AMD Shanghai/Deneb Review Thread

[keithlm]

this is what i have in mind:
1. load 2 or more games on intel/amd rig.
2. run fraps or equivalent utility.
3. alt-tab on different games.
4. observe drop in fps if theres any.
5. vary timing on alt-tabbing.
6. upload the result.

Even better: If someone has machines with various brands of CPU they can make a type of benchmark that CAN be measured and repeated. If they ADD other benchmark(s) at the SAME TIME then they can create some interesting Xtreme conditions. They could try to find something that slows down one machine but not the other AND is repeatable.

For example: What happens on various CPU if you are running Prime95 WHILE playing a game? Does it make one system bog down and the other doesn't even notice? Does the Prime95 go slower?

If you run 4 threads of Prime95 (perhaps 8?), a game benchmark, and everything works without problems... then add something else like a virus scanner. You can keep adding different things that use up various resources one at a time until either one or both machines slow down. Then back off a bit and see if you can determine which machine had problems first. It this process is repeatable then you have found a method to determine "smoothness".

Of course if one brand works better than the other when this process is done... you can expect some people to claim: "I never run all of that stuff at the same time so it doesn't matter to me." However if this kind of thing can be measured AND proven then it is definitely something to consider.

The problem is that this type of "Xtreme" benchmark is not easy to find. In fact various people have already tried the 4xPrime95 + Game test. We didn't really see any differences between brands. So the hypothetical "slowdown" would take a lot of work to actually find and will require more work.

EDIT: BTW: I do actually expect this process to happen... but it very probable that it will be done by somebody testing between the new Intel i7 and and older Intel chip of comparable speed.

[Hornet331]

hmm, i dont really have extensive knowledge about cpus, so i really dont know if that would be true. if it is true then without even going through the tests, we have quantified/qualified that "smoothness" on phenoms is no longer relative/subjective but can be objectively attributed with phenoms having better memory sub system design compared to intel cpu's -except i7's i supposed.
you agree on this?
second, i assume i7's having "smoother feel" also compared to penryn's. any i7 owners can attest to this? esp those who upgrade from penryn's?
edit:btw, as i said im no expert and in no way im trying to sound one. just trying to offer my 2 cents as everytime i read someone mention smoothness, everybody freaks out. and its not just this forum, ive read it across multiple review sites. so...

My CPU history beginning from P4 and skipping everything before 2000
P4->X2->C2D/(A64 for HTPC)->C2D 45nm quad->C2D 45nm->Ci7

Ever since I switched from a P4 to a X2 there never was a feeling that is was more smoother even from dualcore C2D to quad core Ci7.

The only thing i noticed and made a habbit of mine, is that if os older then 6month it tends to get boged down by software installs and all the garbage that accumulates. So i reinstall my my os every 6-8 months (coincidently the last 2 years i buy new hardware every 6-8 months ).

Even better: If someone has machines with various brands of CPU they can make a type of benchmark that CAN be measured and repeated. If they ADD other benchmark(s) at the SAME TIME then they can create some interesting Xtreme conditions. They could try to find something that slows down one machine but not the other AND is repeatable.

For example: What happens on various CPU if you are running Prime95 WHILE playing a game? Does it make one system bog down and the other doesn't even notice? Does the Prime95 go slower?

If you run 4 threads of Prime95 (perhaps 8?), a game benchmark, and everything works without problems... then add something else like a virus scanner. You can keep adding different things that use up various resources one at a time until either one or both machines slow down. Then back off a bit and see if you can determine which machine had problems first. It this process is repeatable then you have found a method to determine "smoothness".

Of course if one brand works better than the other when this process is done... you can expect some people to claim: "I never run all of that stuff at the same time so it doesn't matter to me." However if this kind of thing can be measured AND proven then it is definitely something to consider.

The problem is that this type of "Xtreme" benchmark is not easy to find. In fact various people have already tried the 4xPrime95 + Game test. We didn't really see any differences between brands. So the hypothetical "slowdown" would take a lot of work to actually find and will require more work.

EDIT: BTW: I do actually expect this process to happen... but it very probable that it will be done by somebody testing between the new Intel i7 and and older Intel chip of comparable speed.

I dont see how this will prove anything, if you run max threads of prime on a quad core and then start a game every system will output less fps, it doesn't matter if its intel or amd or any other cpu.

As a matter of fact, i just benched a similar thing on my Ci7.

Running Boinc with 8 threads and benching crysis costs me ~2fps on avarage and 4fps on the min fps as compared to when i only run crysis alone.

But a note: boinc runs as low priority process so if i would run prime which has normal priority it will hurt the game even more.

By your defenition Ci7 would be smother then K10 be default, cause it can handle more threads better due to HT.

This whole "Smoother Myth" (yes i call it a myth cause i never expired it even though i had my share of expirience a lot of processors) is in my eyes more related to other things then just the cpu.

Imho the most important factor is the HDD, i have played around with some SDD and my old Notebook (Singelcore P-M 1,6GHz 1GB RAM) that had an old slow ass 4200rpm drive. Holy moly i never thought how responsive this old thing could get just by upgrading the HDD.

Another thing is RAM, you never can have enough ram, when the system starts swapping it slows down.

[roofsniper]

lets stay on topic if you want to find a way to test for how smooth something is then make another thread.

[sundancerx]

lets stay on topic if you want to find a way to test for how smooth something is then make another thread.

ok, sorry for the off topic. i wasnt trying to derail the thread or try to flamebait anyone. i just grew tired reading about amd system being smoother/or more responsive. i just hope someone or some review site can put and end to this one. again im no expert, so i just offered my 2cents to quantify those "feel". i guess the burden of proof is on the amd side. sorry again.

[iandh]

ok, sorry for the off topic. i wasnt trying to derail the thread or try to flamebait anyone. i just grew tired reading about amd system being smoother/or more responsive. i just hope someone or some review site can put and end to this one. again im no expert, so i just offered my 2cents to quantify those "feel". i guess the burden of proof is on the amd side. sorry again.

I consider it very on topic as long as we continue to discuss how the Phenom/Phenom II system architecture may/may not effect gaming, and keep it civil and intelligent as we have so far.

I think I have explained quite well with what we need to aim for as far as testing, and unfortunately I don't know that this is something that we could easily measure with any benchmark on either side of the fence. We would probably need AMD/intel confidential debugging software and equipment to see how data routing is happening at the platform level, and I seriously doubt any of us will ever be given access to that type of thing to see how video games run, sadly.

Besides speculation, first hand experience, and logical reasoning, we are pretty much left to just trusting each other. I am an intelligent, truthful, and highly skilled individual. I work at an electronics manufacturer that interacts not only with the semiconductor industry, but with the military sector (ONR, ARL), and respected research institutions such as JHU. My current project partner is the group supervisor for the Space and Astrophysical Plasmas group at NASA/JPL and has a Ph.D. in Space Plasma Physics. We just had one of AMD's fellows visit last month to review some equipment they are purchasing from us.

I consider myself to be highly intelligent (never tested less than 99th percentile), and work on a daily basis with people that make me look like a blithering idiot moron. There are many times in research projects that we see behaviors which at first are not easily explained by the math or the data. That does not mean that those behaviors are any less real, or that we are somehow failing at our tasks.

A closed mind, or thinking you already know everything there is to know, is never a good place to be. There is not a single person on this site that knows everything there is to know about computer platforms.

[roofsniper]

I consider it very on topic as long as we continue to discuss how the Phenom/Phenom II system architecture may/may not effect gaming, and keep it civil and intelligent as we have so far.

yes it is but without any facts or anything that can prove anything it will go no where. i don't know how you can prove it but from what i have heard it makes sense. i said about a year ago about this whole phenom smoother thing that it had to do with the fast hyper transport speeds and the fact that the cpu has such low latency and it can just whip things out lightning speed. when you have things like super pi or something that sits in the cpu cache and doesn't really exit the cpu that often intel will get better performance. but with the way amd has designed their platform and made the connection speeds so fast it just seems obvious that it would be smoother. in games the gpu is making the frames the cpu just needs to process them. if they are being sent in at different rates because of the low I/O latency then it will not be as smooth even if you have more frames. if someone can find a way to prove this then so be it. possibly you could get an amd and intel system tuned up perfectly and take fps measurements every second and put them together. from what i would think the intel machine would have a more jagged graph/results while the amd one should have smoother results. if someone can find a way to prove this fine but it seems like all it is going to be is pages of people arguing back and forth because there is no data to present.

[iandh]

yes it is but without any facts or anything that can prove anything it will go no where. i don't know how you can prove it but from what i have heard it makes sense. i said about a year ago about this whole phenom smoother thing that it had to do with the fast hyper transport speeds and the fact that the cpu has such low latency and it can just whip things out lightning speed. when you have things like super pi or something that sits in the cpu cache and doesn't really exit the cpu that often intel will get better performance. but with the way amd has designed their platform and made the connection speeds so fast it just seems obvious that it would be smoother. in games the gpu is making the frames the cpu just needs to process them. if they are being sent in at different rates because of the low I/O latency then it will not be as smooth even if you have more frames. if someone can find a way to prove this then so be it. possibly you could get an amd and intel system tuned up perfectly and take fps measurements every second and put them together. from what i would think the intel machine would have a more jagged graph/results while the amd one should have smoother results. if someone can find a way to prove this fine but it seems like all it is going to be is pages of people arguing back and forth because there is no data to present.

I suppose we could record FPS with rivatuner and expand the graphs out, but it would be rather hard to show it at a decent resolution becuase this is something that happens momentarily over long periods of time, so to be able to view enough sub-second detail to accurately compare the two we would need an expanded graph the length of a schoolbus.


As I said before my hunch is as well that it has to do with platform architecture, but to actually view these things accurately on either side we would need access to equipment that we just can't have access to without signing an NDA, and even then we probably couldn't talk about the results and testing methods afterwards.

[sundancerx]

I consider it very on topic as long as we continue to discuss how the Phenom/Phenom II system architecture may/may not effect gaming, and keep it civil and intelligent as we have so far.

I think I have explained quite well with what we need to aim for as far as testing, and unfortunately I don't know that this is something that we could easily measure with any benchmark on either side of the fence. We would probably need AMD/intel confidential debugging software and equipment to see how data routing is happening at the platform level, and I seriously doubt any of us will ever be given access to that type of thing to see how video games run, sadly.

Besides speculation, first hand experience, and logical reasoning, we are pretty much left to just trusting each other. I am an intelligent, truthful, and highly skilled individual. I work at an electronics manufacturer that interacts not only with the semiconductor industry, but with the military sector (ONR, ARL), and respected research institutions such as JHU. My current project partner is the group supervisor for the Space and Astrophysical Plasmas group at NASA/JPL and has a Ph.D. in Space Plasma Physics. We just had one of AMD's fellows visit last month to review some equipment they are purchasing from us.

I consider myself to be highly intelligent (never tested less than 99th percentile), and work on a daily basis with people that make me look like a blithering idiot moron. There are many times in research projects that we see behaviors which at first are not easily explained by the math or the data. That does not mean that those behaviors are any less real, or that we are somehow failing at our tasks.

A closed mind, or thinking you already know everything there is to know, is never a good place to be. There is not a single person on this site that knows everything there is to know about computer platforms.

you made very good logical points and if i have half of your knowledge and experience, i would be very happy. i have bachelor's degree in electrical engg and little background on electronics too, but i worked on a manufacturing company for almost a decade now. i feel like my technical background is buried and i need to dig way deep just to be able to have little conversation with you guys.

[knopflerbruce]

ok, sorry for the off topic. i wasnt trying to derail the thread or try to flamebait anyone. i just grew tired reading about amd system being smoother/or more responsive. i just hope someone or some review site can put and end to this one. again im no expert, so i just offered my 2cents to quantify those "feel". i guess the burden of proof is on the amd side. sorry again.

I'm switching from 940 to am2 to 775 from time to time, and the most "smooth" setup of those 3 is the am2 setup. Not too much difference, though, and I have a hard time proving anything, but that's just my .02.

What burden of prrof are you talking about, btw? That AMD are smoother than Intel? If the opposite is true, I'd like to see some proof there as well.

[sundancerx]

I'm switching from 940 to am2 to 775 from time to time, and the most "smooth" setup of those 3 is the am2 setup. Not too much difference, though, and I have a hard time proving anything, but that's just my .02.

What burden of prrof are you talking about, btw? That AMD are smoother than Intel? If the opposite is true, I'd like to see some proof there as well.

what im trying to say is that, since these feelings of "smoothness" on amd rigs dont show up on charts and benchmarks (and is very hard to quantify/qualify based on a few comments i read here), pro intel people will just easily dismiss this as a myth.
now if you are pro amd and you really really want this smoothness feel taken as one of the advantages of using amd system, then i think its expected that you are the ones to initiate actions, expreiment on different testing methods, find other benchmarking methods that may quantify this so called smoothness thingy. thsis is what i meant by burden of proof relies on amd. and until then, no amount explanation no matter how logical they maybe will convince hardcore fanboys on the opposite side, that indeed, its not just a myth.
sorry, im no english expert, but i hope you got my point.

[QuietIce]

what im trying to say is that, since these feelings of "smoothness" on amd rigs dont show up on charts and benchmarks (and is very hard to quantify/qualify based on a few comments i read here), pro intel people will just easily dismiss this as a myth.
now if you are pro amd and you really really want this smoothness feel taken as one of the advantages of using amd system, then i think its expected that you are the ones to initiate actions, expreiment on different testing methods, find other benchmarking methods that may quantify this so called smoothness thingy. thsis is what i meant by burden of proof relies on amd. and until then, no amount explanation no matter how logical they maybe will convince hardcore fanboys on the opposite side, that indeed, its not just a myth.
sorry, im no english expert, but i hope you got my point.

While I understand your basic point here I can see what would happen on the other side of the fence. Us AMD guys actually find a way to measure this "smoothness" and show it as a "bench". Meanwhile, the Intel boys go back to their clubhouse and start talking about the AMD-biased program we're using and how it's not a true benchmark and yadda, yadda, yadda!

Hardcore guys simply won't believe AMD is better (P4 sales of 2004-2006 is enough proof of that!) and the fence riders are going to buy whatever PR is best at the time of purchase. I doubt if anything will ever change that ...

[iandh]

when amd users start talking about "responsiveness/smoothness" of their amd gaming rig compared to intel ones, people respond by saying amd users start hallucinating/spewing bs to justify their rig having a better gaming performance that doesnt necessarily show on benchmarks/graphs.
the best way to do this to try to qualify/quantify these "feel" thingy which is very subjective.
since you have both intel and amd rigs, you are in the best opportunity to perform this. im not suggesting/asking you do this, just saying maybe if you want to make this "objective", you can actually do this.
this is what i have in mind:
1. load 2 or more games on intel/amd rig.
2. run fraps or equivalent utility.
3. alt-tab on different games.
4. observe drop in fps if theres any.
5. vary timing on alt-tabbing.
6. upload the result.
then people can see it for themselves which one is 'smoother/more responsive' machine.
just make sure you label the correct video matches the right rig lol.
just my 2 cents. btw, im pro amd .

The problem is that we can't record anything useful by recording average framerate. Does fraps capture jumps between loading map areas or view distance spawns in games like oblivion and fallout 3?

Does is record the jumps in framerate in the large beach are in Crysis once the fireworks start flying?

That is where the difficulty is, and why it is so hard for some people to grasp. People on this site are SO FASCINATED with graphs that they can't see anything else.

I work at a research institution and we encounter things on a daily basis that are not represented well in our test numbers and graphs.

your "test" is nothing more then a artifical stress test for the memory sub system of a cpu.

Guess who will win this.

For my personal experience there was no more "smoothness increasment" after my first A64 X2, all other CPU i ever had since then never felt more smooth or slower or what else then the X2.

I suppose in one way this is going off topic, but in another we are indirectly discussing how the architecture used in the Phenom II will behave in gaming.

I personally have never had an X2 based rig that ran completely smooth in gaming, regardless of title or framerate. Core 2 was marginally better, and C2Q was better still in some games during action intensive scenes, although it didn't provide any framerate increase.

Another thing that really cracks me up in the whole dual vs. quad argument (further off topic), is that people think that just because a quad isn't acheiving higher framerates means it isn't running any better...

My CPU history beginning from P4 and skipping everything before 2000
P4->X2->C2D/(A64 for HTPC)->C2D 45nm quad->C2D 45nm->Ci7

Ever since I switched from a P4 to a X2 there never was a feeling that is was more smoother even from dualcore C2D to quad core Ci7.

The only thing i noticed and made a habbit of mine, is that if os older then 6month it tends to get boged down by software installs and all the garbage that accumulates. So i reinstall my my os every 6-8 months (coincidently the last 2 years i buy new hardware every 6-8 months ).




I dont see how this will prove anything, if you run max threads of prime on a quad core and then start a game every system will output less fps, it doesn't matter if its intel or amd or any other cpu.

As a matter of fact, i just benched a similar thing on my Ci7.

Running Boinc with 8 threads and benching crysis costs me ~2fps on avarage and 4fps on the min fps as compared to when i only run crysis alone.

But a note: boinc runs as low priority process so if i would run prime which has normal priority it will hurt the game even more.

By your defenition Ci7 would be smother then K10 be default, cause it can handle more threads better due to HT.

This whole "Smoother Myth" (yes i call it a myth cause i never expired it even though i had my share of expirience a lot of processors) is in my eyes more related to other things then just the cpu.

Imho the most important factor is the HDD, i have played around with some SDD and my old Notebook (Singelcore P-M 1,6GHz 1GB RAM) that had an old slow ass 4200rpm drive. Holy moly i never thought how responsive this old thing could get just by upgrading the HDD.

Another thing is RAM, you never can have enough ram, when the system starts swapping it slows down.

Just because you haven't personally experienced something doesn't mean it is a myth. I could almost say it is borderline flamebait being that we are so closely approaching an actual logical conversation on this topic.

I will aggree that HDD does have a lot to do with load times, and additionally running your games on a second hard drive from your OS can help quite a bit with stuttering and stability issues, but IME it does not eliminate them completely.

When I use the same HDD and OS install on different mobos and CPU's, and use different ram on each, and still experience different behavior between them, it tells me that there is something else going on that I don't know about yet.






I work at a company that manufactures semiconductor equipment, my brother in law is a process engineer at intel, and I interact with some very sharp folk on an almost daily basis. I can say wholeheartedly that there is a lot more going on in a computer platform than just plain old number crunching. Each platform has its own personality and own strengths and weaknesses due to architectural differences.

I'm sure this whole "smoothness" thing would be a lot easier to sell to an engineer at a server OEM- they work on a little bit different platfom integration level than just slapping some crap together and shooting zombies



edit: Here's a random quote I found while reading the internet; this further reinforces my comment regarding engineers at server OEM's (I wrote that before I found this):

"Here is one explanation why AMD performs better when you run more advanced games and also run them on high res.
It is the same reason as why AMD is strong on servers even if Intel Xeon is higher clocked.
It’s about communication with other hardware (memory and GPU).

When Intel communicates with other hardware it sends data using the FSB. All traffic goes through the FSB. The Latency using the FSB is about 250 clocks. The performance penalty is rather large. If one game is reading memory and/or sends I/O to the GPU using only one thread. Then I don’t think you will see any big differences between AMD and Intel. If the video card is slow you defiantly isn't going see any differences.
Now if the game is using more than one thread the situation changes. If one thread is sending or reading data on Intel, and another thread is sending data to the video card. One of the threads needs to wait. That means that latency goes up. In worst case scenarios it would be double (about 500 clocks).
On AMD this is handled differently. AMD has hypertransport that handles I/O to the video card. If one thread is sending data to the video card it doesn’t compete with memory. On AMD latency is a bit lower too.
Running games on low res or games that isn’t that advanced then this isn’t a problem on Intel. But if you are running a game on high res, have a fast video card (or two). Then this video card is able to handle VERY MUCH I/O. Also if the game is using more memory then more data needs to travel through the FSB on Intel. This situation will lead to more conflicts in the FSB and latency is increased.
Intel is very fast when it isn’t disturbed and is able to use data the cache. More threads, more memory, more communication with video card and maybe more synchronization between threads. Then performance will shift to favor AMD system design."

Regardless of whether you believe it has an effect on gaming or not, intel and AMD platforms running specific applications differently is a FACT, not a MYTH.

[Hornet331]

http://www.xtremesystems.org/forums/...9&postcount=42

riped out to keep this thread clean.