VoodooPC's Rahul Sood responds to the Intel Benchmarks

Status
Not open for further replies.
T

ToastMaster

Limp Gawd
Joined
Nov 2, 2005
Messages
483
I think many of you will find this rather interesting:

Rahul's Blog

While I honestly do believe that Conroe will end up faster than the Athlon 64 series, Rahul digs up some rather interesting issues concerning the benchmark, and since I hadn't yet seen a post relating to this topic, I thought I'd create one.

Oh, and I'm not a AMD "!!!!!!", just an AMD fan.
 
Oh, and for commentary, to those who will say "well why would Intel rig/use old parts knowing that once the systems actually start coming out to independent testers, that they'd see the performance isn't sa great."

Well, let's see:

1) Intel knows that the best group to get out an idea of complete Conroe superiority is to the enthusiast, who will then go off about it to all his/her friends about how this insanely powerful Intel processor is coming out and going to destroy the Athlon 64.

2) By the time Conroe is shipping, AMD will have moved to AM2, will have switched to DDR2-800, and who knows what else? Thus, Intel can concievably say that their increase in performance has decreased as AMD has shipped updated processors.

3) They can also say they used what they had on-hand, although this idea doesn't hold up as well, as Intel can obviously buy the best that is available.

One other thing I'm interested in is seeing a 64-bit comparison between the two.
 
Interesting read.

Be nice to see someone replicate the testing setup with an FX-60 chip on an actual up to date rig. See how close Rahul's estimation on performance increase is.
 
Assume the PR numbers from any company are wrong until proven otherwise.

The article makes good points.

Wait and see.
 
doesn't rahul have financial interest in AMD succeeding since he sells AMD-based PCs?
 
jarthel said:
doesn't rahul have financial interest in AMD succeeding since he sells AMD-based PCs?
Click to expand...

Maybe but like the guy above wisely said don't believe things people say until proven otherwise :p
 
Benchmarks done by the company who makes the product are good for nothing but laugh. There has never been a fair benchmark done this way, ever.
 
where did bit-tech get that 247 for "Intel O/ced FX-60 640x480" from?

At AT they only showed 1280x960 or 1280x1024 benches
 
B/c those are the numbers they got running from the Intel provided boxes, but with different settings. The "stock FX-60" is from their own benches, from a previous article.
 
Looks like Intel is playing Dirty again, I can't wait for AM2 to bitchslap them down :)
 
well you can either focus on intel cheating an fx-60 by oh about 6 fps or you can focus on the overpriced thousand dollar processor getting spanked by a chip half its price
 
osalcido said:
well you can either focus on intel cheating an fx-60 by oh about 6 fps or you can focus on the overpriced thousand dollar processor getting spanked by a chip half its price
Click to expand...

1/2 it's price when it's released. AMD hasn't gone to the 65nm Process yet, so things aren't going to get cheaper. And since I like to live in the Now: 955 still getting pwn3d by the FX60 ;)
 
osalcido said:
well you can either focus on intel cheating an fx-60 by oh about 6 fps or you can focus on the overpriced thousand dollar processor getting spanked by a chip half its price
Click to expand...

If you actually compare past reviews of FX 57s running prior-generation graphics cards, and then look at what an FX 60 running a X1900XT should be capable of doing, you'll see that it's more than "about 6 fps".


You can also then take the $500 and buy an Operton 175, overclock it to close to 2.8Ghz, and thus have essentially an "overclocked FX 60" for the same cost as that Conroe.
 
Mav451 said:
http://www.bit-tech.net/news/2006/03/08/intel_conroe_benchmark_fear/

Bit-tech's numbers raise a bit of suspicion in me. Remember that Intel "overclocked" the AMD machine to "2.8Ghz". Well, don't you find it odd that the 2.8Ghz (using multiplier OC I might add) is getting lower numbers than the stock FX-60 (2.6)?
Click to expand...

Not if it's a motherboard with an outdated bios that doesn't even support the FX 60. ;)
 
I wonder if they used the AMD Dual Core driver and the Windows Hotfix?
 
We need to question what BIOS "support" really means. Does it cripple the chip or does it simply misreport or not know what CPU is installed? Just because it doesn't recognize that chip as an FX-60 doesn't mean the performance was crippled. We also don't know if the latency issue that was mentioned affects this particular setup. Rahul was also wrong in his statements that latency "massively" affects performance. This may have been the case pre-IMC but we now know it doesn't necessarily mean all that much. We have this:

http://techreport.com/etc/2005q4/mem-latency/index.x?pg=9

According to that report memory latency doesn't have much effect in UT2004. So even if the DFI board had higher latencies than reported it doesn't cover the performance difference. The UT2004 scores ARE probably lower on the Intel FX-60 than reported elsewhere. Anand states that an Intel demo was run. Intel probably picked a map or situation that reflected the Conroe in a positive light. It's analogous to the early athlon days when the athlon beat the p3 in quake2 in normal situations but beat it even more when a special demo was run called crusher which stressed the strong FPU. Intel probably did something similar but it's still an accurate representation as that situation could ACTUALLY be found in gameplay.

Another thing to look at is the FEAR performance. Judging by the difference in the speed of the athlon and Conroe the game was CPU limited at that resolution running x1900s in crossfire. Anand was able to run HIS test in FEAR which is the same test he uses to benchmark in-house. It's simply a matter of finding a previous benchmark running a similar config to see if the athlon machine was crippled.

Page A: http://www.anandtech.com/tradeshows/showdoc.aspx?i=2713&p=3

Page B: http://www.anandtech.com/video/showdoc.aspx?i=2679&p=12

Page C: http://www.pcper.com/article.php?aid=197&type=expert&pid=10

There we have it. Anand's tests show that an FX-57 (2.8 GHz) on an RD480 board gets 133 fps at a CPU limited resolution. Intel's machine has an FX-60 at 2.8GHz getting 132 fps. The third link shows that at CPU limited resolutions dual core doesn't affect FEAR performance. It appears that the Intel benchmark is spot on with what Anand gets in-house. The only real difference is that the in-house test uses 2GB of RAM with slightly looser timings.

It's time to just swallow the hard truth. The Conroe really IS faster. I know it's tough to do for some people, but it's time. I know people are under the delusion that AMD will pull something out of a hat within the next four months, but they WILL NOT be able to overcome a 20% deficit without huge modifications to the core itself. Unless they've already taped out new silicon, they simply won't have such a revision ready by the summer.

AM2 will allow the athlon64 to utilize DDR2 memory which will increase the amount of bandwidth available. However, in most situations the athlon is not starved for bandwidth.
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2469&p=5
There simply isn't much return for high bandwidth unless you are doing some heavy multitasking. Let's not forget that DDR2 will offer higher bandwidth at HIGHER latencies which will offset some of the performance improvement. I see a 5-10% improvement in the best cases. Still, not enough to match Conroe. Let's not forget that Conroe is six months away from being released. Many here will discount its performance for that reason alone. However, this almost certainly means that Conroe's performance will almost certainly improve between now and release as it moves to more mature platforms using newer silicon. Don't let company loyalty blind your logic.

The P4 was a loser in general. It was slower when it came out and it was slower when the a64 was released. Granted it was faster in many applications when 3.2C was out and was competing against Barton. However, the last couple of years have belonged to AMD. Therefore, it's no surprise that the Conroe was made not to match the Athlon, but to beat it in nearly every situation. It only makes sense, and frankly isn't much of a surprise. AMD will play catch-up for awhile now until it releases a new core that beats the Conroe. They leapfrog each other and prices drop. I win. That's all I care about really.

BTW, just want to add that I run a 3800+ and have owned nearly every flavor of the athlon since the slot A variant. I also own a 2.8C overclocked to 3.5 which wiped the floor with the t-bred I had in nearly every category, including value. I'm after value over anything.
 
Mav451 said:
http://www.bit-tech.net/news/2006/03/08/intel_conroe_benchmark_fear/

Bit-tech's numbers raise a bit of suspicion in me. Remember that Intel "overclocked" the AMD machine to "2.8Ghz". Well, don't you find it odd that the 2.8Ghz (using multiplier OC I might add) is getting lower numbers than the stock FX-60 (2.6)?
Click to expand...
That's what I thought as well.

Image for clarification:
conroe.png
 
banGerprawN said:
That's what I thought as well.

Image for clarification:
conroe.png
Click to expand...

However, the reference FX-60 results obtained in that graph come from using a different vid card. It adds another variable which might be skewing the results. See my post above for comparisons using the same vid card setup.
 
Yeah, but the thing is, even on the link to the reference "Stock" FX-60, the FX-57 (2.8) only get 283FPS...still quite a bit to go to catch up to 346.
 
I tend to agree with others, the Conroe will probably be faster, I've said this previously. I doubt the difference will be that large once this summer comes along and it is never as large of a difference as whichever company shows it to be in their own benches.

Bottom line, Intel is back in the game, finally got off its Netburst architectures, lets see how AMD answers.

His final comment in that blog was something about it being an interesting year, CPU wars do sell processors.

Akin to the PIII/Thunderbird war, the game is one, we all win.
 
seems everyone is trying to milk a dead horse.

everyone's looking way too hard to try and find something fishy with the setups that were benchmarked. unfortunately though, when it comes right down to it, even if there was something intel could have done better with the AMD system, performance really wouldn't have changed that much. look at the older anand articles being pointed out for example, the benchmarks are pretty much on the money.

everyone should read what BIGDH01 wrote on the first page. that pretty much sums it up.
 
BigDH01 said:
However, the reference FX-60 results obtained in that graph come from using a different vid card. It adds another variable which might be skewing the results. See my post above for comparisons using the same vid card setup.
Click to expand...
A different video card than X1900XTXs in CrossFire? The only "different" that I can imagine would probably be "worse". If so, wouldn't that make the reference FX60 using a "different/worse" video card theoretically perform worse in both GPU and CPU limited games? Which appears to not be the case.
Now I'm all confused, lawlz.
 
Different time demoes anyone ? I hope you do realize that in some places you have 60fps and in other over 300 fps. Average score is heavily influenced about what part of the game you pick as demo. Comparing different tests of different platforms on different timedemoes is plain silly.

If Intel had used different time demoes on the platforms than yes , Intel would be a bitch , but reality is AMD lost.End of story.
 
savantu said:
Different time demoes anyone ? I hope you do realize that in some places you have 60fps and in other over 300 fps. Average score is heavily influenced about what part of the game you pick as demo. Comparing different tests of different platforms on different timedemoes is plain silly.

If Intel had used different time demoes on the platforms than yes , Intel would be a bitch , but reality is AMD lost.End of story.
Click to expand...
AT has updated benchies, here:
http://www.anandtech.com/tradeshows/showdoc.aspx?i=2716&p=1
 
Status
Not open for further replies.
Back
Top