand being unhappy that Techspot says this benchmark is a disappointment for AMD (it IS, but only at face value).... But NO ONE (well, Hardreset probably can) can project where this CPU will fall given that it won't even be released until next year...
You have some difficulties understanding, let me help:
Provide proof for that "released next year" part, thanks. AMD has said long time that small amounts may be available 2016 but 2017 full volume availability.
Squid Surprise
Posts: 5,596 +5,227
Welcome back troll - unsubstantiated.... As in we can't just adjust the frequency and multiply performance (the poster actually admitted this, yet did it anyways), and the article stated that this CPU will see general release in 2017 - that would be next year....
Google AMD Zen launch 2016.
And article stated: AMD has said it doesn’t expect to ship Zen with full market availability until sometime in 2017.
oldtink
Posts: 9 +3
At my age I have grown tired of this type of article where a tech site, tech writer and even the tech editor are willing to "go out on a limb" with dubiously leaked info on an ES part and speculate on what it means for the "future of AMD". I am afraid you have just lost two doses of respect from someone who has watched tech news from when clock speeds were measured in Mhz.
Tip my hat to "Hardreset" for his comments but I realize that for some it is an exercise in futility.
Tip my hat to "Hardreset" for his comments but I realize that for some it is an exercise in futility.
Squid Surprise
Posts: 5,596 +5,227
Full market availability would mean the general launch.... Only the server parts will be available in Q4.... And this is irrelevant to my argument anyways.... My point is, you can't project what something will do before it exists.... And it won't be out for months at the earliest....
Squid Surprise
Posts: 5,596 +5,227
The article took great pains to point out that this was just one set of dubious benchmarks and stated that full conclusions could not be drawn... No need to complain about something that didn't actually happen...
Your knowledge has been proven abysmal again! Server parts are coming LATER than desktop parts ""(y)""
Squid Surprise
Posts: 5,596 +5,227
Actually...
http://www.kitguru.net/components/c...d-cpu-roadmap-appears-online-zen-coming-2017/
We don't have any hard data on launch yet... but again... it doesn't matter!!! The point is (I know, you're just not very good at reading) you can't predict the future!! NO ONE CAN!
AMD said ages ago that desktop Zen comes before server Zen.
This is so funny, you have zero knowledge but accuse others for trolling
amstech
Posts: 2,662 +1,824
There is A LOT riding on Polaris/ZEN.
Atleast one of the two needs to be successful.
Now for some obvious, and good news.
As far as how well the architecture will perform every tech site is making predictions, Techspot is no different. If all you see is a biased site here, its time to put the keyboard down and back away. I'd quote some other statements made by tech sites that will make Techspot look like they work for AMD.
Yall taking your silicone love to the next level!
Ease up everyone god damn.
Atleast one of the two needs to be successful.
fool.com said:
Now for some obvious, and good news.
fool.com said:
As far as how well the architecture will perform every tech site is making predictions, Techspot is no different. If all you see is a biased site here, its time to put the keyboard down and back away. I'd quote some other statements made by tech sites that will make Techspot look like they work for AMD.
Yall taking your silicone love to the next level!
Ease up everyone god damn.
Squid Surprise
Posts: 5,596 +5,227
AMD said ages ago that desktop Zen comes before server Zen.
This is so funny, you have zero knowledge but accuse others for trolling
Again... it doesn't matter - my point was about predicting the future - how far into the future is irrelevant... and AMD has been quite quiet about official releases... so unless you can give me some hard facts (backed up by credible links), don't bother trolling...
cliffordcooley
Posts: 13,141 +6,441
So you are taking a post based on speculation and making even more speculation. All because you feel the desire to uplift AMD in front of people, who have already expressed a desire for AMD to succeed with Zen. Way to go mate!
Another hopeless effort where you try to hide your poor knowledge. Pathetic. AMD's CEO should be credible source.
You have once again proven wrong. Now, would you finally shut up?
Not speculation, but calculation.
cliffordcooley
Posts: 13,141 +6,441
I know. I did some of the very same calculations when making the first comment in the thread. No matter how you look at it, calculation based on speculation is still speculation.
shark975
Posts: 141 +179
What a crap, biased, anti-AMD article.
First off author, let me help you out with a little basic math. The i7 4790 at the top of those charts has a max boost clock of 4 ghz. The Zen chip's max boost is 3.2. That's a 25% difference. If the AMD chip was running at the i7 4790's clock, it would have scored 58X1.25=72.5 FPS, beating the top Intel by over 10%.
Second you claim:
"It’s also worth pointing out that the Zen ES is being compared to Intel’s three-year-old Haswell processors. The Core i5-4670K has just four cores and four threads while the Core i7-4790 features four cores and eight threads and both are built on a 22nm process. The Zen ES reports eight cores, 16 threads and is a 14nm part yet it can’t keep pace with Intel’s old chips?
Making matters even worse for AMD is the fact that Intel is planning to release Cannonlake, its 10nm die shrink of Kaby Lake, next year. AMD has said it doesn’t expect to ship Zen with full market availability until sometime in 2017."
Too bad you're being blatantly disingenuous, because Intel's "brand new" "14nm" Skylake chips barely improve performance on "3 year old, 22nm" Haswell at all. Here's just one example link to prove it http://www.anandtech.com/bench/product/1260?vs=1543. So, if Zen is beating Haswell, it should also beat/compare to "brand new" Skylake. And given that Intel's last several architectures/refreshes have all improved performance little (which is why an ancient 2500k is still fine for gamers), it's likely "Kaby Lake" wont be much faster either though that's speculation.
The fact the Zen part is 8/16 threads supposedly is a bit more troubling, but not entirely. First off does that benchmark scale to more than 8 threads? If it doesn't or scales poorly, it's all moot. Second, AMD has offered double the (pseudo) core count of Intel for a lower price in the past, which is what really matters. For example the 8350. No reason they couldn't do the same with Zen, in theory.
I dont understand why Nvidia/Intel fanboys seem to think (pretend, actually) unrelated things like efficiency are most important. What's most important are price and performance. This came up recently when all the Nvidia fanboys were going gaga over the GTX 1060 offering similar performance (except in you know, the important one of DX 12) to the RX 480 for a TDP of 120 watts vs 150. It's like, I dont know, I'm not a Nun, who cares? I dont care about 30 watts. I care about price and performance. Similarly if AMD offers a 16 thread chip that is putting up similar performance to an 8 thread Intel chip for a lower price, IT DOESNT MATTER if omgosh it's getting half the performance per thread, people will buy it.
That's all hypothetical in a huge way at this point, of course.
Or if you wanted to flip the debate to Nvidia fanboy terms, you could ask why GTX 1060 is offering more than 30% less teraflops for only 20% less power use, in other words it's less efficient than AMD RX 480. Except said Nvidia fanboys would suddenly become rational and correctly point out, Flops doesn't really matter, game performance does. But somehow they become blind to what's important when it suits them.
Just tired of reviewers damning AMD for whatever biased irrelevant stat they want to focus on today, be it thread count, so-called power efficiency, process node, etc. Those are ancillary things, price and performance on offer are the big things.
First off author, let me help you out with a little basic math. The i7 4790 at the top of those charts has a max boost clock of 4 ghz. The Zen chip's max boost is 3.2. That's a 25% difference. If the AMD chip was running at the i7 4790's clock, it would have scored 58X1.25=72.5 FPS, beating the top Intel by over 10%.
Second you claim:
"It’s also worth pointing out that the Zen ES is being compared to Intel’s three-year-old Haswell processors. The Core i5-4670K has just four cores and four threads while the Core i7-4790 features four cores and eight threads and both are built on a 22nm process. The Zen ES reports eight cores, 16 threads and is a 14nm part yet it can’t keep pace with Intel’s old chips?
Making matters even worse for AMD is the fact that Intel is planning to release Cannonlake, its 10nm die shrink of Kaby Lake, next year. AMD has said it doesn’t expect to ship Zen with full market availability until sometime in 2017."
Too bad you're being blatantly disingenuous, because Intel's "brand new" "14nm" Skylake chips barely improve performance on "3 year old, 22nm" Haswell at all. Here's just one example link to prove it http://www.anandtech.com/bench/product/1260?vs=1543. So, if Zen is beating Haswell, it should also beat/compare to "brand new" Skylake. And given that Intel's last several architectures/refreshes have all improved performance little (which is why an ancient 2500k is still fine for gamers), it's likely "Kaby Lake" wont be much faster either though that's speculation.
The fact the Zen part is 8/16 threads supposedly is a bit more troubling, but not entirely. First off does that benchmark scale to more than 8 threads? If it doesn't or scales poorly, it's all moot. Second, AMD has offered double the (pseudo) core count of Intel for a lower price in the past, which is what really matters. For example the 8350. No reason they couldn't do the same with Zen, in theory.
I dont understand why Nvidia/Intel fanboys seem to think (pretend, actually) unrelated things like efficiency are most important. What's most important are price and performance. This came up recently when all the Nvidia fanboys were going gaga over the GTX 1060 offering similar performance (except in you know, the important one of DX 12) to the RX 480 for a TDP of 120 watts vs 150. It's like, I dont know, I'm not a Nun, who cares? I dont care about 30 watts. I care about price and performance. Similarly if AMD offers a 16 thread chip that is putting up similar performance to an 8 thread Intel chip for a lower price, IT DOESNT MATTER if omgosh it's getting half the performance per thread, people will buy it.
That's all hypothetical in a huge way at this point, of course.
Or if you wanted to flip the debate to Nvidia fanboy terms, you could ask why GTX 1060 is offering more than 30% less teraflops for only 20% less power use, in other words it's less efficient than AMD RX 480. Except said Nvidia fanboys would suddenly become rational and correctly point out, Flops doesn't really matter, game performance does. But somehow they become blind to what's important when it suits them.
Just tired of reviewers damning AMD for whatever biased irrelevant stat they want to focus on today, be it thread count, so-called power efficiency, process node, etc. Those are ancillary things, price and performance on offer are the big things.
cliffordcooley
Posts: 13,141 +6,441
It was rumored speculation for the 100th time.
Ohh and we are to take your claim based on speculation even more seriously?
As others have pointed out to you, the numbers I provided are indeed substantial, but only to a point. As I said, without actually knowing the full performance/clock curve of the architecture, we won't really "know", but we can project, that is for sure, how you cannot follow the basic math from the % of increased clock speed, relative to the performance score in AoTS, I don't quite understand, but not everyone is good at math, which is understandable, but please trust me that it is a legitimate calculation, with a result that may be off by a few percentage points give or take, but that's just assuming a linear frequency to performance relationship.
As far as projections being possible, prior to this benchmark being leaked, I would agree with you. We haven't had any numbers to work off of, however, if (and that's a big if), the benchmark is a legitimate benchmark from a legitimate Zen ES, then yes, anyone can actually project performance numbers using the data that was provided from that benchmark. If you want me to break it down for you...
If you take the benchmark averages as shown here: http://cdn.wccftech.com/wp-content/uploads/2016/08/AMD-Zen-ES-AotS-Benchmarks.jpg
We see:
i7 4790 at 3.6GHz base with a score of 65.4fps
Zen ES at 2.8GHz base with a score of 58fps
i5 4670K at 3.4GHz base with a score of 52.6fps
FX8350 at 4.0GHz base with a score of 42fps
Now, seeing as how Vishera ES's that went out ranged anywhere from 1.8GHz to 3.3GHz. The Cinebench 11.5 score for the 3.3GHz ES was 5.73 vs the Final FX8350 product of 6.89. That is roughly a 20% boost over the ES, and differences have been even greater throughout history between ES and Production (up to 50%).
If we take the low end of that potential improvement range at just 20% and add it to the performance of this ES in this game, FPS jumps up to 70.6fps, which makes it 8% faster than the i7. That is just using 20% as a base, since we've seen this time and again from ES to production.
If we instead take frequency, and bump Zen to 4.0GHz, that is a 43% increase in clock speed. If we use frequency alone, assuming a linear frequency/performance relationship (**which may not be the case, but hypothetically here**), and increase the score of the Zen ES by 43%, we get 82.94fps which is 27% better than the i7 4790.
Now if we take the higher end of historical data and improve performance of the Zen ES by upwards of 50%, we end up with 87fps which is 33% faster than the i7 4790.
So, in conclusion, yes, we can indeed "project" with the above numbers the following:
Zen with 20% improvement over ES could have a score of 70.6fps (8% faster than i7 4790)
Zen at 4GHz base could have a score of 82.94fps (27% faster than i7 4790)
Zen with 50% improvement over ES could have a score of 87fps (33% faster than i7 4790)
Let me say, however, before trolls try to troll, that "projected" figures and actual figures may vary. Projections are just that, a forecasting based on current information, historical data, and assumed figures to fill in the gaps. We will not truly know anything until the production part of Zen is released, WHICH, will be at the end of 2016 in small quantities, but with general availability in Q1 of 2017. So, you can take my calculations with a grain, I don't mind, that's what they're for, but for those that understand the historical differences between ES's and production units, these projection may be more substantial than others may think.
Last edited:
Neat, another product from AMD that offers performance similar to someone else's tech from three years ago, yet the AMD fanboys are comically still dancing in the streets. It's like a bunch of hillbillies bragging that their wife/sister just graduated from high school.
Impressive to them, but nothing special to the rest of us.
Impressive to them, but nothing special to the rest of us.
Feel free to go create your own tech site, then spew all you want, nerd.
Squid Surprise
Posts: 5,596 +5,227
I don't think you understand the definition of "evidence".... just giving a quote means nothing.... Give us some facts backed up by actual links... You claim AMD's CEO as your source - where's the link?
Now... you're once again just trolling - this article is speculating.... as we all are... because all we have is one set of leaked benchmarks that may or may not be valid.
NO ONE can predict the future - if AMD could, they'd be making profits instead of losing gazillions each year...
In my experience, those who scream the word "fanboy" 4x in the same paragraph are usually even bigger ones themselves. Seriously, calm down and wait for the reviews like the rest of the grown ups...
Last edited:
Similar threads
Latest posts
-
Nintendo DMCA lawyers shut down everything Mario on Garry's Mod- VariableSpike replied
