Next Gen VS. Current Gen CPU: Crysis Benchmark

kindergarden seems busy these days
time to close it (again)

well, some unbelievable scores from these new cpus, but i'd like to ask and i wouldnt let this pass

Are my scores any good for a quad? i am running 9 x 325 (3.13 ghz) and stock Ultra(if i flash my Ultra with the Ultra KO settings, it wont be stable :rolleyes: ) see rig bellow for reference:

1280x960, "very high settings trick" and xp64
http://img61.imageshack.us/img61/389...enchku2.th.jpg

1024x760, high settings and xp64
http://img124.imageshack.us/img124/3...1024qz0.th.jpg

I'd like some tips on this, if its a poor score or not. I am running this bench on xp 64 and it runs soft with modded settings to emulate dx10(very high trick) and 1280x960, average of 25, 30 fps as in the bench, so its pretty playabe.

I reckon if i change cpu, how much would it increase. :eek:

i am sick and tired of this whole wait for the 20th and then AMD is going to bring out something that is going to crush intel and be way better, blah blah blah.

WE HAVE K10 NUMBERS!!!!! IT'S CALLED BARCELONA AND IT"S BEEN OUT FOR MORE THAN 1 MONTH AND IT IS SLOWER CLOCK PER CLOCK TO INTEL!!!!!!!

have you people not seen? Phenom and Barcelona are the EXACT same and guess what Barcelona is slower than Core 2 Quad. so how is it that you people are saying Phenom is faster than Core 2 Quad?????? It's IMPOSSIBLE!!!!!! get out of your dream AMD world and face facts K10 is the new R600.

i have not seen any benches that show K10 faster than Core 2 Quad in the real world and yet many that show that Core 2 Quad is faster than K10 in the real world.

and this is a CPU bound test. it may not be the best but look at the difference in numbers. the Intels are very close to each other (which was expected) and phenom is a good lot slower than the Intels. exact same bench exact same detail and resolution settings. AMD is SLOWER what more do you need?

and this is a reputable site. it brought us the first 8800gt numbers and guess what they were right on. and now they are just confirming what we know already k10 slower than core 2 quad.

Quote:

---

Originally Posted by [XC] hipno650

i am sick and tired of this whole wait for the 20th and then AMD is going to bring out something that is going to crush intel and be way better, blah blah blah.

WE HAVE K10 NUMBERS!!!!! IT'S CALLED BARCELONA AND IT"S BEEN OUT FOR MORE THAN 1 MONTH AND IT IS SLOWER CLOCK PER CLOCK TO INTEL!!!!!!!

have you people not seen? Phenom and Barcelona are the EXACT same and guess what Barcelona is slower than Core 2 Quad. so how is it that you people are saying Phenom is faster than Core 2 Quad?????? It's IMPOSSIBLE!!!!!! get out of your dream AMD world and face facts K10 is the new R600.

i have not seen any benches that show K10 faster than Core 2 Quad in the real world and yet many that show that Core 2 Quad is faster than K10 in the real world.

and this is a CPU bound test. it may not be the best but look at the difference in numbers. the Intels are very close to each other (which was expected) and phenom is a good lot slower than the Intels. exact same bench exact same detail and resolution settings. AMD is SLOWER what more do you need?

and this is a reputable site. it brought us the first 8800gt numbers and guess what they were right on. and now they are just confirming what we know already k10 slower than core 2 quad.

---

Sadly, +1 :(

http://www.expreview.com/img/news/07...nomoc_29xx.png
http://www.expreview.com/img/news/07...oc_30xx_ht.png

http://bbs.expreview.com/thread-9585-1-1.html

ok i wont say anything.

just read the pic and judge by yourself.(install a Crysis demo and run that time demo and compare.. )

Hey! You re-ran the test! The original one had way too low Min FPS.

Maybe run another benchmark coz Crysis benchmarks results are too fluctuent for me :shrug:
Something everyone trust and understand...

Looks like he increased the fsb on the Phenom from 200 to 265 and 267, increase the memory clock from 375 to around the 500 range and got a nice increase in overall performance in doing so. Also a little better memory timmings in the second screenie. Not sure which was more important, but it looks like HT link has no effect. Probably the memory as the seem to be the main difference between the two, but the Phenom is still laging in fsb which may play a factor. Very nice and thanks for sharing

can you please run 3d06 cpu test on both?
Other then that thanks for a great review.
Any luck getting that phenom to 4G on air?v;) Cause that's what it would need to compete with 45nm Intel cpu.

can you run the es phenom with 266mhz x 12.5 for 3.3ghz . nb speed at 2.1ghz memory at 485mhz 4-4-5 2t ... vcore at 1.48v.. that should max out your es phenom

hows the ram bandwidth for amd?
looking closer and closer clock for clock now.

Quote:

---

Originally Posted by PhilDoc

Looks like he increased the fsb on the Phenom from 200 to 265 and 267, increase the memory clock from 375 to around the 500 range and got a nice increase in overall performance in doing so. Also a little better memory timmings in the second screenie. Not sure which was more important, but it looks like HT link has no effect. Probably the memory as the seem to be the main difference between the two, but the Phenom is still laging in fsb which may play a factor. Very nice and thanks for sharing

---

The 2nd is also @ 3.07GHz vs 2.91GHz for the first one, I'm sure that helps a bit as well as the memory timings. :up:

Quote:

---

Originally Posted by Epsilon84

Those are pretty crap memory timings dude, AFAIK slow timings hurt AMD more than it does Intel. Why are you using such loose timings? Hell even my crappy Corsair Value Select DDR2-667 can do 1000MHz @ 4-4-4-12 2T @ 2.3V... did you even try to tighten up the timings at all?

Still, assuming there is a ~10% gain with better timings for AMD as opposed to only ~5% for Intel, it will still put X4 ~20% slower clock for clock than C2Q. Not good. :(

---

AGREED 100%

btw: I have some valueselect, and it maxes out at 5-5-5-15, DDR2-900, 2.2v...bench stable is all. I should put the stuff in my DFI :D

Quote:

---

Originally Posted by AbelJemka

Maybe run another benchmark coz Crysis benchmarks results are too fluctuent for me :shrug:
Something everyone trust and understand...

---

oh kay ok
now i am writing 3DMark06

pls wait for a sec

So I guess the phenom wont deliver higher performance then Intel's c2d.

So does this mean AMD will again be forced to fight with the price war. But who knows, the price might be just right.

Quote:

---

Originally Posted by cookerjc

oh kay ok
now i am writing 3DMark06

pls wait for a sec

---

w00t!

:banana: :banana: :up:

Quote:

---

Originally Posted by [XC] hipno650

i am sick and tired of this whole wait for the 20th and then AMD is going to bring out something that is going to crush intel and be way better, blah blah blah.

WE HAVE K10 NUMBERS!!!!! IT'S CALLED BARCELONA AND IT"S BEEN OUT FOR MORE THAN 1 MONTH AND IT IS SLOWER CLOCK PER CLOCK TO INTEL!!!!!!!

have you people not seen? Phenom and Barcelona are the EXACT same and guess what Barcelona is slower than Core 2 Quad. so how is it that you people are saying Phenom is faster than Core 2 Quad?????? It's IMPOSSIBLE!!!!!! get out of your dream AMD world and face facts K10 is the new R600.

i have not seen any benches that show K10 faster than Core 2 Quad in the real world and yet many that show that Core 2 Quad is faster than K10 in the real world.

and this is a CPU bound test. it may not be the best but look at the difference in numbers. the Intels are very close to each other (which was expected) and phenom is a good lot slower than the Intels. exact same bench exact same detail and resolution settings. AMD is SLOWER what more do you need?

and this is a reputable site. it brought us the first 8800gt numbers and guess what they were right on. and now they are just confirming what we know already k10 slower than core 2 quad.

---

First of all, Phenom != Barcelona. Other one is ran with server hardware, other one is not. It makes a difference, doesn't it?

While I agree with you, I'd like to point out that all these tests are ran with BA/B1 revisions. B2 should be better. AMD has never been better or faster than Intel, especially not in SuperPI alike (has there ever been ANY 1M world record made with AMD CPU?), which people like to use as a "speed-o-meter".

Remember how the first K8's were? SLOW, at least when compared what they were with S939, so there was quite an improvement there, and I wouldn't be surprised if they managed to make it again. At this point I have to say that I have no clue what kind of improvement has Intel made since C2D launch, so even if AMD can improve, is the improvement enough to fight Intel?

AMD can always TRY to fight with price, but it won't be so good for them, bad yeilds and cheap prices, need to say more? Though, if I understood correctly, AMD makes the chips always a full X4, and then if the yeild was good, it will be sold as X4, if one core is danaged, it will be sold as X3, if half the die is damaged (2 cores) it will be sold as X2 and so on.

Also as I said in other thread, AMD is pretty much screwed, it can't really sell alot since Intel just pays resellers not to sell AMD, which it did for a great time. All hail corruption.

Quote:

---

Originally Posted by Calmatory

First of all, Phenom != Barcelona. Other one is ran with server hardware, other one is not. It makes a difference, doesn't it?

---

What people are comparing is IPC, and for that Phenom and Barcelona are identical in architecture and memory timings will make AT BEST only a 1-2% improvement. So no, Phenom = Barcelona for these comparisons.

Quote:

---

While I agree with you, I'd like to point out that all these tests are ran with BA/B1 revisions. B2 should be better. AMD has never been better or faster than Intel, especially not in SuperPI alike (has there ever been ANY 1M world record made with AMD CPU?), which people like to use as a "speed-o-meter".

---

Actually, before C2D, SuperPi was one of THE AMD benchmarks. Netburst got creamed in Super Pi.

Quote:

---

Remember how the first K8's were? SLOW, at least when compared what they were with S939, so there was quite an improvement there, and I wouldn't be surprised if they managed to make it again. At this point I have to say that I have no clue what kind of improvement has Intel made since C2D launch, so even if AMD can improve, is the improvement enough to fight Intel?

---

K8s were never THAT slow to begin with. Over the span of 3 years, the different revisions only added MAYBE 8-10% TOTAL improvement over those 3 years (and most of that came in the one step of going to a dual memory bus on Socket 939). How many revisions since the E4/E6 cores have shown IPC improvements? None. And Brisbane was actually a small step BACKWARDS.

Quote:

---

AMD can always TRY to fight with price, but it won't be so good for them, bad yeilds and cheap prices, need to say more? Though, if I understood correctly, AMD makes the chips always a full X4, and then if the yeild was good, it will be sold as X4, if one core is danaged, it will be sold as X3, if half the die is damaged (2 cores) it will be sold as X2 and so on.

---

Only partially true. Quad and Tri-cores come from the same wafers. Dual cores are a completely different entity and are made as dual cores to start with.

Quote:

---

Also as I said in other thread, AMD is pretty much screwed, it can't really sell alot since Intel just pays resellers not to sell AMD, which it did for a great time. All hail corruption.

---

Intel is not stupid enough to be paying anyone SQUAT right now (or for the past 18 months). They are in the middle of being investigated for this in multiple countries. I'm sure they are walking the "straight and narrow" right now. After the investigation, who knows. But you can bet they are playing nice right now (well, except for the fact the C2D is kicking everythings arse).

Quote:

---

Originally Posted by mstp2009

What people are comparing is IPC, and for that Phenom and Barcelona are identical in architecture and memory timings will make AT BEST only a 1-2% improvement. So no, Phenom = Barcelona for these comparisons.

---

so you think that going from ECC DDR-667 to regular DDR2-1066 is nothing more than a 1-2% improvement?
:rofl: :rofl:

Quote:

---

Originally Posted by mstp2009

K8s were never THAT slow to begin with. Over the span of 3 years, the different revisions only added MAYBE 8-10% TOTAL improvement over those 3 years (and most of that came in the one step of going to a dual memory bus on Socket 939). How many revisions since the E4/E6 cores have shown IPC improvements? None. And Brisbane was actually a small step BACKWARDS.

---

If i remember it correctly,the first K8 was an opteron @ 1.4 GHz - 1.6GHz
he is not talking about the architecture being slow,he is talking about clockspeeds......

Quote:

---

Originally Posted by GoThr3k

so you think that going from ECC DDR-667 to regular DDR2-1066 is nothing more than a 1-2% improvement?
:rofl: :rofl:



If i remember it correctly,the first K8 was an opteron @ 1.4 GHz - 1.6GHz
he is not talking about the architecture being slow,he is talking about clockspeeds......

---

And I was talking about IPC. IPC doesn't change significantly with memory speed you n00b. :stick:

Plus, with the 2MB L3 caches we see in K10, Phenom should be LESS limited by memory speed than K8 was (with it's relatively small caches). Once you increase cache size, memory interface speed becomes FAR LESS critical in determining performance, at least on 1 and 2P systems.

No, he was talking about ARCHITECTURE, not clockspeed.

Oh, and GoThr3k, thanks for quoting my comment on 3GHz C2D Penryn chips.

Based on preliminary wattage info here, a Penryn Quad at 3GHz consumes an actual 73W at load (even though TDP is listed as 125W IIRC):

http://www.tomshardware.com/2007/10/...ng/page14.html
http://media.tomshardware.com/2007/1..._power_max.png

72.98W on the QX9650 at full load is almost 1/2 the wattage we saw on the 65nm Quads.

That would put an UNBINNED DUAL CORE at 38.5W at full load.

25W at 3GHz for a DUAL CORE based on Penryn, with proper binning, is WELL WITHIN the possibilities for Intel's 45nm HMGK process.


Now, how dumb do you look and feel? http://www.xcpus.com/forums/images/smilies/biggrin2.gif http://www.xcpus.com/forums/images/smilies/wideeyed.gif

Quote:

---

Originally Posted by mstp2009

And I was talking about IPC. IPC doesn't change significantly with memory speed you n00b. :stick:

---

if you were talking about IPC,and IPC stays the same,how can it improve 1-2% according to you? :rolleyes: :rolleyes:
so you were talking about performance,and from ECC DDR2-667 to DDR2-1066 will gain more then 1-2%
who is the dumbass after all??

Quote:

---

Originally Posted by mstp2009

25W at 3GHz for a DUAL CORE based on Penryn, with proper binning, is WELL WITHIN the possibilities for Intel's 45nm HMGK process.

---

According to you maybe....
we shall see what Q1 brings,3ghz at 25W would be a real accomplishment if you ask me,would be some nice notebook chip
i still haven't seen any proof of a 3ghz 25W part

Quote:

---

Originally Posted by GoThr3k

if you were talking about IPC,and IPC stays the same,how can it improve 1-2% according to you? :rolleyes: :rolleyes:
so you were talking about performance,and from ECC DDR2-667 to DDR2-1066 will gain more then 1-2%
who is the dumbass after all??

---

It's statistical error in the measurements genius.

You really do have3 $h!t for brains, don't you?

Quote:

---

Originally Posted by mstp2009

It's statistical error in the measurements genius.

---

what does that have to do with it?

Quote:

---

Originally Posted by GoThr3k

what does that have to do with it?

---

ANY 1-2% variance is within the level of statistical error.


Basic math.


You can't get a significant p value between two data sets that vary by only 1-2% unless the number of measurements within EACH data set is HUGE. In this case we would need about 2-500 INDIVIDUAL measurements from EACH (Phenom and Barcelona) to say with 95% certainty there is a STATISTICALLY SIGNIFICANT difference between the two (i.e. a P <= 0.05, which is universally recognized as the MINIMUM cutoff to say something is significantly different or not).