anand's jaguar article

[Exophase]

How many times I have to say that we are using GFLOPs because we have no other data now?

Sorry, I don't understand what that has to do with my response..

Those 18 CUs are split in a 14+4 balance. Albeit the 4 CUs can switch mode to rendering if needed, the main task of the 4 CU will be on computing non-graphics stuff: e.g. physics effects. In one sense those 4 CU are like the streamlined coprocessors of the Cell

http://www.vgleaks.com/world-exclusive-orbis-unveiled-2/

That was a vague rumor that has basically been completely debunked by Sony's reveal. It's almost certainly wrong. Doesn't seem to stop it from constantly being repeated though.

If you want to introduce the GPU of the i7 comparison then introduce the full GPU of the PS4 as well. In that case the whole performance of the PS4 is not the 512 GFLOPs that I wrote above but 512 + 1433 GFLOPs. This is a total of 1945 GFLOPs, which is usually quoted as "2 TFLOPs".

I don't want to use the full power of the GPU on an i7, I want to use the full power of a GPU that a gamer will likely use with the i7.. which isn't going to be the i7's integrated GPU. With PCs you have that option, with PS4 you have what's there. It makes it awfully hard to try to do a useful comparison.

But I'd rather we just stuck to a CPU comparison like the original one you were making before you were aware that the numbers you used were incorrect.

One single jaguar core in kabini gives slightly more performance per cycle than a SB core in an i3.

Unixbench single-core index
Jaguar (kabini) @ 2.0 GHz: 403.8
SB (i3-2100T) @ 2.5 GHz: 495.5

Wow, now you're saying that Jaguar has better single-threaded perf/MHz than Sandy Bridge?

Three purely single-threaded tests (Cinebench ST, 7-zip ST, and Kraken) were done here:

http://www.anandtech.com/show/6974/amd-kabini-review/3

Normalized for clock speed, the IB i5's relative IPC advantage is:

2.42, 1.88, 2.77

Going from an i5 IB to an i3 SB isn't going to change the single threaded performance by that much. Typically going from SB to IB only improves things 5-10%.

IMO the CPU benchmarking done in this review is quite poor, particularly by AT standards (hope Anand or Brian Klug does something..) but still a lot better than cherry picking a number from a website somewhere without even citing it. That number could be faulty or unrepresentative for any number of reasons.

Any very basic common sense understanding of the uarchs (SB vs Jaguar) would make it easy to conclude that the SB would have better, usually substantially better average IPC in most real world programs. Accepting a number that shows the opposite is resisting any understanding of how CPUs work.

 

[Riek]

Sorry, I don't understand what that has to do with my response..



That was a vague rumor that has basically been completely debunked by Sony's reveal. It's almost certainly wrong. Doesn't seem to stop it from constantly being repeated though.



I don't want to use the full power of the GPU on an i7, I want to use the full power of a GPU that a gamer will likely use with the i7.. which isn't going to be the i7's integrated GPU. With PCs you have that option, with PS4 you have what's there. It makes it awfully hard to try to do a useful comparison.

But I'd rather we just stuck to a CPU comparison like the original one you were making before you were aware that the numbers you used were incorrect.



Wow, now you're saying that Jaguar has better single-threaded perf/MHz than Sandy Bridge?

Three purely single-threaded tests (Cinebench ST, 7-zip ST, and Kraken) were done here:

http://www.anandtech.com/show/6974/amd-kabini-review/3

Normalized for clock speed, the IB i5's relative IPC advantage is:

2.42, 1.88, 2.77

Going from an i5 IB to an i3 SB isn't going to change the single threaded performance by that much. Typically going from SB to IB only improves things 5-10%.

IMO the CPU benchmarking done in this review is quite poor, particularly by AT standards (hope Anand or Brian Klug does something..) but still a lot better than cherry picking a number from a website somewhere without even citing it. That number could be faulty or unrepresentative for any number of reasons.

Any very basic common sense understanding of the uarchs (SB vs Jaguar) would make it easy to conclude that the SB would have better, usually substantially better average IPC in most real world programs. Accepting a number that shows the opposite is resisting any understanding of how CPUs work.

Although your conclusion is correct in saying that sb has a higher ipc than jaguar... you have to take turboboost of i5 into account... which is 2.6GHz.

So for single threaded in Cinebench the difference is ~60% higher performance for IVB/clock. not > 2 times

 

[Exophase]

Although your conclusion is correct in saying that sb has a higher ipc than jaguar... you have to take turboboost of i5 into account... which is 2.6GHz.

So for single threaded in Cinebench the difference is ~60% higher performance for IVB/clock. not > 2 times

Thanks for the correction. I thought the numbers looked abnormally high >_> I wouldn't have needed it if the review actually said this instead of repeatedly referring to it as 1.7GHz :| I thought maybe
turbo was turned off or something.

So the proper normalized numbers should be:

1.582, 1.23, 1.81

The benchmarking part of this review really needs a ton of work. A table of hardware configurations for the things you're comparing against, some mention of things like impact of turbo on the results, and way more actual benchmarks.

 

[Riek]

Thanks for the correction. I thought the numbers looked abnormally high >_> I wouldn't have needed it if the review actually said this instead of repeatedly referring to it as 1.7GHz :| I thought maybe
turbo was turned off or something.

So the proper normalized numbers should be:

1.582, 1.23, 1.81

The benchmarking part of this review really needs a ton of work. A table of hardware configurations for the things you're comparing against, some mention of things like impact of turbo on the results, and way more actual benchmarks.

Indeed. When you see any cpu number indication it is best to google it an see the max boost. (you still don't know the behavior during benchmarking or how it behaves over multiple cores in that situation)
Would be great if they did something similar as they did when benchmarking gpu with boost functionality.. e.g. list the average clockspeed during the benchmark.

 

[Exophase]

I'd love to see logs for utilization per CPU core and clock speed vs power consumption for every benchmark. Or averages if logs are too much. They could be hidden under reveal buttons if they seem too cluttering. I hope a site takes initiatives towards making this part of their standard procedure for all tests. Would be even better if they could throw in a profile report that gives you an idea of cache misses, branch mispredictions, and average IPC. While that sounds like digging deep the tools are available that would make this not a lot of additional effort if it became standard/habitual.

When I read reviews I'd really like to get a thorough understanding of what the hardware's relative strengths and weaknesses are, and not just where it wins and where it loses.

 

[galego]

You are using GFLOPS because you think it supports your position. We aren't using it because we know the difference between theoretical numbers and real life performance.

If you have real-world benchmarks of the PS4 8 jaguar cores please share them with us. If instead you have synthetic benchmarks please share them with us. Whereas this is going to happen I will continue discussing available theoretical stuff such as GFLOPs.

That was a vague rumor that has basically been completely debunked by Sony's reveal. It's almost certainly wrong. Doesn't seem to stop it from constantly being repeated though.

I don't want to use the full power of the GPU on an i7, I want to use the full power of a GPU that a gamer will likely use with the i7.. which isn't going to be the i7's integrated GPU. With PCs you have that option, with PS4 you have what's there. It makes it awfully hard to try to do a useful comparison.

But I'd rather we just stuck to a CPU comparison like the original one you were making before you were aware that the numbers you used were incorrect.


Wow, now you're saying that Jaguar has better single-threaded perf/MHz than Sandy Bridge?

Three purely single-threaded tests (Cinebench ST, 7-zip ST, and Kraken) were done here:

http://www.anandtech.com/show/6974/amd-kabini-review/3

Normalized for clock speed, the IB i5's relative IPC advantage is:

2.42, 1.88, 2.77

Going from an i5 IB to an i3 SB isn't going to change the single threaded performance by that much. Typically going from SB to IB only improves things 5-10%.

IMO the CPU benchmarking done in this review is quite poor, particularly by AT standards (hope Anand or Brian Klug does something..) but still a lot better than cherry picking a number from a website somewhere without even citing it. That number could be faulty or unrepresentative for any number of reasons.

Any very basic common sense understanding of the uarchs (SB vs Jaguar) would make it easy to conclude that the SB would have better, usually substantially better average IPC in most real world programs. Accepting a number that shows the opposite is resisting any understanding of how CPUs work.

When I added the 4 CUs for compute, several people here suggested me to add the i7 GPU to the i7. I did, but then I also added the rendering 14 CUs to the PS4. Now suddenly this changes to ignore the i7 GPU and adds the dGPU. Wow! Fine, but then I could add HSA and hUMA and second bus and volatile bit... to the PS4 chip.

I have no problem with the original numbers. It just happens that whereas in the CPU+dGPU that you are trying now there is a clear distinction between CPU and dGPU, this distinction is blurred in the PS4. We have a 8+4+14 'core' design. If all the CUs are devoted to rendering then we have something like CPU=8 GPU=18, but if 4 are used as coprocessors then we have something like CPU=8+4 GPU=14. The CPU performance varies.

Look to the single-core benchmark given. 201.9 vs 198.2

Cinebench? Kraken on Chrome?... No thanks.

 

[Phynaz]

Cinebench? Kraken on Chrome?... No thanks.

Of course not, they don't fit your argument.

I'm really trying to understand what you are trying to accomplish by posting here. You must be aware that your posts have been completely discredited.


Oh, BTW, what systems are you running?

 

[galego]

Of course not, they don't fit your argument.

It is more related to if those benchmarks represent real world performance or not. E.g. CB uses ICC.

But you can believe what you want.

 

[Enigmoid]

If you have real-world benchmarks of the PS4 8 jaguar cores please share them with us. If instead you have synthetic benchmarks please share them with us. Whereas this is going to happen I will continue discussing available theoretical stuff such as GFLOPs.

You realize that
synthetic benchmarks >>>>>>>>>>>> theoretical GFLOPS
right?

You can get a very good estimation of the ps4 performance using kabini. Same jaguar core, same number of GFLOPS normalized to frequency and core count allows accurate estimation.

 

[Riek]

You realize that
synthetic benchmarks >>>>>>>>>>>> theoretical GFLOPS
right?

You can get a very good estimation of the ps4 performance using kabini. Same jaguar core, same number of GFLOPS normalized to frequency and core count allows accurate estimation.

different bandwidth, different software. So no you cannot estimate what Ps4 can run based on Kabini.

 

[Idontcare]

It is more related to if those benchmarks represent real world performance or not. E.g. CB uses ICC.

And I'm confident that people who use CB are glad to know it has been compiled with the best compiler available. Who wants to buy software only to be told it has been compiled with a sub-par compiler?

 

[Abwx]

And I'm confident that people who use CB are glad to know it has been compiled with the best compiler available. Who wants to buy software only to be told it has been compiled with a sub-par compiler?

Well , as a sarcasm it would be adequate...

In those scenarios where Intel has a distinct architectural or optimization advantage, the Phenom II doesn't stand a chance. That is unfortunately one downside AMD has to deal with. Intel's developer relations team is much larger and works much harder to ensure that more applications and workloads are optimized as best as possible for Intel CPUs

http://www.anandtech.com/show/2754/12

 

[Enigmoid]

different bandwidth, different software. So no you cannot estimate what Ps4 can run based on Kabini.

Compared to using FLOPS its much better.

Bandwidth really only affects the igp (the much more powerful 3770k [about 5-6x more powerful] has no problem running 1066 dual channel ram (outside winrar)).

Software is a different story.

Of course the fact that both console and pc are x86 means they can be compared fairly well (outside specific optimizations for the console that will probably take years to be implemented).

 

[galego]

You realize that
synthetic benchmarks >>>>>>>>>>>> theoretical GFLOPS
right?

I suppose you mean that synthetic benchmarks are a better tool to measure performance than theoretical GFLOPS. If the benchmark is accurate then I agree. There are misleading benchmarks that cheat the results, their names are well-known. Some are listed here:

http://sharikou.blogspot.com.es/2009/12/ftc-accuses-intel-of-rigging-benchmarks.html

Now, I prefer to evaluate the performance of the jaguar cores using theoretical values and benchmarks as the one that cited in a previous post whereas I wait for more benchmarks.

 

[Phynaz]

Lol @ Sharikou. I totally forgot about that quack. Remember his 15 minutes of fame when he said Intel was going to go bankrupt?

Now I know you're just trolling us all galego. Hey, maybe you are Sharikou!

 

[galego]

Lol @ Sharikou. I totally forgot about that quack. Remember his 15 minutes of fame when he said Intel was going to go bankrupt?

Now I know you're just trolling us all galego. Hey, maybe you are Sharikou!

The important part are the quotes from the FTC report, not what the author of the blog says or said or not said.

Specially relevant for evaluating correctly the performance of jaguar cores are the points 61 and 64 of the FTC report.

 

[Idontcare]

Well , as a sarcasm it would be adequate...

http://www.anandtech.com/show/2754/12

I don't understand your reply.

If I am using software application ABC then I really don't care what specific reasoning can be chalked up to making it perform fast or slow unless I have some control over the reason. I care about what makes it perform fast - should I buy Intel or AMD, etc - but I don't care about the stuff I have no control over (who compiled it, what compiler settings they used, and who makes the compiler).

If Intel brings its resources to bare such that Cinebench and MS Excel users will benefit by powering their software applications with an Intel CPU then more power to the Cinebench and MS Excel users.

Likewise for AMD, I really do not care how or why my application of interest (Gaussian98 and Gaussian03) performs so well with an AMD based Piledriver CPU, but it does and so that is what I use it for. (as do others)

^ performance? Yes please!

 

[Lepton87]

All you have to do is compare Jaguars vs Silvermonts reviews, counting total words and examining phrasing and tone to really determine Anands favorite CPU provider, he just doesnt care enough about AMD. :biggrin:

Agreed, it's painfully obvious, silvermont didn't even had a review and he praised it like it was the best low-power CPU ever that no one can match. He does not even try to hide his preference.

 

[galego]

If I am using software application ABC then I really don't care what specific reasoning can be chalked up to making it perform fast or slow unless I have some control over the reason. I care about what makes it perform fast - should I buy Intel or AMD, etc - but I don't care about the stuff I have no control over (who compiled it, what compiler settings they used, and who makes the compiler).

If Intel brings its resources to bare such that Cinebench and MS Excel users will benefit by powering their software applications with an Intel CPU then more power to the Cinebench and MS Excel users.

Except this is/was not the point. Nobody is worried if optimizations for a specific processors or family help to obtain better performance. I repeat, this is/was not the point. This is an extract from the FTC report (link give above):

59. To the public, OEMs, ISVs, and benchmarking organizations, the slower performance of non-Intel CPUs on Intel-compiled software applications appeared to be caused by the non-Intel CPUs rather than the Intel software. Intel failed to disclose the effects of the changes it made to its software in or about 2003 and later to its customers or the public. Intel also disseminated false or misleading documentation about its compiler and libraries. Intel represented to ISVs, OEMs, benchmarking organizations, and the public that programs inherently performed better on Intel CPUs than on competing CPUs. In truth and in fact, many differences were due largely or entirely to the Intel software. Intel’s misleading or false statements and omissions about the performance of its software were material to ISVs, OEMs, benchmarking organizations, and the public in their purchase or use of CPUs. Therefore, Intel’s representations that programs inherently performed better on Intel CPUs than on competing CPUs were, and are, false or misleading. Intel’s failure to disclose that the differences were due largely to the Intel software, in light of the representations made, was, and is, a deceptive practice. Moreover, those misrepresentations and omissions were likely to harm the reputation of other x86 CPUs companies, and harmed competition.

[...]

67. Intel publicized the results of the benchmarking to promote sales of products containing its x86 CPUs even though it knew the benchmarks were misleading.

If you want evaluate the real performance of jaguar chips, you cannot use benchmarks that do not really measure that performance.

Open source benchmarks show that one jaguar core is fastest than one SB core @ same GHz. The difference between both increases in multi-threaded benchmark

MS Excel is not compiled with ICC. Gaussian is not compiled with ICC. In fact, almost no real software is compiled with it. ICC is mostly used in several useless synthetic benchmarks.

 

[NTMBK]

In fact, almost no real software is compiled with it. ICC is mostly used in several useless synthetic benchmarks.

You really think so?

 

[Phynaz]

Galego, if Jaguar is so great (faster than SB), why is AMD positioning it as their low-end CPU? You realize it is supposed to to compete against Atom, right?

 

[BallaTheFeared]

Looks great!

 

[MightyMalus]

The next gen consoles are so weak that we need more powerful and expensive hardware to be able to play them!

The reason is the same reason that iOS>Android development. Console>PC development.
Always will be like this.
Even Flash>HTML5.

Yes, Android, PC and HTML5 are awesome and cool and whatever. But iOS, Console and Flash will always be "better". (Unless Adobe keep being jerks...)

Oddly enough, that's what Microsoft wants to do but "people" don't like that. Lmao!

Which is also what Steam wants to do! And people want it! rofl...nothing makes sense, its the same reason!

Closed, manageable systems/targets will win. Always. They offer the better experience. (iOS, Flash, Consoles. All primerally closed in one way or another, yet, ubiquitious!)

That's why most "real" PC games now are on VM's... .Net, Mono/Unity, Flash, Java, etc. And published directly to an App Store of sort, which are mostly closed.

("Real" in the context of games developed for the PC first.)

The more you treat the PC like a Console the more you "Win". That's what Steam has been doing on PC. Which is what MS wants to do on Windows!

PC Gamers usually have a hard time believing this.

Point is, closed systems have always won and will always win. Because, world peace will never be achieved unless world peace is compromised.

 

[galego]

You really think so?

Yes, and I am not alone.

Galego, if Jaguar is so great (faster than SB), why is AMD positioning it as their low-end CPU? You realize it is supposed to to compete against Atom, right?

I don't need to evaluate AMD portfolio, marketing, plans... neither anything as that.

I am evaluating the Jaguar benchmarks that I know, whereas waiting for more.

 

[NTMBK]

Yes, and I am not alone.

You may not be alone, but you are certainly incorrect.