Why AMD’s Performance Claims Are Losing Credibility

$Advanced Micro Devices(AMD)$ 1 year ago, AMD showed us slides where MI300 beats H100 by 30-40% on inferencing. Now we have some MLPerf benchmarks showing us that MI300 and H100 are on par in inferencing.

You wonder why nobody is thrilled? Because AMD has shown numbers last year which couldn't be confirmed by other benchmarks. They basically showed false numbers or what I suspect, they didn't show full potential of the competitor. Now they show us again such slides vs. H200. If AMD's slides are correct than the MI325 is supposed to be 2.2x as fast as MI300?? The reasoning is that H200 is 40% faster in inferencing than MI300 in MLPerf. So if AMD claims MI325 being 30% faster than H200 then it's essentially 2.2x faster than MI300. AMD doesn't show any such slides as they have very cherry picked slides.

What AMD misses to do is, they should stop showing relative performance comparison but simply do some real world benchmarks. Take your GPU server and run any common benchmark and post the real result of MI300 and MI325 so people can see the direct improvement. Comparison with Nvidia are stupid as reality has shown. AMD is losing trust with slides which I personally no longer believe anymore.

What I wonder, if Nvidia shows us a slide with Blackwell being 3x faster than MI300, do we believe them then? I mean it seems everybody believes AMD slides easily despite their huge disappointment on MLPerf and them obviously showing false data last year.

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Comments

We need your insight to fill this gap

Leave a comment

1

{"i18n":{"language":"en_US"},"detailType":1,"isChannel":false,"data":{"magic":2,"id":361080609009744,"tweetId":"361080609009744","gmtCreate":1729182707544,"gmtModify":1729182720675,"author":{"id":3479274792179683,"idStr":"3479274792179683","authorId":3479274792179683,"authorIdStr":"3479274792179683","name":"jazzyco","avatar":"https://static.tigerbbs.com/d2d1609584743604ecdfa6404fa6f8a4","vip":8,"userType":8,"introduction":"运营账号","boolIsFan":false,"boolIsHead":false,"crmLevel":1,"crmLevelSwitch":0,"individualDisplayBadges":[],"fanSize":101,"starInvestorFlag":false},"themes":[],"images":[],"coverImages":[],"title":"Why AMD’s Performance Claims Are Losing Credibility","html":"<html><head></head><body><p> <a data-mention-id=\"AMD\" class=\"teditor-mention\" data-mention-name=\"Advanced Micro Devices\" href=\"https://ttm.financial/S/AMD\">$Advanced Micro Devices(AMD)$</a> <v-v data-views=\"0\"></v-v>1 year ago, AMD showed us slides where MI300 beats H100 by 30-40% on inferencing. Now we have some MLPerf benchmarks showing us that MI300 and H100 are on par in inferencing.</p><p>You wonder why nobody is thrilled? Because AMD has shown numbers last year which couldn't be confirmed by other benchmarks. They basically showed false numbers or what I suspect, they didn't show full potential of the competitor. Now they show us again such slides vs. H200. If AMD's slides are correct than the MI325 is supposed to be 2.2x as fast as MI300?? The reasoning is that H200 is 40% faster in inferencing than MI300 in MLPerf. So if AMD claims MI325 being 30% faster than H200 then it's essentially 2.2x faster than MI300. AMD doesn't show any such slides as they have very cherry picked slides.</p><p>What AMD misses to do is, they should stop showing relative performance comparison but simply do some real world benchmarks. Take your GPU server and run any common benchmark and post the real result of MI300 and MI325 so people can see the direct improvement. Comparison with Nvidia are stupid as reality has shown. AMD is losing trust with slides which I personally no longer believe anymore.</p><p>What I wonder, if Nvidia shows us a slide with Blackwell being 3x faster than MI300, do we believe them then? I mean it seems everybody believes AMD slides easily despite their huge disappointment on MLPerf and them obviously showing false data last year.</p></body></html>","htmlText":"<html><head></head><body><p> <a data-mention-id=\"AMD\" class=\"teditor-mention\" data-mention-name=\"Advanced Micro Devices\" href=\"https://ttm.financial/S/AMD\">$Advanced Micro Devices(AMD)$</a> <v-v data-views=\"0\"></v-v>1 year ago, AMD showed us slides where MI300 beats H100 by 30-40% on inferencing. Now we have some MLPerf benchmarks showing us that MI300 and H100 are on par in inferencing.</p><p>You wonder why nobody is thrilled? Because AMD has shown numbers last year which couldn't be confirmed by other benchmarks. They basically showed false numbers or what I suspect, they didn't show full potential of the competitor. Now they show us again such slides vs. H200. If AMD's slides are correct than the MI325 is supposed to be 2.2x as fast as MI300?? The reasoning is that H200 is 40% faster in inferencing than MI300 in MLPerf. So if AMD claims MI325 being 30% faster than H200 then it's essentially 2.2x faster than MI300. AMD doesn't show any such slides as they have very cherry picked slides.</p><p>What AMD misses to do is, they should stop showing relative performance comparison but simply do some real world benchmarks. Take your GPU server and run any common benchmark and post the real result of MI300 and MI325 so people can see the direct improvement. Comparison with Nvidia are stupid as reality has shown. AMD is losing trust with slides which I personally no longer believe anymore.</p><p>What I wonder, if Nvidia shows us a slide with Blackwell being 3x faster than MI300, do we believe them then? I mean it seems everybody believes AMD slides easily despite their huge disappointment on MLPerf and them obviously showing false data last year.</p></body></html>","text":"$Advanced Micro Devices(AMD)$ 1 year ago, AMD showed us slides where MI300 beats H100 by 30-40% on inferencing. Now we have some MLPerf benchmarks showing us that MI300 and H100 are on par in inferencing. You wonder why nobody is thrilled? Because AMD has shown numbers last year which couldn't be confirmed by other benchmarks. They basically showed false numbers or what I suspect, they didn't show full potential of the competitor. Now they show us again such slides vs. H200. If AMD's slides are correct than the MI325 is supposed to be 2.2x as fast as MI300?? The reasoning is that H200 is 40% faster in inferencing than MI300 in MLPerf. So if AMD claims MI325 being 30% faster than H200 then it's essentially 2.2x faster than MI300. AMD doesn't show any such slides as they have very cherry picked slides. What AMD misses to do is, they should stop showing relative performance comparison but simply do some real world benchmarks. Take your GPU server and run any common benchmark and post the real result of MI300 and MI325 so people can see the direct improvement. Comparison with Nvidia are stupid as reality has shown. AMD is losing trust with slides which I personally no longer believe anymore. What I wonder, if Nvidia shows us a slide with Blackwell being 3x faster than MI300, do we believe them then? I mean it seems everybody believes AMD slides easily despite their huge disappointment on MLPerf and them obviously showing false data last year.","highlighted":1,"essential":1,"paper":2,"likeSize":1,"commentSize":0,"repostSize":0,"favoriteSize":0,"link":"https://ttm.financial/post/361080609009744","repostId":0,"isVote":1,"tweetType":1,"viewCount":46,"commentLimit":10,"likeStatus":false,"favoriteStatus":false,"reportStatus":false,"symbols":["AMD"],"verified":2,"subType":0,"readableState":1,"langContent":"EN","currentLanguage":"EN","warmUpFlag":false,"orderFlag":false,"shareable":true,"causeOfNotShareable":"","featuresForAnalytics":[],"commentAndTweetFlag":false,"andRepostAutoSelectedFlag":false,"upFlag":false,"length":1210,"xxTargetLangEnum":"ORIG"},"commentList":[],"isCommentEnd":true,"isTiger":false,"isWeiXinMini":false,"url":"/m/post/361080609009744"}