SSE4 and gaming

[Carfax]

Is it just me, or has Conroe's gaming performance has denigrated a bit from the first previews, to the release of AM2?

I'm not totally sure, but it just seems that way...particularly as some of the reviews/previews out around the web feature the 6800 Extreme which should by all means smack the living crap out of an FX-62.

Reasons as to why this may be the case? Perhaps it's because the initial previews used XFire setups and weren't graphically bottlenecked..

Or, maybe it's due to reviewers using too high a setting and thus limiting CPU involvement?

Early previews/reviews are also famous for using leaked ES samples, which typically perform less than actual revisions..

The possibilities are many, if my premise is true.

Or maybe it's due to SSE4?

Let me explain the last one. If anyone read the first Anandtech preview, they mentioned that ATI had made some special drivers for Conroe..

Lots of people were postulating about the nature of these special drivers, yet that was all we could do..

I'm now wondering whether those ATI drivers had SSE4 optimizations in them, which caused Conroe's gaming performance to be 5 > 10% higher than it normally would be?

Conroe has great SIMD capability, and both ATI's and Nvidia's graphics drivers use SSE2; and older but more widespread SIMD implementation.

Nobody knows the exact nature of the new SSE4 instructions, other than they are for multimedia purposes..

Gaming falls under the multimedia category, and perhaps Intel collaborated with ATI to produce some early drivers which took advantage of the new instructions to increase gaming performance.

Nothing wrong there ofcourse, as optimization is not cheating.

Anyone want to add their thoughts to this, go right ahead.

[Theta]

I don't see anything indicating that Conroe is losing any footing versus the FX-60/62.

Where are you basing this off of - did a new comparison emerge with lower scores, etc?

No offense intended at all, but Conroe is still squarely roughing up AMD's fleet even at the entry level, let alone the X6800.

[Carfax]

I don't see anything indicating that Conroe is losing any footing versus the FX-60/62.

Where are you basing this off of - did a new comparison emerge with lower scores, etc?

No offense intended at all, but Conroe is still squarely roughing up AMD's fleet even at the entry level, let alone the X6800.

The review which springs most to mind which made me think about this, was HKEPC's latest review which pits Conroe 6800 Extreme vs the FX-62.

Read the review here (http://www.hkepc.com/hwdb/x6800vsfx62-10.htm)

It even has Crossfire results, which when compared to what we saw from Anandtech and the rest, had seemingly diminished.

I'm not EXACTLY SURE ofcourse, but it just seems that way.

Some people at Aces also speculated on this, so I'm not the only one who's noticed.

Regardless, Conroe is still an awesome CPU and I will be ordering one myself.

[Theta]

I see your point, but the gains are still very high.

Perhaps these results are 10% off of some other reviews (Anandtech, TH, etc.) but I think we can classify that as a margin of error.

I really highly doubt that SSE4 has impacted the results. I may be completely off base, though. :p:

[SepheronX]

I personaly dont trust a site that doesnt actualy have any english to it. I rather base my opinion when i see a proper review done in english. I dont know what they are saying!

[andyisc00l]

I've noticed that too, some sites show conroe laying it down and some are just kind of laying it down..veryyy strange.

[SepheronX]

yeah, i get a little curious sometimes. But with user benchmarks here, it shows that Conroe is just wasting AMD, but we need actual game benchmarks, wich we can easily do if you have the right programs.

I cant wait for conroes release, im getting ready to buy one. I have lots of faith, since its a built off of the Pentium M architecture (the Core and Core Duo) wich actualy performes either on par or greater with AMD clock for clock wise, so this has to be better!

[M1nDH4cK]

Carfax, i think it's just you. There's no way that SSE4 can influence game benchies in such a manner ;)

[Theta]

Carfax, i think it's just you. There's no way that SSE4 can influence game benchies in such a manner ;)

To build on that thought, I don't think games lacking SSE4 support (everything currently out) would have any change when messing around with SSE4.

Maybe drivers could be tweaked, but I just don't see how SSE4 could even enter the game benching equation.

[Thorburn]

The review which springs most to mind which made me think about this, was HKEPC's latest review which pits Conroe 6800 Extreme vs the FX-62.

Read the review here (http://www.hkepc.com/hwdb/x6800vsfx62-10.htm)

It even has Crossfire results, which when compared to what we saw from Anandtech and the rest, had seemingly diminished.

I fail to see your point.
Anandtech and HKEPC use completely different demos on Crossfire and 7800GTX v.s. X1900 Crossfire with both at different resolutions is hardly comparible.

[Carfax]

Carfax, i think it's just you. There's no way that SSE4 can influence game benchies in such a manner ;)

How do you know? Do you know the exact nature of the new SSE4 instructions?

Anandtech and HKEPC use completely different demos on Crossfire and 7800GTX v.s. X1900 Crossfire with both at different resolutions is hardly comparible.

In the first previews, Intel sent out setups which used CrossFire, at similar resolutions. The demos were different I grant you, but the delta between the 2.66ghz Conroe and the 2.8ghz FX-62 using DDR-400 was rather hefty.

I guess everyone thought that the Conroe XE would just increase that lead even more, but it doesn't seem that way.

The PCper (http://www.pcper.com/article.php?aid=265&type=expert) review also had similar scores to the HKEPC review.

Like I said, maybe I'm just seeing things, but others have picked up on it aswell.

The reason could be as you said though, different demos.

[duploxxx]

if you don't program to use those instruction sets, there just not used... as simple as is ........ so no current games will handle the sse4

[Carfax]

if you don't program to use those instruction sets, there just not used... as simple as is ........ so no current games will handle the sse4

Intel could have easily worked with ATI to optimize the drivers for SSE4..

Optimizing games for SSEn doesn't yield much in terms of performance, but optimizing graphics drivers definitely helps.

The P4 got a huge boost in performance when Nvidia optimized their drivers for SSE2.

[Vric]

Intel is cheating... been a while since I heard that one..

[Carfax]

Intel is cheating... been a while since I heard that one..

Heard my foot.. You didn't hear anything. You came to that conclusion by yourself!

[mesyn191]

yeah, i get a little curious sometimes. But with user benchmarks here, it shows that Conroe is just wasting AMD

Except for one or 2 in game benches here and there most of the tests done on this site are with synthetic benches which have almost always favored Intel and were never worth a damn anyways....

I cant wait for conroes release, im getting ready to buy one. I have lots of faith, since its a built off of the Pentium M architecture (the Core and Core Duo) wich actualy performes either on par or greater with AMD clock for clock wise, so this has to be better!

Thats a silly reason, its got about has much in common with the PM as it does the P4...

Only reason I'm buying it as Intel has already stated it'll have a general performance IPC advantage over the A64 of roughly 20% and with the overclocks that people have been getting with it I'm expecting to get to around 3Ghz at least on air and 3.5Ghz or so on phase with the E6600.

[grimREEFER]

holy crap! it looks like the difference between the p4 and fx62 is less than the difference between the fx62 and core 2 6800....
few months ago, ppl considered the 965 a crappy gaming cpu...buying an fx-62 now is even worse than buying netburst back then!

[savantu]

I expect Conroe optimized drivers to provide a healthy boost.Most likely with current drivers Conroe falls into the "P3/Pentium M" and its vast capabilities aren't used.

[Theta]

I expect Conroe optimized drivers to provide a healthy boost.Most likely with current drivers Conroe falls into the "P3/Pentium M" and its vast capabilities aren't used.

Honestly, then, the benchmarks will only get better with time.

Any anomaly at this point in testing is due to system / test differences. What I think we can all agree on is that pre-production Conroe is beating the :banana::banana::banana::banana: out of the FX line. Imagine what will happen when retail chips properly optimized are benched. :slapass: :slapass: :slapass:

[burningrave101]

An E6600 bests an FX-62 in gaming so to think an FX-62 is going to keep up with the X6800 is just folly thinking :). I dont know what AMD is going to do with pricing when Intel will have a ~$316-$319 processor that beats their $1,000+ FX-62 at stock speeds. Granted the E6600 will likely be priced in the $350-$400 range at launch because of availability at first and the demand but thats still a far cry from $1,000+.